![Applied Energy 287 (2021) 116563
2
the participation of end-users in demand response programmes is projected by running uniform price demand
response auctions. The most salient distinction of the proposed two-stage (wholesale and retail) demand-side
management market model is the continual process of trading, with incentive prices unique to each transaction.
Nomenclature
Indices and sets
c ∈ C = {PV,WT,MH,T,E,FC,HT,BP,B,SC,S,FCEV2G} micro-grid
components, the optimal size of which is under
investigation: photovoltaic panels (PV), wind turbines
(WT), micro-hydro turbines (MH), transformer (T),
electrolyser (E), fuel cell (FC), hydrogen tank (HT),
biopower plant (BP), battery bank (B), super-capacitor
bank (SC), hydrogen station (S), and fuel cell electric
vehicle-to-grid unit (FCEV2G)
d ∈ D = {1,2,⋯,365} day of the year-round micro-grid operation
D*
= {d*
1,d*
2,⋯,d*
K} set of best-response load reductions contributed
by all the customers
D
j,*
LA = {dk,j,*
,d− k,j,*
} set of best-response strategies of all the
customers signed up with the j-th aggregator
es ∈ ES = {B,SC,HT,FCEV} energy storage media: battery bank (B),
super-capacitor bank (SC), hydrogen tank (HT), and
aggregated fuel cell electric vehicles’ tanks (FCEV)
I*
LA = {I1,*
LA ,I2,*
LA ,I3,*
LA ,I4,*
LA ,I5,*
LA } set of best-response incentive payments
of aggregators
j ∈ J responsive load aggregators
k ∈ NJ customers enrolled with aggregator j
K set of all the micro-grid’s customers
p ∈ Pd⊂T day-specific peak consumption hour
t ∈ T = {1,2,⋯,8760} time-step of the year-round micro-grid
operation
Parameters
c
k,j
1 discomfort tolerance coefficient of customer k of
aggregator j [$/kWh2
]
c
k,j
2 discomfort tolerance coefficient of customer k of
aggregator j [$/kWh]
C900, C115, C33 nameplate capacities of inverters [kW]
CV gross calorific value of biomass feedstock [kWh/kg]
δj load type-dependent demand response procurement factor;
sectoral elasticity of customer-supplied demand response
capacity
Δt time-step length [h]
d
k,j
cr (t), d
k,j
ncr(t) critical/non-critical portion of the load power
demanded by the k-th customer subscribed to aggregator j
at time-step t [kWh]
d
k,j
full(t) full load power demanded by the k-th customer subscribed
to aggregator j at time-step t [kWh]
disk,j,min
, disk,j,max
lower/upper limit of the discomfort cost imposed
on the k-th customer of aggregator j [$]
DF derating factor of the PV module [%]
ξCO2
social cost of CO2 emissions [$/tCO2]
ηes round-trip efficiency of storage medium es [%]
ηFCEV2G efficiency of the operation of the fuel cell electric vehicles
in the vehicle-to-grid mode [%]
ηPV, ηMH, ηBP, ηT, ηI, ηSC, ηB, ηE, ηHT, ηFC, ηS efficiency of the PV plant/
micro-hydro plant/biopower plant/transformer/inverter/
super-capacitor/battery/electrolyser/hydrogen tank/fuel
cell/hydrogen station [%]
ηPV,DC/DC, ηMH,AC/DC, ηBP,AC/DC PV plant’s DC/DC converter efficiency,
micro-hydro power plant’s AC/DC converter efficiency,
biopower plant’s AC/DC converter efficiency [%]
ECO2
CO2 emission factor of the biopower plant [kg-CO2/kg-
feedstock]
Ees,min, Ees,max lower/upper capacity limit of storage medium es
[kWh]
F(t) river streamflow rate at time-step t [m3
/s]
g acceleration of gravity [m/s2
]
h wind turbine hub height [m]
hg micro-hydro turbine gross head [m]
href reference height of wind speed records [m/s]
HHVH2
higher heating value of hydrogen [kWh/kg]
iMGO step size for the micro-grid operator-determined incentive
[$/kWh]
IG(t) global solar irradiance on the horizontal surface at time-
step t [kW/m2
]
I
j,min
LA , I
j,max
LA lower/upper limit of the incentives determined by
aggregator j [$/kWh]
Imin
MGO, Imax
MGO lower/upper limit of the micro-grid operator-offered
incentives [$/kWh]
Iref reference solar irradiance [kW/m2
]
Itermax maximum number of iterations
K DC gain of the transfer function
Kp PV module’s temperature coefficient [%/◦
C]
LPSPmax
e , LPSPmax
H2
maximum allowable loss of power supply
probability in supplying electricity/hydrogen [%]
MBD(t) biomass feedstock mass consumption rate at time-step t
[kg/h]
N
j
cust number of customers enrolled with aggregator j
Nmax
c upper limit of the size (capacity/quantity) of component c
NSA number of search agents of the optimisation algorithm
NMOT nominal PV module operating temperature [◦
C]
ρ water density [kg/m3
]
Pch,max
es , Pdch,max
es upper limit of the charging/discharging rate of
storage medium es [kW]
Pch,min
es , Pdch,min
es lower limit of the charging/discharging rate of
storage medium es [kW]
Pmax
FCEV2G(t) maximum V2G power at time-step t [kW]
PFC,r,PE,r rated capacity of each fuel cell/electrolyser stack [kW]
PL(t) load power demand at time-step t [kW]
PL,max maximum electrical load on the micro-grid [kW]
PMH,r, PBP,r rated capacity of each micro-hydro turbine/biopower
plant [kW]
PPV(t), PWT(t), PMH(t), PBP(t) power output from the PV/wind
turbine/micro-hydro/biopower plant at time-step t [kW]
PPV,r rated capacity of the PV module under standard test
conditions [kW]
PS(t) hydrogen power demand of the station at time-step t [kW]
penconst penalty term added to the life-cycle cost function where
constraints are not met [$]
πex, πim(t) per-unit income from electrical energy exports [$/kWh],
per-unit cost of electrical energy imports at time-step t
[$/kWh]
πFCEV2G per-unit premium tariff rate for V2G power [$/kWh]
Q quality factor of the low-pass energy filter
tup minimum up-time of the electrolyser, fuel cell, and
biopower plant [h]
Ta(t) ambient temperature at time-step t [◦
C]
Tm(t) PV module temperature at time-step t [◦
C]
S. Mohseni et al.](https://image.slidesharecdn.com/mohseni2021-211027145406/85/Mohseni2021-2-320.jpg)
![Applied Energy 287 (2021) 116563
3
1. Introduction
One of the principal advantages of making the electricity grid
“smart” is that it enables consumers to proactively engage in electricity
markets and benefit from demand-side management (DSM) schemes
designed and incentivised by utilities to curtail/interrupt or shift a
proportion of electricity demand, and thereby flatten the load power
profile–and improve the load factor. While demand response (DR) pro
grammes have been in use to improve the energy efficiency of industrial
consumers for years, the expansion of the concept to include less energy-
TSTC PV module temperature under standard test conditions
[◦
C]
Vh normalised wind speed profile to the wind turbine hub
height [m/s]
Vref reference wind speed profile [m/s]
ω0 cut-off frequency [dB]
γ wind shear exponent
Variables
costem(t) total penalties imposed for emissions at time-step t [$]
costFCEV2G(t) cost associated with the FCEV2G operations at time-
step t [$]
costim(t) cost of electricity import at time-step t [$]
dk,j
(t) load reduction contributed by the k-th customer of
aggregator j at time-step t [kWh]
dk,j,*
(t) best-response strategy taken by the k-th customer
subscribed to the j-th aggregator for load reduction at time-
step t [kWh]
Ddef (t) capacity deficit to meet the loads at time-step t [kWh]
D
j
LA(t) load reduction contributed by aggregator j at time-step t
[kWh]
disk,j
(t) discomfort cost imposed on the k-th customer subscribed to
aggregator j at time-step t [$]
Ees(t) energy content of storage medium es at time-step t [kWh]
ESC(t), EB(t), EHT(t) energy content of the super-capacitor/battery
bank/hydrogen tank at time-step t [kWh]
I
j
LA(t) incentive payment offered by aggregator j for load
reduction at time-step t [$/kWh]
I
j,*
LA(t) best-response incentive payment for load reduction offered
by aggregator j at time-step t [$/kWh]
IMGO(t) rate of micro-grid operator-posted incentive payments for
load reduction at time-step t [$/kWh]
I*
MGO(t) globally-optimum incentive payment for load reduction
offered by the MG operator at time-step t [$/kWh]
incomeex(t) income from electricity export at time-step t [$]
LPSPe, LPSPH2
loss of power supply probability in supplying
electricity/hydrogen [%]
mHT(t) mass of hydrogen stored in the tank at time-step t [kg]
NB optimal capacity of the overall battery bank [kWh]
NFCEV2G optimal capacity of the fuel cell electric vehicle-to-grid
system [kW]
NHT optimal capacity of the hydrogen tank [kg]
NI optimal capacity of the electrical loads’ overall power
inversion system [kW]
NPV, NWT, NMH, NBP, NE, NFC, NSC optimal quantity of PV modules/
wind turbines/micro-hydro turbines/biopower units/
electrolyser stacks/fuel cell stacks/super-capacitor
modules
NS optimal capacity of the hydrogen refuelling station [kg-H2/
h]
NT optimal capacity of the transformer [kVA]
N900, N115, N33 optimal quantity of 900-kW/115-kW/33-kW
inverters
N1600, N400, N100 optimal quantity of 1600-kWh/400-kWh/100-
kWh battery packs
NPCc net present cost of component c [$]
NPCI net present cost of the inverter [$]
OCMG(t) operational cost of offsetting power deficit at time-step t
[$]
OC*
MG(t) globally-optimum operational cost of the micro-grid to
address the shortage of power generation capacity at time-
step t [$]
Pch(t), Pdch(t) total charging/discharging power of the hybrid
battery/super-capacitor storage system at time-step t [kW]
Pch,HF2, Pdch,HF2 charging/discharging power of the super-capacitor
bank [kW]
Pch,LF2, Pdch,LF2 charging/discharging power of the battery bank
[kW]
PE(t) power consumed by the electrolyser at time-step t [kW]
PE− HT(t) hydrogen power directed from the electrolyser to the
hydrogen tank at time-step t [kW]
Pch
es (t), Pdch
es (t) charging/discharging rate of energy storage medium
es at time-step t [kW]
PFC(t) power generated by the fuel cell at time-step t [kW]
PFCEV2G(t) aggregated vehicle-to-grid power provided by fuel cell
electric vehicles at time-step t [kW]
PHT− FC(t) hydrogen power directed from the hydrogen tank to the
fuel cell at time-step t [kW]
PHT− S(t) hydrogen power directed from the hydrogen tank to the
station at time-step t [kW]
Pim(t), Pex(t) imported/exported electricity at time-step t [kW]
PSH(t), PEX(t) shortage/excess of renewable power generation at
time-step t [kW]
PSH− LF1(t), PSH− HF1(t) low-/high-frequency component of the
renewable power shortage signal at the first low-pass filter
output at time-step t [kW]
PSH− LF2(t), PSH− HF2(t) low-/high-frequency component of the
renewable power shortage signal at the second low-pass
filter output at time-step t [kW]
PEX− LF1(t), PEX− HF1(t) low-/high-frequency component of the
renewable power excess signal at the first low-pass filter
output at time-step t [kW]
PEX− LF2(t), PEX− HF2(t) low-/high-frequency component of the
renewable power excess signal at the second low-pass filter
output at time-step t [kW]
Pr
j
LA(t) profit gained by aggregator j at time-step t [$]
QL(t), QH2
(t) unmet electrical/hydrogen load demand at time-step t
[kW]
Uk,j
(t) utility of the customer k serviced by aggregator j at time-
step t [$]
Functions
H(s) low-pass energy filter transfer function
NPV
20− yr
(z) net present value of cost component z over the 20-year life
of the project [$]
NPCc
20− yr
net present cost of micro-grid component c over the 20-
year life of the project [$]
NPCI
20− yr
net present cost of the overall power inverter over the 20-
year life of the project [$]
OCMG hourly operational cost function of the micro-grid [$]
⌊⋅⌋ floor function
⌈⋅⌉ ceiling function
S. Mohseni et al.](https://image.slidesharecdn.com/mohseni2021-211027145406/85/Mohseni2021-3-320.jpg)
![Applied Energy 287 (2021) 116563
4
dense demand sectors, namely the residential, agricultural, and com
mercial sectors, as well as electrified transport, is enabled by recent
advancements in information and communications technology (ICT),
which have substantially contributed to the development of advanced
metering infrastructure [1,2]. Recent studies have revealed that the
consideration of DSM strategies in the optimum investment planning
phase of renewable and sustainable energy systems (RSESs) for domestic
applications can offer cost savings of about 15% to nearly 35%
(depending on the participation rate of end-users in the DR pro
grammes), whilst preserving consumer comfort standards [3,4,5]. That
is, the proper integration of DR programmes into RSESs would result in a
win–win–win situation–the third winner being the environment, as they
will accelerate the transition to a low-carbon energy economy and a
world run on green energy.
1.1. Long-term, demand response-integrated micro-grid infrastructure
planning background
A reformed formulation of the micro-grid (MG) equipment capacity-
planning problem is required to make effective use of the economic
opportunities offered by DSM processes to support decision-making in
developing cost-effective RSESs [6]. A solution to the optimal DR-
integrated MG design problem identifies the least-cost combination of
the size of the components of the system over a decades-long–often
spanning 20–30 years–investment planning horizon to meet the pro
jected demand for energy, while leveraging the potential of responsive
loads [7,8].
Recent review studies have focused on discussing methods and
trends for harvesting the potential of the demand-side flexibility to
contribute significantly to energy affordability in energy networks with
a high penetration of distributed renewables. Gelazanskas and Gamage
[9], Haider et al. [10], Esther and Kumar [11], Wang et al. [12], Robert
et al. [13], as well as, more recently, Jordehi [14] have scrutinised
various approaches to implementing DR arrangements, while optimally
designing RSESs, with a particular focus on residential DR resources.
Moreover, various types of DSM strategies have been incorporated in the
formulation of the MG capacity-optimisation models. This implies that
DR programmes are well-analysed for the planning of RSESs, a state
ment that has likewise been made in the context of different DSM
business models in electricity markets [15,16], as well as for the optimal
operational scheduling (energy management) of RSESs [12].
There have also been attempts to exploit other types of DR structures
for the optimal capacity planning of RSESs. For instance, Kahrobaee
et al. [17] devised a particle swarm optimisation (PSO)-based planning
model for a smart home nano-grid that utilises the real-time pricing
(RTP) scheme, which allows for leveraging the historical records of the
price elasticity of demand for personalised dynamic pricing. In another
instance, Yu et al. [18] proposed a robust flexible-programming
approach for the integration of renewables into a municipal energy
system, which runs a critical peak pricing (CPP) rate structure. More
over, Varasteh et al. [19] employ a hybrid direct load control-time-of-
use (DLC-ToU) DR framework to drive down the whole-life cost of a
grid-tied combined heat and power (CHP) MG.
In addition, some studies have explored the potential of vehicle-to-
grid (V2G) technologies and electric vehicle (EV) charging/discharg
ing coordination through DSM mechanisms in driving economic sus
tainability improvement for renewable energy development projects.
For instance, Cardoso et al. [20] have proposed a DLC decision model for
the aggregated energy scheduling of EVs and demonstrated its distinc
tive contribution to reducing the lifetime cost of a multiple energy
carrier MG, while considering the uncertainty associated with the EV
driving schedules. In another instance, Hosseinnia et al. [21] have
provided further evidence of the utility and economic benefits of EV fleet
trip level energy management and V2G connectivity in the context of
sustainable energy system design and planning. Moghaddas-Tafreshi
et al. [22] have also underlined the potential of optimal charging/
discharging scheduling of plug-in hybrid EVs in improving the profit
ability of an energy hub and reaping cost-savings for vehicle owners,
while addressing the uncertainty associated with the power consump
tion of vehicles during trips. Table 1 summarises the most rigorous
studies carried out to date on the integration of demand-side resources
(for the strategic planning of energy demand) in the long-term capacity
optimisation models of RSESs (listed in ascending order of publication
date), whilst additionally situating this study in the context of the
existing literature.
1.2. Demand response-integrated life-cycle planning of micro-grids:
knowledge gaps and proposition
As Table 1 indicates, there is a growing body of literature lending
support to the integration of DSM frameworks into the design phase of
RSESs. However, as far as can be ascertained, no single study has eval
uated the attitude of neither end-users nor electricity providers in
relation to adopting these practices during the optimal design and
planning process of RSESs. Accordingly, oversimplified assumptions
have commonly been made in the literature regarding the available
capacity of responsive loads, which have substantially reduced the ac
curacy of projections. That is, many hypotheses regarding the degree of
end-users’ participation in the DR schemes are not well-grounded. To
aid the associated asset-allocation decision-making procedure, a long-
term, DR-integrated MG investment planning approach needs to
model the involvement of aggregator-mediated customers in the DR
programmes in a systematic, market-driven approach. The market-
driven approach needs to capture the dynamic nature of strategic in
teractions between rational, utility-maximising active economic agents
in an aggregator-mediated DSM market. More specifically, the approach
needs to identify the reaction and commitment of different classes of
customers mediated by third-party demand response aggregators
(DRAs), when exposed to variations in the economic incentives for load
curtailment/shifting. In this context, the DRAs round up parcels of
interruptible loads to enable them to reach the sufficient scale required
for selling services to the system operator(s) [43,44,45]. In addition,
more work is needed to evaluate the effect of different levels of
discomfort experienced by different customer classes on the economic
feasibility of renewable energy projects as the characterisation of
aggregator-mediated customer comfort constraints during the planning
phases of RSESs is less well explored. To assist decision-makers in
designing cost-optimal sustainable energy systems consistent with the
expectations of their customers, it is critically important to devise ac
curate models aimed at reflecting user values and preferences (which
furnish the basis for service flexibility) in the design of MG projects. This
brings to light the need for an investment decision-making framework
that accommodates end-users’ preferences (which could be derived from
their energy service needs and the relative values they place on them)
within the long-term MG capital-investment plans.
1.3. Objective
The main objective of the paper is to demonstrate the potential of
aggregator-mediated, incentive-based, market-driven DSM programmes
tailored to small- to medium-scale end-consumers in improving the
economic viability of community-scale MG systems. Accordingly, the
paper expands the boundaries of knowledge and understanding of the
positive impacts of altering energy consumption behaviour of different
types of electrical loads–through effective incentive-based DR pro
grammes–on the cost-optimal design of MGs. Also, a secondary objective
of the paper is to ascertain the technological competence and cost-
competitiveness of utilising hydrogen as an energy vector in
community-scale MGs for niche applications–inter-seasonal energy
storage to meet seasonal demand, and hydrogen mobility to decarbonise
the transport sector.
More specifically, the paper contributes to the trend of the
S. Mohseni et al.](https://image.slidesharecdn.com/mohseni2021-211027145406/85/Mohseni2021-4-320.jpg)
![Applied
Energy
287
(2021)
116563
5
Table 1
Summary of the studies on the provision of the DSM procurements in the long-term investment planning of the RSESs.
Reference Test-case system
configuration
DSM
scheme
Flexible loads V2G
capabilities
DR-inherent
uncertainties
Multi-temporal
reserve
procurement
Aggregator-mediated
customer comfort
characterisation
Modelling
approach
Objective(s) Solution
algorithm
Case study area
Martins and
Borges, 2011
[23]
A typical active
distribution grid with
a high share of
renewables
ICSs Unspecified × × × × Stochastic LCCM GA A typical
distribution
network in
Brazil
Kahrobaee et al.,
2013 [17]
A grid-connected WT/
BESS nano-grid
RTP SRAs × × × × Stochastic LCCM PSO A typical house
in the U.S.
Cardoso et al.,
2014 [20]
A grid-connected PV/
ST/ICE/MT/GT/FC/
BESS/AC MG
DLC EV-charging √ × × × Stochastic LCCM DER-CAM
tool
San Francisco,
CA, U.S.
Zhu et al., 2015
[24]
An off-grid PV/WT/
BESS/DG MG
DLC HVAC × × × × Deterministic LCCM NP Shanghai, China
Atia and Yamada,
2016 [25]
A grid-tied PV/WT/
BESS MG
DLC SRAs and EV-charging × × × × Stochastic LCCM MILP Okinawa, Japan
Pazouki and
Haghifam, 2016
[26]
A WT/MCHP/boiler/
BESS/TESS energy
hub
DLC Unspecified × × × × Stochastic LCCM MILP Unspecified
Schachter et al.,
2016 [27]
A typical smart
distribution grid with
deep renewable
penetration
DLC Unspecified × √ × × Stochastic LCCM SDM Unspecified
Yu et al., 2017
[18]
A WT/PV/BP/coal/
gas municipal energy
system
CPP Unspecified × × × × Stochastic LCCM RFP Qingdao, China
Chauhan and
Saini, 2017 [28]
A stand-alone PV/
WT/BESS/DG/BP/
MHPP MG
DLC Smart appliances of
the residential,
commercial,
agricultural, and
community sectors
× × × × Deterministic LCCM DHS Chamoli, India
Nojavan et al.,
2017 [29]
The standard IEEE 33-
bus distribution
network
ToU Unspecified × × × × Stochastic LCCM and
reliability
maximisation
MINLP Unspecified
Amrollahi and
Bathaee, 2017
[30]
A grid-connected PV/
WT/BESS
DLC Unspecified × × × × Deterministic LCCM MILP An unnamed
forestry camp,
northwest of
Iran
Chen et al., 2018
[31]
A grid-tied PV/WT/
BESS MG
DLC SRAs and HVAC × × × × Stochastic LCCM and
reliability
maximisation
MILP Unspecified
Zheng et al., 2018
[32]
A grid-tied PV/BP/
boiler MG
ToU Unspecified × × × × Stochastic LCCM LP Davis, CA, U.S.
Xiao et al., 2018
[33]
A modified IEEE 33-
bus distribution
network with deep
penetration of
renewables
Hybrid
DLC-
ICSs
Unspecified thermal
and electrical loads
× × × × Stochastic LCCM MBGO Unspecified
Husein and Chung,
2018 [34]
An on-grid PV/WT/
MHPP/GPP/BP MG
ToU Unspecified × √ × × Stochastic LCCM ESM Seoul, South
Korea
Gazijahani and
Salehi, 2018
[35]
A modified IEEE 33-
bus distribution
network with high
penetration of
renewables
CPP Unspecified √ × × × Deterministic LCCM RMILP Unspecified
(continued on next page)
S.
Mohseni
et
al.](https://image.slidesharecdn.com/mohseni2021-211027145406/85/Mohseni2021-5-320.jpg)
![Applied
Energy
287
(2021)
116563
6
Table 1 (continued)
Reference Test-case system
configuration
DSM
scheme
Flexible loads V2G
capabilities
DR-inherent
uncertainties
Multi-temporal
reserve
procurement
Aggregator-mediated
customer comfort
characterisation
Modelling
approach
Objective(s) Solution
algorithm
Case study area
Amir et al., 2018
[36]
A grid-connected PV/
BESS/boiler/TESS/
MCHP MG
ToU Unspecified × × × × Deterministic LCCM GA Unspecified
Mohseni et al.,
2018 [37]
An off-grid PV/WT/
battery MG
DLC EV-charging × × × × Deterministic LCCM GA Kish Island, Iran
Nazari and
Keypour, 2019
[38]
An on-grid PV/WT/
BESS/MT MG
DLC SRAs, PWS, and HVAC × × × × Stochastic LCCM MILP Unspecified
Prathapaneni and
Detroja, 2019
[39]
A stand-alone PV/
BESS/DG MG
DLC EV-charging and PWS × × × × Stochastic LCCM MINLP Hyderabad,
India
Hosseinnia et al.,
2019 [21]
An on-grid PV/WT/
BESS/boiler/MCHP/
TESS MG
ToU SRAs √ × × × Stochastic LCCM and
GHGEM
TSA Unspecified
Bhamidi and
Sivasubramani,
2019 [40]
A grid-connected PV/
WT/BESS/MT/DG MG
ToU SRAs and EV-charging × × × × Deterministic LCCM and
GHGEM
NSGA-II San Angelo, TX,
U.S.
Varasteh et al.,
2019 [19]
A multi-carrier PV/
WT/CCHP/boiler/
BESS MG
Hybrid
DLC-
ToU
SRAs × × × × Deterministic LCCM MINLP Unspecified
Mohseni et al.,
2019 [41]
A grid-independent
PV/WT/BESS MG
DLC SRAs and EV-charging × × × × Deterministic LCCM MFOA Hengam Island,
Iran
Amir and Azimian,
2020 [8]
A grid-connected PV/
MCHP/BESS/TESS
multiple energy
carrier MG
Hybrid
DLC-
ToU
Unspecified × × × × Stochastic LCCM GA-MINLP Unspecified
Salyani et al., 2020
[42]
The standard IEEE 33-
bus distribution
network
RTP EV-charging √ √ × × Stochastic LCCM and
GHGEM
MINLP Unspecified
This study A grid-tied PV/WT/
MHPP/BP/FC/BESS/
SC MG
ICSs SRAs and FCEVs √ √ √ √ Deterministic LCCM MFOA Ohakune, New
Zealand
Key: AC = Absorption Chiller, BESS = Battery Energy Storage System, BP = Biopower Plant, CA = California state, CPP = Critical Peak Pricing, CCHP = Combined Cooling, Heating, and Power, DER-CAM = Distributed
Energy Resources-Customer Adoption Model, DG = Diesel Generator, DHS = Discrete Harmony Search, DLC = Direct Load Control, ESM = Enumeration Search Method, EV = Electric Vehicle, FC = Fuel Cell, FCEV = Fuel
Cell Electric Vehicle, GA = Genetic Algorithm, GHGEM = Greenhouse Gas Emissions Minimisation, GPP = Geothermal Power Plant, GT = Gas Turbine, HVAC = Heating, Ventilation, and Air Conditioning, ICE = Internal
Combustion Engine, ICSs = Interruptible/Curtailable Services, LCCM = Life-Cycle Cost Minimisation, LP = Linear Programming, MBGO = Metamodel-Based Global Optimisation, MCHP = Micro-Combined Heat and
Power, MG = Micro-Grid, MHPP = Micro-Hydro Power Plant, MILP = Mixed-Integer Linear Programming, MINLP = Mixed-Integer Nonlinear Programming, MFOA = Moth-Flame Optimisation Algorithm, MT = Micro-
Turbine, NP = Nonlinear Programming, NSGA-II = Non-dominated Sorting Genetic Algorithm II, PSO = Particle Swarm Optimisation, PV = Photovoltaic, PWS = Pumped Water Storage, RFP = Robust Flexible Pro
gramming, RMILP = Robust MILP, RTP = Real-Time Pricing, SC = Super-Capacitor, SDM = Supply-Demand Matching, SRAs = Smart Residential Appliances, ST = Solar Thermal, TESS = Thermal Energy Storage System,
ToU = Time-of-Use, TSA = Tabu Search Algorithm, TX = Texas state, WT = Wind Turbine.
S.
Mohseni
et
al.](https://image.slidesharecdn.com/mohseni2021-211027145406/85/Mohseni2021-6-320.jpg)
![Applied Energy 287 (2021) 116563
7
conservation of energy through procuring DSM provisions for the stra
tegic decision-making related to the optimal mix of distributed energy
resources (DERs) to be integrated into RSESs–which is discussed in the
literature review in Section 1.1. Accordingly, this study puts forward a
novel long-term, comfort-preserving MG equipment capacity-planning
decision-making framework that offers a new solution to fill the litera
ture voids identified in Section 1.2. Notably, this paper makes the
following key contributions:
• The strategic interactions between the MG operator (utility), mo
nopoly DRAs, and end-consumers are characterised using an equi
table market model for DR aggregation in community-scale
renewable energy projects using tools borrowed from non-
cooperative game theory [46] and the endogenous Stackelberg
leader–follower relationships1
[47]. The proposed DSM market
model is designed on the basis of interruptible DR programmes and
accounts for the elasticity of customer-supplied DR capacity (load
type-dependent DR procurement factor).
• The proposed DSM market design is integrated into a standard model
of long-term, meta-heuristic-based capacity planning of grid-
connected MGs to elucidate the contributions of more accurate DR
resource projections in improving the economic viability of MG
development projects.
• A novel hydrogen-based MG system is conceptualised, which is the
first to capture the potential of the fuel cell electric vehicles in
vehicle-to-grid operation (FCEV2G) technology in improving the
dispatchability of 100%-renewable MG systems and, in turn,
ensuring the economic sustainability of strategic MG investment
planning decisions.
• The application of the energy filter-based approach to scheduling
energy storage infrastructure is expanded to multiple energy storage
technologies, namely: hydrogen storage, vanadium redox flow bat
teries, and super-capacitors (SCs). This provides a platform to more
efficiently address the intermittency of renewables by economically
dispatching different backup systems running at various temporal
resolutions, namely: seasonal, inter- and intra-day, and transient.
1.4. Structure of paper
The rest of this paper is organised as follows. Section 2 mathemati
cally defines the conceptualised stand-alone, multi-energy-storage-
technology MG architecture employed as a test-case to evaluate the
utility and effectiveness of the proposed two-stage market-driven DSM
business model. The proposed interruptible DR scheduling framework is
presented in Section 3. Section 4 integrates the proposed DSM frame
work into a standard meta-heuristic-based MG capacity planning model.
A case study analysis is carried out in Section 5. Finally, conclusions are
made in Section 6.
A schematic outline of the paper, which illustrates the steps followed
in this study and their interconnectedness, is set out in Fig. 1.
2. Test-case micro-grid system
The conceptualised grid-connected, DC-coupled, multiple energy
carrier MG test-case system (see Fig. 2) is envisioned to supply green
power and transportation fuel to communities residing in the vicinity of,
or within relatively short distances from, the main power grids. Also, it
serves five different categories of energy demand: (1) residential, (2)
agricultural, (3) commercial, and (4) industrial load power demands, as
well as (5) the demand for hydrogen (through dedicated hydrogen
refuelling infrastructure) from fuel cell electric vehicles (FCEVs). The
test-case is used to verify the effectiveness of the proposed DR-integrated
energy planning optimisation model.
2.1. Micro-grid equipment
For the purposes of this study, the leading brands of equipment in
New Zealand’s renewable energy asset market were chosen based on the
first author’s judgement of prevalence. The following sub-sections
mathematically model the system equipment.
2.1.1. Photovoltaic plant
Canadian Solar’s CS6K-280P poly-crystalline photovoltaic (PV)
modules [48], which have a nominal power of 280 W are employed in
this study for PV power generation. The power output from the PV plant
at each time-step, PPV(t) [kW], can be estimated as follows [25,49,50]:
Tm(t) = Ta(t) + IG(t) ×
NMOT − 20
0.8
, (1)
PPV (t) = NPV × PPV,r × ηPV,DC/DC × DF ×
IG(t)
ISTC
×
(
1 −
Kp
100
× (Tm(t)
− TSTC )
)
, (2)
where NPV is the optimum quantity of the modules; PPV,r is the rated
capacity of the module under the standard test conditions (STC);
ηPV,DC/DC is the PV plant’s DC/DC converter efficiency; Kp is the tem
perature coefficient of the module; Tm, Ta, and TSTC respectively repre
sent the PV module temperature, ambient temperature, and the module
temperature at the STC; IG and ISTC respectively denote the global solar
irradiance on the horizontal surface and the solar irradiance at the STC;
and NMOT and DF respectively stand for the nominal module operating
temperature and derating factor. The tilt angle is assumed as 30◦
and the
Meteonorm software [51] is used to normalise the values of IG to this tilt
angle. Also, the numeric values 20 and 0.8 respectively represent the
ambient temperature [◦
C] and solar irradiance [kW/m2
] at which the
NMOT is defined.
2.1.2. Wind plant
The wind turbine (WT) ECO 48/750, which has a rated power of 750
kW is considered for wind power generation [52]. The turbine’s
manufacturer-provided characteristic power-wind speed curve is shown
in Fig. 3. The wind plant’s output power at each time-step, PWT(t) [kW],
can be obtained by multiplying the optimal quantity of the WTs, NWT, by
each turbine’s output power estimated from the power curve presented
in Fig. 3. Also, since the power curve of the WT is characterised for its
hub height wind speed, Eq. (3) can be used to normalise the wind speed
data measured at other heights to the turbine’s hub height [53].
Vh = Vref ×(
h
href
)γ
, (3)
where Vref denotes the reference wind speed collected at the height of
href and γ ∈ [0.1, 0.25] is the wind shear exponent, which varies with
respect to the terrain [54].
2.1.3. Micro-hydro plant
Suneco Hydro’s XJ50-100SCTF6-Z 100-kW micro-hydro turbines are
selected to be integrated into the run-of-the-river plant as part of the MG
system [55]. The power output from the plant at each time-step [kW]
can be estimated from Eq. (4) [56,57].
PMH(t) =
NMH × ηMH,AC/DC × ηMH × ρ × g × hg × F(t)
1000
, (4)
where NMH denotes the optimum quantity of turbines, ηMH is the total
efficiency of the plant (including the turbine, generator, and water
1
In game theory, a Stackelberg duopoly is a non-symmetric, strategic,
sequential game with one party, or a group of parties, taking over the leading
position and the other(s) acting as follower(s).
S. Mohseni et al.](https://image.slidesharecdn.com/mohseni2021-211027145406/85/Mohseni2021-7-320.jpg)
![Applied Energy 287 (2021) 116563
8
wheel efficiency), ηMH,AC/DC is the efficiency of the plant’s AC/DC con
verter, ρ represents the density of water, g is the acceleration due to
gravity, hg is the gross head (which is defined as the difference between
the head race and tail race levels when water is not flowing), F(t) is the
flow rate at time-step t [m3
/s], while the numeric value of 1000 converts
the unit of measurement from Wh to kWh.
2.1.4. Biomass plant
The integrated biomass gasifier-generator system PP30 Cogen-CS
manufactured by All Power Labs [58] is utilised in this study. The
plant, the flow diagram of which is shown in Fig. 4, is a commercially
available, off-the-shelf component with an electrical rated power of 25
kW. The power output from the biomass plant at each time-step [kW]
can be calculated from Eq. (5) [59].
PBP(t) = NBP × ηBP,AC/DC × ηBP × CV × MBP(t), (5)
where NBP represents the optimal quantity of the considered biopower
units, ηBP,AC/DC is the efficiency of the plant’s AC/DC converter, ηBP is the
overall bio-electricity generation efficiency of the system, CV stands for
the gross calorific value of the biomass feedstock, and MBP(t) denotes the
feedstock mass consumption rate at time-step t [kg/h].2
Furthermore, the system is characterised with a carbon emission
factor of 1.53 kg-CO2 per kg of feedstock used [60]. Accordingly, the
social cost of the carbon emissions needs to be factored into the decision-
making–for an eco-design of the MG system. The following equation can
be used to calculate the life-cycle penalty imposed on the MG for CO2
emissions:
Fig. 1. Overview of the section-wise modelling procedure employed in this paper for the aggregator-mediated, market-driven integration of flexible demand re
sources in the long-term planning of MGs.
2
Note that the rated powers of micro-hydro turbines and biopower plants,
are incorporated into the model and the decision-making process in an indirect
manner using the power rating-dependent parameters–hg in the case of micro-
hydro turbines, and MBP in the case of biopower units–as well as specifically
developed terminal constraints (refer to Section 4.2.6 for more details).
S. Mohseni et al.](https://image.slidesharecdn.com/mohseni2021-211027145406/85/Mohseni2021-8-320.jpg)
![Applied Energy 287 (2021) 116563
9
costem =
ξCO2
1000
× ECO2
×
∑
T
t=1
MBP(t), (6)
where ξCO2
[$/tCO2] denotes the social cost of CO2 emissions used as a
reference to account for life-cycle GHG impacts of the biopower plant in
the model, and ECO2
represents the CO2 emission factor of the plant [kg-
CO2/kg-feedstock].
2.1.5. Upstream power grid
The MG system is tied to the upstream electricity network through a
dedicated bidirectional MV/LV transformer, the optimal capacity of
which is under investigation. The cost imposed by purchasing electricity
from the grid at each time-step could be represented by Eq. (7), while the
income generated by the MG’s electricity exports is obtained from Eq.
(8) [61].
costim(t) = πim(t) × Pim(t) × Δt, (7)
incomeex(t) = πex × Pex(t) × Δt, (8)
where πim(t) represents the (time-varying) wholesale electricity market
price at time-step t [$/kWh], πex is the utility grid’s single-tier (flat) buy-
back rate [$/kWh], Pim(t) is the amount of power imported from the
utility grid at time-step t, Pex(t) is the amount of power exported to the
Fig. 2. Micro-grid system architecture and streams of energy driven by renewables and the upstream grid.
Fig. 3. Power curve of the ECO 48/750. Data Source: [52].
Fig. 4. Schematic diagram of the considered integrated biomass gasifier-
generator system. Source: [60].
S. Mohseni et al.](https://image.slidesharecdn.com/mohseni2021-211027145406/85/Mohseni2021-9-320.jpg)
![Applied Energy 287 (2021) 116563
10
main grid at time-step t, and Δt represents the length of each time-step.
The power exchange is expected to adhere to the following con
straints:
Pim(t)/ηT ≤ NT , (9)
Pex(t)/ηT ≤ NT , (10)
where ηT denotes the transformer’s efficiency and NT represents the
rated capacity of the transformer, which is to be optimised.
The generic solid-state transformer, designed by Qin and Kimball
[62], is used in this study. The size of the transformer is characterised by
the apparent power [kVA] and, as a simplifying assumption, the power
factor is assumed to be 0.95.
2.1.6. Power conversion apparatuses
As shown in Fig. 2, the MG system is equipped with several con
verters to serve the purpose of coupling the equipment to a common DC
busbar. For electrical loads, Leonics’ GTP-519-S 900-kW, GTP-506 115-
kW, and GTP-501 33-kW inverters are considered in this study [63]. To
calculate the size of the electrical loads’ inverters, first, the following
equation is used to determine the nominal power of the overall power
inversion system:
NI =
PL,max
ηI
, (11)
where PL,max represents the demanded annual peak electrical loads and
ηI identifies the power inversion equipment’s efficiency.
Then, NI is rounded up to the next integer and the number of each
inverter model is identified by the following equations, which give
priority to higher-rated inverters as they carry a lower per-unit cost:
N900 =
⌊
NI
C900
⌋
, (12)
N115 =
⌊
NI − (N900C900)
C115
⌋
, (13)
N33 =
⌈
NI − (N900C900)− (N115C115)
C33
⌉
, (14)
where N900, N115, and N33 respectively denote the quantity of the 900-
kW, 115-kW, and 33-kW inverters, while C900, C115, and C33 indicate
their respective rated capacities.
2.1.7. Internal backup energy storage
The proposed system leverages the temporal characteristics of
various DERs providing backup power, or energy storage. To this end,
this study expands on the idea proposed by Akram et al. [64] that low-
pass energy filters could be used to calculate the share of each energy
storage medium in supplying load power demand on a representative
MG. Accordingly, the power mismatch signal is first broken down into
the low- and high-frequency components using a low-pass filter with a
transfer function given in Eq. (15).
H(s) =
Kω2
0
s2 + (ω0/Q)s+ω2
0
, (15)
where ω0 denotes the cut-off frequency, K represents the DC gain, and
Q = 1/2ξ identifies the quality with ξ indicating the damping factor.
Then, the low-frequency signal is directed to the hydrogen system
(including the electrolyser, hydrogen tank, and the fuel cell), while the
high-frequency signal is transferred to the hybrid battery-SC system.
Subsequently, another low-pass filter with a lower cut-off frequency
identifies the contribution of the battery and SC banks in serving loads or
storing surplus power.
The technical rationale underlying this power allocation approach is
the longer cycle life, higher round-trip efficiency, and more rapid
response capability of SCs (batteries) to balance out generation-demand
mismatches than batteries (the hydrogen system). That is, the shortest
and longest periods of surplus or shortage of electricity are addressed
using the SC bank and hydrogen system, respectively, while the battery
bank bridges the gap between these two storage media.3
2.1.7.1. Super-capacitor bank. Eaton’s 48-V, 166-F XLR-48R6167-R SC
modules [65], which are of the type electrochemical double-layer
capacitor (EDLC), are used to address short-term renewable power and
load demand mismatches–and improve the MG’s dynamic response and
overall efficiency. The SC bank’s energy content at each hour of the MG
operation can be calculated as follows:
ESC(t) = ESC(t − 1) +
(
Pch,HF2 −
(
Pdch,HF2
ηSC
) )
× Δt, (16)
where ηSC represents the SC’s round-trip efficiency, while Pch,HF2 and
Pdch,HF2 are the high-frequency components of the outputs of the second-
stage filtered charging and discharging signals, respectively.
2.1.7.2. Battery bank. CellCube’s vanadium redox flow-based battery
bank [66] is used in the conceptualised MG. Likewise to the inverter
system, three different battery product models are selected and the same
procedure is followed to apportioning the total optimal size of the bat
tery bank to different model types, following the same logic. The battery
product models are: FB 10–100 (100 kWh), FB 200–400 (400 kWh), and
FB 400–1600 (1600 kWh). The battery bank’s energy content at each
hour can be obtained as follows:
EB(t) = EB(t − 1) +
(
Pch,LF2 −
(
Pdch,LF2
ηB
) )
× Δt, (17)
where ηB is the battery bank’s round-trip efficiency, while Pch,LF2 and
Pdch,LF2 denote the low-frequency components of the outputs of the
second-stage filtered charging and discharging signals, respectively.
2.1.7.3. Hydrogen storage. The hydrogen-based storage system mainly
includes polymer electrolyte membrane (PEM) electrolyser stacks, a
medium-pressure (20 bar) hydrogen reservoir, and stationary PEM fuel
cell stacks. H-TEC Systems’ S 30/50 5-kW electrolyser stacks [67], a
generic hydrogen reservoir (which needs to be customised), and Bal
lard’s FCgen-1020ACS 3.3-kW fuel cell stacks [68] are used as part of the
hydrogen storage system. The hydrogen power directed from the elec
trolyser outlet to the reservoir at time-step t can be obtained as follows:
PE− HT (t) = PE(t) × ηE, (18)
where PE is the electrolyser’s consumed power, which is controlled by
the low-frequency component of the output of the first-stage filtered
charging signal, while ηE denotes the electrolyser’s efficiency.
The mass of hydrogen, mHT [kg], stored in the reservoir at each time-
step can be calculated as follows:
EHT (t) = EHT (t − 1) +
(
PE− HT (t) −
(PHT− FC(t) + PHT− S(t))
ηHT
)
× Δt, (19)
mHT (t) =
EHT (t)
HHVH2
, (20)
where EHT represents the reservoir’s energy level, PE− HT is the directed
3
Note that the backup power allocation strategy employed in this study is
tailored towards long-term capacity planning, at which stage long-term fore
casted data are available. A forward-looking predictive modelling approach
(using a critic network, for example) is indispensable for the real-time operation
phase.
S. Mohseni et al.](https://image.slidesharecdn.com/mohseni2021-211027145406/85/Mohseni2021-10-320.jpg)
![Applied Energy 287 (2021) 116563
11
hydrogen power from the electrolyser to the reservoir, PHT− FC and PHT− S
respectively denote the hydrogen power consumption of the fuel cell and
the FCEV parking lot, ηHT represents the tank’s round-trip efficiency, and
HHVH2
stands for the higher heating value of hydrogen.
The electric power output from the high-energy-density fuel cell at
time-step t, which is controlled by the low-frequency component of the
output of the first-stage filtered discharging signal, can be obtained
using Eq. (21).
PFC(t) = PHT− FC(t) × ηFC, (21)
where PHT− FC represents the fuel cell’s consumed hydrogen power and
ηFC denotes its electrical efficiency, which is defined as the ratio between
the electricity generated and the hydrogen consumed.
2.1.8. Fuel cell electric vehicle parking lot
The hydrogen refuelling infrastructure of the parking lot mainly
consists of a medium-pressure (20/350 bar) compressor, a buffer stor
age, a cryogenic pump, as well as a vaporiser, a refrigeration unit, and
some dispensers to deliver gaseous hydrogen fuel to FCEVs [69]. The
refuelling infrastructure is modelled by its overall efficiency, which is
denoted by ηS. To this end, the Pure Energy Centre’s customised
hydrogen refilling station [70] is considered for integration into the
proposed MG.
2.1.8.1. Selected fuel cell electric vehicles. A fleet of ultra-light-duty
personal passenger vehicles is planned for integration into the envi
sioned system through the coordinated use of the refuelling infrastruc
ture. Accordingly, vehicles are assumed to be refuelled on a first-come/
first-served basis using the multi-server Erlang-C queuing model [71],
where C identifies the optimal number of dispensers. Also, FCEVs are
assumed to be of the model Riversimple Rasa.
2.1.8.2. Fuel cell electric vehicles in vehicle-to-grid operation. To provide a
platform for exploiting the V2G capabilities of the FCEVs, the FCEV2G
setup designed in [72] is used in this study. The setup enables the
conversion of the DC power of the vehicle’s fuel cell engine into AC that
can be directed to the input port of the electrical loads’ inverter after
frequency synchronisation, with an overall efficiency of ηFCEV2G.
Accordingly, modulation of the power output from each FCEV, the
owner of which aspires to participate in the V2G operations, can be
made from 0 to 8.5 kW DC–in compliance with the nominal capacity of
Rasa’s built-in fuel cell. This means the costs arising from payments
made to FCEV owners to provide V2G power at each time-step–under a
feed-in-tariff style programme–can be calculated by the following
equation:
costFCEV2G(t) = πFCEV2G × ηFCEV2G × PFCEV2G(t) × Δt, (22)
where πFCEV2G represents the per-unit premium tariff rate for V2G power
[$/kWh] and PFCEV2G(t) denotes the amount of V2G power used for
operational scheduling at time-step t.
For the sake of simplification, it was assumed that at each time-step
of the MG operation, the maximum amount of available V2G power that
can be provided by the station at each time-step, Pmax
FCEV2G(t), equals 25%
of the load reduction potential of the station at that time-step.
2.1.9. Data: Selected product models
The values of the underlying system scalars, defined above, are
presented in Table 2. Also, the techno-economic specifications of the MG
equipment, namely the capital, replacement, and operation and main
tenance (O&M) costs, as well as the estimated service life and efficiency
of the equipment are summarised in Table 3.
2.2. Operational strategy
A rule-based, hourly-basis, cycle-charging operational strategy is
adopted in this study for the dispatch of energy within the MG system,
which is illustrated by the flowchart in Fig. 5. In the devised energy
scheduling plan, (1) energy storage devices and FCEVs are recharged/
refilled using only the surplus non-dispatchable renewable power, (2)
non-dispatchable renewable power and electrical loads are partitioned
into the ultra-high, high, and low-frequency components and then
stored/supplied within/using the SC bank, battery bank, and the
hydrogen tank/fuel cell, respectively, (3) the dispatchable biopower
plant can only be operated during the time-slots stamped as peak hours
to partially or wholly offset the lack of sufficient fuel cell power,4
(4) the
upstream grid serves as the ultimate guarantor of the perfect satisfaction
of the electric load demand, and (5) the FCEV2G capability is considered
as a resource to compensate for at least part of the electricity left un
served by the fuel cell and the biopower plant, or the shortage of battery
and SC capacity to meet the load power demand. To this end, morning
and evening peak demand were assumed to occur between the hours of 6
a.m. to 10 a.m. and 5 p.m. to 9 p.m., respectively–in compliance with
historical records of electricity consumption in New Zealand.
Moreover, the key assumptions made in conceptualising the pro
posed MG system and conducting the life-cycle analysis are listed in
Supplementary Material (Additional File 1: Key assumptions underlying
the conceptualised micro-grid model).
3. Aggregator-mediated, incentive-based demand-side
management market design
This section presents a mathematical formulation of a two-stage,
aggregator-mediated, incentive-based DSM market model specifically
developed for integration into standard MG capacity planning ap
proaches. Building on the interruptible load programmes, the model is
designed specifically to improve the accuracy of projections of the small-
to medium-scale DR resource availability across different end-use sec
tors–residential, commercial, industrial, agricultural, and electrified
transportation. More specifically, it characterises the interactions be
tween a MG operator, DRAs, and end-consumers. To this end, the model
consistently treats these three sets of actors as rational, utility-
Table 2
Data values and sources for the proposed micro-grid system scalars.
Scalar Value Source Scalar Value Source
CV 5.07 kWh/kg [73] Kp − 0.40%/◦
C [48]
Δt 1 h (this
paper)
NMOT 43 ◦
C [48]
ηPV,DC/DC 95% [34] ρ 1000 kg/m3
−
ηMH,AC/DC,
ηBP,AC/DC
95% [34] PBP,r 25 kW [58]
DF 85% [74] πex $0.05/kWh [75]
ECO2
1.53 kg-CO2/
kg-feedstock
[60] πFCEV2G $0.05/kWh (this
paper)
g 9.81 m/s2
− PL,max 7.31 MW (this
paper)
h 55 m [52] PMH,r 100 kW [55]
hg 10 m [55] PPV,r 0.28 kW [48]
href 10 m [76] TSTC 25 ◦
C [78]
HHVH2 39.7 kWh/kg [77] γ 0.15 [79]
ISTC 1 kW/m2
[78] ξCO2
$42/tCO2,
$50/tCO2*
[80]
*
A central value of $42/tCO2 is applied for the first 10-year planning horizon
(covering the years 2020 to 2030), which rises to $50/tCO2 for the second half of
the projected lifespan of the project in accordance with the Obama adminis
tration’s central estimates [80].
4
This assumption can be explained by the relatively long cold start-up time of
the biopower plant (i.e. ~10–15 min) and the inefficiency of leaving the bio
power plant on standby at all times.
S. Mohseni et al.](https://image.slidesharecdn.com/mohseni2021-211027145406/85/Mohseni2021-11-320.jpg)
![Applied Energy 287 (2021) 116563
12
maximising (self-interested), active economic agents. The proposed
market design provides a forum for these economic agents to negotiate
on how to mutually optimise their objective functions in non-
cooperative (strategic) game settings under the Stackelberg competi
tion. It also identifies the minimum operational MG costs based on
hourly priced DR products and the wholesale power price. In this way,
the model enables all the active agents within the MG to be involved in
co-designing a business model for more independent energy procure
ment. Fig. 6 displays a schematic of the overall structure of the model
with the sequence of incentive price/DR supply communications be
tween the market participants overlaid [96].
As Fig. 6 shows, the market-based, aggregator-mediated DSM strat
egy is modelled as an interactive hierarchical decision-making process,
which consists of two levels of leader–follower relationships, namely
between the MG operator and the DRAs (wholesale DSM market), and
between the DRAs and their customers (retail DSM market). Although
the DSM market participants are hierarchically related with respect to
DR service, each has an independent viewpoint on the problem, which is
modelled by specific objective functions in the following sub-sections.
3.1. Micro-grid operator
It is assumed that the conceptualised MG, laid out in Section 2, runs
on an energy-as-a-service business model in that not only does a third-
party (private company) own the MG, but it also provides an over
arching framework for energy management (through effective incentive
Table 3
Data values and sources for techno-economic specifications of the conceptualised system’s components.
Component Manufacturer
part number
Nameplate
rating
Capital cost* Replacement
cost†
Operation and
maintenance
cost†
Expected
service life
Nominal efficiency Source
Per unit Per standard
unit of
generation/
storage/
conversion
capacity
Notation Value
[%]
PV module CS6K-280P 280 W $210/
unit
$750/kW $200/unit $1/unit/year 25 years ηPV 17.11 [48]
Wind turbine ECO 48/750 750 kW $1.096
m/unit
$1.46 k/kW $0.822 m/
unit
$21 k/unit/
year
20 years N/A‡
N/A‡
[52]
Micro-hydro
turbine
XJ50-100SCTF6-
Z
100 kW $56 k/
unit
$560/kW $17 k/unit $2 k/unit/year 25 years ηMH 78 [55]
Biopower
unit§
PP30 Cogen-CS 25 kW $32 k/
unit
$1.28 k/kW $23 k/unit $0.01/unit/
hour
10 k hours ηBP 23 [58]
Transformer Generic − − $65/kVA $55/kVA $2/kVA/year 30 years ηT 93 [62,81]
Inverter GTP-501 33 kW $12 k/
unit
$364/kW $12 k/unit $85/unit/year 15 years ηI 96 [63]
GTP-506 115 kW $38 k/
unit
$330/kW $38 k/unit $250/unit/
year
GTP-519-S 900 kW $270 k/
unit
$300/kW $270 k/unit $1.9 k/unit/
year
Super-
capacitor
module
XLR-48R6167-R 166F, 48 V
≡ 0.054
kWh
$1.3 k/
unit
$24.1 k/kWh $1.3 k/unit $13/unit/year 1 m cycles ηSC 97 [65]
Battery pack FB 10–100 100 kWh $110 k/
unit
$1.1 k/kWh $110 k/unit $220/unit/
year
20 years
with
unlimited
cycles
ηB 80 [66,82]
FB 200–400 400 kWh $400 k/
unit
$1 k/kWh $400 k/unit $840/unit/
year
FB 400–1600 1600 kWh $1.442
m/unit
$901/kWh $1.442 m/
unit
$4 k/unit/year
Electrolyser
stack
S 30/50 5 kW $6 k/
unit
$1.2 k/kW $6 k/unit $120/unit/
year
20 years ηE 75 [67]
Hydrogen
tank
Generic − − $500/kg $500/kg $1/kg/year 20 years ηHT 95 [83]
Fuel cell
stack
FCgen-1020ACS 3.3 kW $5 k/
unit
$1.52 k/kW $5 k/unit $0.02/unit/
hour
10 k hours ηFC 40 [68]
Hydrogen
station
Generic (Pure
Energy Centre)
− − $10 k/(kg-H2/h) $5 k/(kg-H2/
h)
$350/(kg-H2/
h)/year
20 years ηS 95 [69,70,84]
Generic (The
Energy
Technology
Section, TU
Delft)¶|
− − $155/kW $95/kW $32/kW/year 20 years ηFCEV2G 44#
[85,86,87,88]
¶
In view of the assumption that the DC power provided by the FCEVs is fed into the electrical loads’ inverter, the costs associated with the FCEV2G technology only
include the costs of modifying the vehicles with a V2G DC outlet plug.
*
All of the reported capital costs represent the actual cost of buying the selected components in New Zealand’s energy asset market as of October 2019–which were
adjusted to 2019 U.S. dollars. In October 2019, US$1 = NZ$1.56.
†
All of the replacement and O&M costs were calibrated in accordance with the component-specific ratios of capital to replacement and O&M costs reported in
[82,83,89,90,91,92,93,94,95].
‡
Not applicable as the wind turbine efficiency is reflected in its power curve shown in Fig. 3.
§
To value the positive impact of the biopower plant on the internal dispatchability of the MG, the total discounted cost of pellet feedstock was considered to be an
exogenous variable, which is determined outside the model based on the imposed emission credits (see Eq. (6)) with respect to the total discounted energy output of the
plant (see Eq. (5)).
#
The V2G infrastructure’s efficiency in this paper represents a tank-to-DC-bus efficiency (units converted based on the higher heating value of hydrogen).
S. Mohseni et al.](https://image.slidesharecdn.com/mohseni2021-211027145406/85/Mohseni2021-12-320.jpg)
![Applied Energy 287 (2021) 116563
14
reduction,
∑
j∈JD
j
LA; IMGO is the MG operator-posted incentive price
signal to the wholesale DSM market, with superscripts min and max
representing its lower and upper limits, respectively; Ddef denotes the net
energy deficit of the system; and D
j
LA is the total load reduction procured
by DRA j ∈ J.
3.2. Demand response aggregators
The DRAs serve as a go-between, interfacing with the smaller DR
providers and the broader MG system operator so as to maintain the
visibility of the small-scale DR products. The independence of the DRAs
is fully preserved in the proposed model as they are precluded from
ownership of the energy infrastructure. Precisely, third-party aggre
gators enlist end-consumers to take part in interruptible load pro
grammes. To this end, they take a percentage of the MG operator-offered
incentive as compensation, passing the rest on to their customers. More
specifically, the DRAs aim to maximise the objective (profit) function in
Eq. (26) subject to the constraints in Eqs. (27) and (28) [97]:
Prj
LA(t) =
(
IMGO(t) − Ij
LA(t)
)
× Dj
LA(t)∀j, t, (26)
Ij,min
LA ≤ Ij
LA(t) ≤ Ij,max
LA ∀j, t, (27)
Dj
LA(t) =
∑
k∈NJ
dk,j
(t)∀j, t, (28)
where I
j
LA is the incentive rate posted by the j-th aggregator to the retail
DSM market; dk,j
denotes the capacity of DR product supplied by
customer k subscribed to aggregator j; NJ is the set of customers serviced
by aggregator j, which is a proper subset of set of all the customers
within the MG system’s operational territory, K; and I
j,min
LA and I
j,max
LA
respectively represent the lower and upper bounds of the incentive
payments offered by aggregator j.
3.3. End-consumers
End-consumers, who are activated by third-party DRAs, have the
opportunity to take full advantage of their flexibility potential, whilst
adhering to a set of discomfort cost constraints. To this end, the end-
consumers determine the optimum supply of their DR resources with
respect to the DRA-offered incentive prices by maximising the utility
function expressed in Eq. (29) subject to Eqs. (30) and (31).
Uk,j
(t) = dk,j
(t) × Ij
LA(t) − disk,j
(t)∀k, t, (29)
0 ≤ dk,j
(t) ≤ dk,j
ncr(t)∀k, t, (30)
dk,j
full(t) = dk,j
cr (t) + dk,j
ncr(t)∀k, t, (31)
where disk,j
denotes the cost of discomfort (inconvenience) associated
with load reductions as a measure of the value of electricity, which can
be obtained from Eq. (32)5
[98,99], and must lie within a certain range,
as constrained by Eq. (33); d
k,j
full is the full (original) load demanded by
customer k of aggregator j; d
k,j
cr is the critical portion of the original load,
any shedding of which results in impaired reliability; and d
k,j
ncr is the non-
critical (dispatchable) demand, which can be interrupted by making
effective incentive payments to customers for curtailing load.
disk,j
= ck,j
1 (dk,j
)2
+ ck,j
2 (1− δj)dk,j
∀k, t, (32)
disk,j,min
≤ disk,j
≤ disk,j,max
∀k, t, (33)
where c
k,j
1 and c
k,j
2 are positive individual-level parameters specified by
end-consumers that characterise their sensitivity to load reductions, for
customers indifferent to incentive payment, c
k,j
1 ,c
k,j
2 →∞; 0 ≤ δj ≤ 1 is the
sector-level elasticity of customer-supplied DR capacity, for a hypo
thetical completely inelastic customer category, δj→0; while disk,j,min
and
disk,j,max
respectively denote the minimum and maximum allowable
Fig. 7. Sequence diagram of implementing the proposed DSM model in the context of the conceptualised MG system.
5
The customer discomfort cost function can be viewed as the second-order
best-fit equation to individual-level, user-specified data points representing
ordered pairs of DR capacity supply and the associated discomfort cost
incurred.
S. Mohseni et al.](https://image.slidesharecdn.com/mohseni2021-211027145406/85/Mohseni2021-14-320.jpg)
![Applied Energy 287 (2021) 116563
16
power and hydrogen demand on the system are scaled-down (modified)
through running the proposed DR scheduling framework for the specific
peak hours of each day of the representative year before being fed to the
hourly operational scheduling strategy outlined in Fig. 5. The process
continues by transmitting the aggregator’s incentives for load reduction
to their corresponding customers, and completes by clearing the DSM
markets respectively at the local (retail) and wholesale levels. As
mentioned above, this procedure is repeated for each hour of a repre
sentative hourly-basis, one-year operational timeframe. To this end, the
year-long demand profiles are decomposed into daily profiles so as to be
used in the day-ahead DR management plan of the MG (see Fig. 7), the
DR-adjusted values of which are then used in the course of the hourly
energy management of the system (see Fig. 5).
4. Micro-grid capacity-optimisation model
This section explains the deterministically estimated life-cycle cost of
the conceptualised MG system before describing how the proposed non-
cooperative game-theoretic DR management scheme is integrated into
the MG sizing model. The MG capacity-optimisation model consists of
three key elements: (1) the net present cost (NPC) and net present value
(NPV) methods utilised to formulate the total discounted system cost
function, (2) the loss of power supply probability (LPSP) technique to
quantify the reliability of the system in servicing the electrical and
hydrogen load demands, and (3) the moth-flame optimisation algorithm
(MFOA) [100] as a single-objective meta-heuristic optimisation algo
rithm to find the globally optimum solution to the problem by mini
mising the life-cycle cost of the MG, whilst adhering to the technical,
reliability, and systemic constraints (see Supplementary Material
(Additional File 2: Techniques used in the micro-grid capacity-optimi
sation model) for details). The superiority of the single-objective MFOA
to the well-established meta-heuristics in the MG investment planning
literature–for instance, the genetic algorithm (GA) [101] and the PSO
[102]–as well as to a wide variety of state-of-the-art meta-heuristics in
terms of nearing the globally optimum solution is supported in previous
studies [41,84,103,104,105].
4.1. Objective function
A static analysis of expected future cash flows for the underlying
project lays the basis for the mathematical formulation of the objective
function. The whole-life cost of the MG based on the NPC and NPV
calculations, which is to be minimised, can be expressed as follows:
minWLC =(
∑
c∈C
NPCc) + NPCI + NPV
(
∑
8760
t=1
OCMG(t)
)
+ NPV
(
∑
8760
t=1
costem(t)
)
+ NPV
(
∑
8760
t=1
costFCEV2G(t)
)
− NPV
(
∑
8760
t=1
incomeex(t)
)
+ penconst,
(39)
Where NPCc represents the NPC of the components, the optimal size of
which is under investigation and are indexed by c ∈ C = {PV,WT,MH,T,
E, FC, HT, BP, B, SC, S, FCEV2G}; NPCI denotes the NPC incurred by the
inverter; OCMG is the operational cost of the MG to serve the unmet
loads, either by paying incentives for load reduction or purchasing
power from the upstream grid, as defined in Eq. (7); costem is the cost
imposed on the system for buying emission credits on account for
running the biopower plant, as given in Eq. (6); costFCEV2G denotes the
cost resulting from the provision of FCEV2G services, as expressed in Eq.
(22); incomeex is the income generated by selling the surplus power to
the main grid, as expressed in Eq. (8); while the term penconst enforces the
solutions to meet the constraints set out in Section 4.2.
In this context, the useful life of the project was considered to be 20
years and the real interest rate was set to 3.7%. The real interest rate was
projected by taking the mean of the historical records in New Zealand
over a 10-year period, between 2010 and 2019 [106].
4.2. Problem constraints
The objective function presented above is subject to various sets of
constraints along the following lines.
4.2.1. System reliability
The LPSP reliability metric is employed to characterise the system
performance over its projected 20-year life span. To this end, two
separate LPSP indices are used to evaluate the reliability of electricity
and hydrogen supply, which are constrained by Eqs. (40) and (41),
respectively.
LPSPe ≤ LPSPmax
e , (40)
LPSPH2
≤ LPSPmax
H2
, (41)
where LPSPmax
e and LPSPmax
H2
denote the imposed upper bounds on LPSPe
and LPSPH2 , respectively.
4.2.2. System-wide power balance
According to Eq. (42), at each time-step of the system operation, the
sum of all of the internally generated energy components, energy re
leases from the storage media, energy imports from the main grid, and
any unmet load must be equal to the sum of the total energy consumed
within the MG (to serve the loads or to charge the energy storage de
vices) and any power sold to the upstream grid.
PPV (t) + PWT (t) + PMH(t) + PBP(t) + Pdch(t) + PFC(t) + Pim(t) + PFCEV2G(t)
+
QL(t)
ηI
+
QH2
(t)
ηS
= Pch(t) + PE(t) + Pex(t) +
PL(t)
ηI
+
PS(t)
ηS
∀t,
(42)
where QL(t) and QH2 (t) respectively represent the unmet electrical and
hydrogen demands at time-step t, which are used in the LPSP
calculations.
4.2.3. Demand response scheduling
As mentioned previously, under equilibrium conditions of the pro
posed two-stage, aggregator-mediated, market-driven DR arrangement,
the constraints in Eqs. (24), (25), (27), (28), (30), (31), (33) must be
relaxed.
4.2.4. Energy storage systems and fuel cell electric vehicles
The optimisation of the MG equipment capacity must additionally
adhere to some constraints in terms of charge/discharge rate limits of
the energy storage media and FCEVs, bounding the state of charge/
hydrogen of the storage systems and vehicles, as well as the state of
energy reserves in the first and last operating hours, which could be
expressed mathematically as:
Ees,min ≤ Ees(t) ≤ Ees,max∀t, es, (43)
Pch,min
es ≤ Pch
es (t) ≤ Pch,max
es ∀t, es, (44)
Pdch,min
es ≤ Pdch
es (t) ≤ Pdch,max
es ∀t, es, (45)
Ees− {FCEV}(0) = 0.5 × Ees− {FCEV},max∀es, (46)
Ees− {FCEV}(8760) ≥ Ees− {FCEV}(0)∀es, (47)
where Ees(t) is the energy content of the energy storage technology es ∈
ES = {B, SC, HT, FCEV} at time-step t; Ees,min and Ees,max respectively
denote the minimum and maximum allowable energy content of energy
storage technology es; Pch
es (t) and Pdch
es (t) respectively represent the
S. Mohseni et al.](https://image.slidesharecdn.com/mohseni2021-211027145406/85/Mohseni2021-16-320.jpg)
![Applied Energy 287 (2021) 116563
17
charging and discharging rates of storage technology es at time-step t;
Pch,max
es and Pdch,max
es are the maximum charging and discharging rates of
storage technology es, respectively; and Pch,min
es and Pdch,min
es are the min
imum charging and discharging rates of storage technology es, respec
tively.
The maximum allowable energy content of the battery bank, SC
bank, and hydrogen tank are defined by their optimised capacity at each
iteration of the optimisation process, whereas the maximum total energy
content of the releasable hydrogen stored in the FCEVs’ tanks
(max(Pmax
FCEV2G(t)Δt) where t ∈ T) is limited by the maximum (optimal)
capacity of the FCEV2G setup (as part of the hydrogen station) in
addition to the stored hydrogen in the vehicles’ tanks at time-step t. That
is, the variables Ees,max, es ∈ ES are treated as endogenous variables in the
model. Also, the same principle holds for the variables Pch,max
es and
Pdch,max
es .
Moreover, in the interest of preventing the performance degrada
tion–and mitigating the energy losses–during the start-up and shut-
down cycles of the electrolyser, fuel cell, and biopower plant, a spe
cific constraint preserves the durability of their operation. To this end,
when the electrolyser, fuel cell, and biopower plant are started up, they
are constrained to continue to run for at least tup time-steps–as a mini
mum up-time constraint–at operating points equal to, or greater than the
initially adjusted operating points. Accordingly, the power output from
the fuel cell and biopower plant are treated as negative loads in the
course of the MG operation on the extra hours mentioned above, whilst
also being allowed to take higher operating point values where appro
priate.
In addition, to avoid severe pressure drops in the hydrogen tank,
complete releases of hydrogen are prevented by enforcing EHT,max not to
fall short of 5% of the optimised capacity of the tank. Also, to ensure that
the design pressure of the tank is not exceeded, the upper limit on the
energy content of the tank is set as 95% of its optimum capacity [107].
4.2.5. Energy exchange
The MG’s transactions of energy with the upstream power network is
constrained by Eqs. (9) and (10) to adhere to the optimal size of the
transformer connecting the MG system with the upstream grid at the
point of common coupling (PCC).
4.2.6. Decision variables
Specific upper bounds are set for maximum values the non-negative
design variables can take, as represented in Eq. (48). These bounds are
adjusted commensurate with the practical feasibility of implementing
the conceptualised MG system in the considered area. For example, land
limitations, characteristics of the catchment sites, available biomass as a
feedstock, and acceptable emissions limits (from the biopower plant)
could constrain the feasible solution space.
Nc ≤ Nmax
c ∀c, (48)
where subscript c ∈ C indicates the MG components, the optimal size of
which is under investigation, while the superscript max denotes the
maximum permissible value of the optimum quantity/capacity of the
equipment (Nc).6
4.3. Meta-heuristic optimisation algorithm
Mathematically, the underlying MG capacity-planning model is a
nonlinear, non-convex, non-deterministic polynomial time-hard (NP-
hard) decision problem at its core, as indicated by Chen et al. [108].
Consequently, it cannot be solved exactly or by enumerating the entire
search space explicitly or implicitly, but meta-heuristic techniques could
be used effectively to solve the problem.
As noted earlier, the MFOA is employed to optimise a solution to the
formulated MG capacity-optimisation problem on account of its well-
proven superior performance to a wide range of both the well-
established and state-of-the-art meta-heuristics in the MG planning
context. Furthermore, owing to the mixed-discrete-continuous structure
of the formulated problem, the technique proposed by Chowdhury et al.
[109] is employed to modify the original continuous MFOA to make it
applicable to the problem at hand. Moreover, the control parameters of
the algorithm were adjusted as suggested by its developer [100], while
the number of search agents, NSA, and the maximum number of itera
tions, Itermax, were set based on the findings of Khan and Singh [110] on
the appropriate values to ensure the convergence of a broad spectrum of
meta-heuristic optimisation algorithms–including both the well-
established and state-of-the-art meta-heuristics–in the context of MG
design optimisation and capacity planning.
4.4. Data: Adjusted demand-response integrated micro-grid equipment
capacity planning model parameters
Table 4 lists the chosen data values for the parameters used to build
the proposed DR-integrated MG equipment capacity-planning model.
4.5. Overview of the proposed solution algorithm
The flowchart of the proposed MG equipment capacity-planning
model, which uses the proposed two-stage, aggregator-mediated
market-driven DR model to realistically project the customer engage
ment in incentive-based DR programmes–based on an economically
stable allocation of the profits generated from interruptible load pro
grammes between the sole energy service provider, DSM aggregators,
and end-users–is presented in Fig. 8. As can be seen from the figure, the
solution algorithm integrates the proposed DR provision framework (the
yellow block) and applies the developed rule-based hourly-basis oper
ational scheduling strategy (the light coral block), while taking an
iterative approach to optimise the discounted MG investment cost with
which to determine the respective size of the equipment (the blue
blocks).
Table 4
Data values for the demand response-integrated micro-grid equipment capacity
planning model parameters.
Scalar Value Scalar Value
Ees,max (endogenous variable) Nmax
FCEV2G 5,000 kW
Ees− {HT},min 0 kWh* Nmax
HT 50,000 kg
EHT,min (endogenous variable) Nmax
MH 30
iMGO $0.02/kWh Nmax
PV 20,000
Imin
MGO
$0.02/kWh NSA 100
Imax
MGO $0.32/kWh Nmax
S 100 kg-H2/h
Ij,min
LA
$(0.02− ε†
)/kWh Nmax
SC 10,000
I
j,max
LA
$(0.32− ε†
)/kWh Nmax
T 8,000 kVA
Itermax 500 Nmax
WT 15
LPSPmax
e 0% Pch,max
es
(endogenous variable)
LPSPmax
H2
5% Pch,min
es
ε†
kW
Nmax
B 20,000 kWh Pdch,max
es
(endogenous variable)
Nmax
BP 50 Pdch,min
es
ε†
kW
Nmax
E 1,000 penconst (1/ε†
)
Nmax
FC 2,000 tup 3 h
*
Note that the depth of discharge capability of the vanadium redox flow
battery is 100% and the total energy content of the FCEVs’ tanks is assumed to be
aware of the specific level of hydrogen expected (desired) by each FCEV owner
at the scheduled departure time.
†
The symbol ε denotes a small positive infinitesimal quantity.
6
The maximum permissible values of the design variables are aware of the
rated powers of the corresponding components.
S. Mohseni et al.](https://image.slidesharecdn.com/mohseni2021-211027145406/85/Mohseni2021-17-320.jpg)
![Applied Energy 287 (2021) 116563
19
Furthermore, the step-wise representation of the integrated simula
tion platform to optimally design the conceptualised MG, while man
aging the DR resources using the proposed DR scheduling approach is
summarised in Fig. 9. After the procurement and pre-processing of the
input data, the model is built up in a multi-layered structure, which
consists of (from bottom to top): (1) a rule-based hourly energy sched
uling strategy, (2) a two-stage, aggregator-mediated, DSM market
design to arrange the delivery of the DR resources on a day-ahead basis,
(3) various constraints the objective function is subjected to, and (4) the
derived fitness function representing the whole-life cost of the system,
which is to be optimised using the MFOA.
5. Case study
To confirm the proposition put forward in Section 1 on the effec
tiveness of integrating the proposed DR management framework into
the standard meta-heuristic-based MG capacity planning approach, as
well as the viability of the conceptual test-case MG system, laid out in
Section 2, this section presents the results of the case study analysis
conducted for the town of Ohakune, New Zealand. To this end, first, the
validity of the model is confirmed through a direct comparison of the
extreme-case model results with those of a business-as-usual (BAU), non-
game-theoretic interruptible DR scheduling framework. Then, the eco
nomic viability of integrating the developed DSM strategies into the
long-term MG investment decision-making processes is benchmarked
against two cases where: (1) the DSM market is cleared without
employing ideas from non-cooperative game theory for interactive
decision-making regarding the practical capacity of DR resources, and
(2) no provision is made to employ the responsive loads as a backup
resource in the proposed MG system. Finally, a financial appraisal
assessment demonstrates the economic sustainability of the proposed
renewable energy project. Numerical simulations were carried out using
the MATLAB software (version 9.5, R2018b) [111].
5.1. Case study site: The town of Ohakune, New Zealand
The notional MG system proposed in this study is envisioned to
decarbonise the energy economy of the town of Ohakune, which is sit
uated in the central part of the North Island of New Zealand–latitude
39.4180◦
S, longitude 175.3985◦
E [112].
The forecasted hourly-basis, year-long climatic input data streams,
are presented as monthly mean 24-h profiles in 3D plots in Fig. 10 [76].
Also, the forecasted monthly averaged profile for biomass availability is
shown in Fig. 11, assuming that the amount of monthly available
biomass is evenly distributed over the days of the months [113,114].
The forecasted one-year load power demand on the system, which is
represented as a monthly mean 24-h profile for greater clarity, is shown
in a 3D plot in Fig. 12 (a) [115,116]. Also, the forecasted monthly mean
Fig. 10. CliFlo-compliant forecasted meteorological input data (at an hourly resolution) for Ohakune, New Zealand: (a) solar irradiance; (b) ambient temperature; (c)
wind speed; and (d) streamflow.
S. Mohseni et al.](https://image.slidesharecdn.com/mohseni2021-211027145406/85/Mohseni2021-19-320.jpg)
![Applied Energy 287 (2021) 116563
20
24-h profile for the hydrogen demand of the refuelling sta
tion–considering the seasonal variation in demands for transportation
fuel as suggested in [117]–is shown in a 3D plot in Fig. 12 (b).
The forecasted hourly-basis, year-long wholesale electricity price
input data stream, πim(t), obtained using the weighted rolling average
method, is shown as a monthly averaged daily profile in Fig. 13 [118].
More details of the case study site and the complete details on how
the forecasted one-year profiles for climatological, load demand, and
wholesale electricity price data are derived, can be found in Supple
mentary Material (Additional File 3: Case study details).
Table 5 presents the data values and sources for all parameters of the
proposed two-stage, aggregator-mediated, incentive-based DSM frame
work. In addition to the values of the model parameters defined in
Section 3, Table 5 presents the number of customers signed up with each
aggregator, which is denoted by N
j
cust.
Moreover, given the New Zealand government’s aspirations of
electrifying transport to help meet its target of net-zero greenhouse gas
emissions by 2050, as well as the recent government-funded ‘Warmer
Kiwi Homes’ programme offering up to 90% heat pump grants to low-
income home owners, the penetration levels of light-duty FCEVs and
heat pumps were assumed to be 40% and 60%, respectively at the time
of commitment. Accordingly, smart charging of FCEVs and control of
heat pump demand is of utmost importance to smooth and manage the
overall load during peak periods.
5.2. Validation of the proposed demand-side management market
To validate the effectiveness of the proposed two-stage aggregator-
mediated DSM market model, two instances of day-ahead energy man
agement analysis are conducted and the obtained results are compared
with the case where the aggregator-mediated interruptible/curtailable
DR resources are scheduled in a BAU way. Accordingly, the non-market-
driven (BAU) procurement of aggregator-activated interruptible/cur
tailable responsive loads excludes the ability to adaptively update the
incentives offered by the MG operator, based on which the aggregators
post their incentives to the retail DSM market, and subsequently the end-
consumers select their participation rate in load reduction programmes.
More specifically, the MG operator offers a fixed, day-specific rate of
incentive to the aggregators, who also offer fixed levels of incentives to
their customers–for load reduction during the peak hours of electricity
consumption. Subsequently, the end-users and aggregators respond to
the aggregator-determined and MG operator-offered incentive rates,
respectively. In this way, the retail and wholesale DSM markets are
sequentially cleared for the day-specific incentives by stacking the cus
tomers’ and aggregators’ bids, low to high, and allocating demand
reduction schedules to the customers and aggregators in the merit order
irrespective of whether the power shortage is addressed with the best
compromise between load reduction and imported electricity for each
hourly period. Expectedly, as there exists no mechanism to update the
Fig. 11. Monthly mean profile for the estimated total biomass available per
month at the site: Ohakune, New Zealand.
Fig. 12. Forecasted monthly mean 24-h profiles for the energy demand of the town Ohakune: (a) load power demand; and (b) hydrogen demand.
Fig. 13. Forecasted monthly mean 24-h profile for the wholesale power price.
S. Mohseni et al.](https://image.slidesharecdn.com/mohseni2021-211027145406/85/Mohseni2021-20-320.jpg)
![Applied Energy 287 (2021) 116563
21
initial strategy of the MG operator, the efficiency of such a framework is
particularly sensitive to the choice of the MG operator-offered incentive
rate. Hence, the model response is determined for various day-specific
MG operator-offered incentive rates. Accordingly, Table 6 summarises
the results obtained by simulating the above-described BAU interrupt
ible DR mechanism when applied to the DR provision problem at hand in
two extreme scenarios with the MG operator-offered incentive payment
ranging from $0.02/kWh to $0.32/kWh in intervals of $0.02/kWh.
Specifically, the two days that represent the most intense peak and
trough on the year-round, mean daily load profile (consisting of the
mean of the load power demand forecasts for 24 equidistant times in the
course of each continuous 24-hour period of the representative year),
namely July 21st and February 14th, were chosen for scenario analysis.
The table, furthermore, presents the results of the suggested market-
driven interruptible DR model for the extreme days considered.
The following observations can be made from a comparative analysis
of the proposed model and BAU model results presented in Table 6:
1. The systematic updating of the MG operator-offered incentive for
load reduction–for the time-steps at which the system is predicted to
be under stress–using an aggregator-mediated, market-driven DSM
market model, can play a pivotal role in unlocking the full potential
of demand-side resources by finding the economically efficient DR
allocation solutions. In other words, the lack of a systematic
Table 5
Data values and assumption sources for the two-stage, aggregator-mediated, incentive-based demand-side management framework.
Parameter Aggregator
Residential Commercial Industrial Agricultural FCEV-refuelling
δj* Value 0.48 0.51 0.57 0.63 0.76
Source [119] [119] [119] [119] [120]
c
k,j
1 [$/kWh2
] Range [1.08 × 10-3
, 1.15 × 10-3
] [1.04 × 10-3
, 1.07 × 10-3
] [0.99 × 10-3
, 1.03 × 10-3
] [0.95 × 10-3
, 0.98 × 10-3
] [0.91 × 10-3
, 0.94 × 10-3
]
Source†
[121,122] [121,122] [121,122] [121,122] [121,122]
ck,j
2 [$/kWh] Range [11.49 × 10-3
, 11.70 × 10-3
] [11.31 × 10-3
, 11.48 × 10-3
] [11.71 × 10-3
, 11.86 × 10-3
] [11.25 × 10-3
, 11.30 × 10-3
] [11.40 × 10-3
, 11.57 × 10-3
]
Source†
[121,122] [121,122] [121,122] [121,122] [121,122]
d
k,j
full [kWh] Range [8, 30] [20, 100] [100, 200] [30, 65] [5, 30]
Source (this paper) (this paper) (this paper) (this paper) (this paper)
d
k,j
ncr [kWh] Range [2.5, 16.5] [5, 60] [20, 84] [10, 46.2] [4, 25.5]
Source (this paper) (this paper) (this paper) (this paper) (this paper)
N
j
cust
Value(s) 250 65 10 55 {1, 2, …, 150}‡
Source (this paper) (this paper) (this paper) (this paper) (this paper)
*
The load type-dependent DR procurement factor (sectoral elasticity of customer-supplied DR capacity) for the residential, commercial, industrial, and agricultural
loads (normalised to the range [0, 1]) were adjusted in proportion with the weighted average values of unserved energy for various durations of interruption in a New
Zealand context [119], while this factor for the FCEV-refuelling load was adjusted based on the plug-in EVs’ value of lost load reported in [120].
†
The range of values the discomfort tolerance coefficients of customers can take were arbitrarily selected. The chosen values were guided by those used in [121,122]
for the customer outage cost function coefficients for the relevant customer categories. Additionally, the range of sector-wide customer discomfort tolerance co
efficients was normalised with respect to the corresponding load type-dependent DR procurement factor (in an inversely proportional manner).
‡
Since the number of FCEVs that utilise the filling station’s equipment varies with the time of day, it was modelled as a range of possible scenarios, i.e. the number of
vehicles.
Table 6
Comparative analysis of the proposed and BAU realisations of the interruptible DR programme on the extreme days: February 14th and July 21st.
MG operator-offered
incentive* (IMGO)
[$/kWh]
Total daily incentive
payment to the aggregators
(Ip
MGO(
∑
p∈Pd
∑
j∈JD
j,p
LA))
[$/d]
Total daily incentive
payment to the customers
(
∑
p∈Pd
∑
j∈JI
j,p
LAD
j,p
LA) [$/d]
Total daily load reduction
procured by the customers
(
∑
p∈Pd
∑
j∈J
∑
k∈NJ
dk,j,p
)
[kWh/d]
Total daily cost of
electricity imports
(
∑
p∈Pd
costp
im) [$/d]
Total daily operational cost of
the MG
(
∑
p∈Pd
OCp
MG)
∑
p∈Pd
OCp
MG)
[$/d]
Feb. 14th Jul. 21st Feb. 14th Jul. 21st Feb. 14th Jul. 21st Feb. 14th Jul. 21st Feb. 14th Jul. 21st Feb. 14th Jul. 21st
Business-as-usual interruptible DR scheduling approach
0.02 0.02 10.5 43.3 3.9 15.6 525.0 2165.0 870.8 1997.7 881.3 2041.0
0.04 0.04 22.5 102.1 9.2 41.9 562.5 2552.5 863.2 1916.1 885.7 2018.2
0.06 0.06 44.8 162.8 18.8 70.0 746.7 2713.3 824.3 1882.4 869.1 2045.2
0.08 0.08 81.7 390.4 35.9 171.8 1021.3 4880.0 766.6 1427.4 848.3 1817.8
0.1 0.1 180.9 528.5 83.2 232.5 1809.0 5285.0 601.2 1342.3 782.1 1870.8
0.12 0.12 217.1 634.2 91.8 310.8 1809.0 5285.0 601.2 1342.3 818.3 1976.5
0.14 0.14 264.7 787.8 105.9 409.7 1890.7 5627.1 584.1 1270.5 848.8 2058.3
0.16 0.16 302.5 900.3 115.8 459.2 1890.7 5627.1 584.1 1270.5 886.6 2170.8
0.18 0.18 377.3 1013.7 188.7 547.4 2096.1 5631.7 540.9 1269.5 918.2 2283.2
0.2 0.2 421.5 1341.0 219.2 643.7 2107.5 6705.0 538.5 1044.1 960.0 2385.1
0.22 0.22 486.0 1611.0 233.3 757.2 2209.1 7322.7 517.3 914.4 1003.3 2525.4
0.24 0.24 551.8 2093.1 253.8 879.1 2299.2 8721.3 498.3 620.7 1050.1 2713.8
0.26 0.26 708.6 2593.3 311.8 959.5 2725.4 9974.2 408.9 357.6 1117.5 2950.9
0.28 0.28 1040.2 2906.8 436.9 1133.6 3714.9 10381.4 201.2 272.1 1241.4 3178.9
0.3 0.3 1401.8 3503.2 560.7 1191.1 4672.7 11677.3 0 0 1401.8 3503.2
0.32 0.32 1495.2 3736.7 586.2 1195.7 4672.7 11677.3 0 0 1495.2 3736.7
Proposed market-driven interruptible DR scheduling approach
0.17 0.15 566.3 1327.8 230.7 488.6 3253.8 8155.2 50.8 76.9 617.1 1404.7
Values in bold indicate the total daily operational cost of the MG in the best performance of the BAU interruptible DR management framework.
*
Given the variability of the best-strategy incentive offered by the MG operator at different peak hours of the day in the proposed market-driven model, the mean
daily value of the optimal incentive rate offered by the MG operator (Ip,*
MGO) is presented for the proposed model.
S. Mohseni et al.](https://image.slidesharecdn.com/mohseni2021-211027145406/85/Mohseni2021-21-320.jpg)
![Applied Energy 287 (2021) 116563
22
framework to enable the DR programme administrator to vary the
rate of incentive payment to increase or decrease the supply of DR
capacity, either results in an overpayment for access to the DR re
sources, or leads to the under-trading of the responsive loads. More
specifically, the proposed model has outperformed the BAU model by
at least ~21.1% (equating to a saving of $165) and ~22.7%
(equating to a saving of $413.1) in terms of the daily operational cost
of the MG (
∑
p∈Pd
OCp
MG) respectively for the February 14th and July
21st scenarios.
2. The BAU realisation of the interruptible DR programme has failed to
exploit the full potential of the demand-side flexibility resources
available. The most crucial factor underpinning this under-
utilisation of the responsive loads in this model is the lack of inter
action between the MG operator and responsive load aggregators, as
well as between aggregators and end-consumers to dynamically re-
render the incentives for load reduction at different times of the
day. This is evident from Table 6, where increasing the MG operator-
posted incentive rate from $0.1/kWh to $0.12/kWh, and also from
$0.14/kWh to $0.16/kWh, has only led to an increase in the total
daily payment to the aggregators despite no increase in the net load
reduction in both the scenarios considered.
3. In contrast to the proposed model where the daily operational cost of
the system strictly decreases as the MG operator-offered incentive
rate increases up to a saturation point, the BAU model’s response to
variations in the MG operator-offered incentive rate does not tend to
follow a particular pattern. For example, increasing the MG operator-
offered incentive rate from $0.02/kWh to $0.04/kWh in the case of
July 21st has resulted in a reduction of the daily operational cost of
the MG by ~1.1%, then increasing the incentive rate from $0.04/
kWh to $0.06/kWh has increased the daily operational cost of the
MG by ~1.3%, and then increasing the incentive rate from $0.06/
kWh to $0.08/kWh has substantially driven down the daily opera
tional cost of the MG system–to the globally optimal level. Much of
the reason for such an erratic behaviour of the BAU model lies in the
fact that the participation of the aggregators depends on meeting
certain threshold levels of profits. Put differently, increasing the rate
of incentive payments leads to a worthless overpayment unless it
triggers the participation of a further MG customer, provided that a
lower incentive rate than the per-unit cost of electricity import is
deemed sufficient by the customer. However, the interactive DSM
market-clearing mechanism embedded in the proposed DSM market
model–implemented using the proposed interactive value iteration
solution approach (refer to Algorithm 1)–has addressed such a source
of unreliability.
To evaluate the weather-sensitivity of each model, the analysis is
expanded to include all the days in which the interruptible DR pro
gramme is executed. Table 7 summarises the descriptive statistics for the
DR scheduling variables during the hours of peak demand for which a
net energy deficit is predicted. Note the change in temporal resolutions
of the dependent variables compared to Table 6. Specifically, the results
are presented for the morning peak (MP) and evening peak (EP) hours
across the seasons to provide insight into the temporal distribution of
utilising the DR resources.
The table is revealing in the following ways:
1. The DR events occur more frequently in autumn (734 times) and
winter (832 times) than in spring (388 times) and summer (284
times). A comparison of the total number of DR event observations
during the morning and evening peak periods across different sea
sons offers the following insights: (i) two distinctive daily periods of
positive net load demand–the total electric demand on the system
minus local generation–can be identified for autumn and winter;
while (ii) the net load demand in spring and summer is characterised
by one period, namely the MP. This change in the capacity deficit
pattern is mainly driven by weather conditions; the warmer months
reduce the necessity of utilising electric space heating systems. Other
seasonal covariates, including daylight saving and longer daylight
hours in spring and summer, which lead to both lower lighting use
and higher solar PV generation in the early evening, also contribute
to this variation, albeit to a lesser degree.
2. Although the number of DR events that occurred during the MP
period is lower than the corresponding EP period in the colder
months, the average hourly load reduction procured is nearly the
same for the morning and evening peak periods in autumn and
winter. This implies that the profile of the net load demand has a
shorter, sharper peak in the morning, but a longer, flatter peak in the
evening in autumn and winter. This is not only due to the
Table 7
Summary statistics for the DR scheduling variables.
Variable Spring Summer Autumn Winter
MP EP MP EP MP EP MP EP
MG operator-offered incentive [$/kWh] Avg. 0.159 0.202 0.140 0.168 0.120 0.097 0.128 0.183
Med. 0.160 0.200 0.140 0.173 0.120 0.097 0.120 0.189
SD 0.031 0.034 0.026 0.027 0.015 0.038 0.039 0.029
Obs. 291 97 208 76 344 390 400 432
Incentive payment to the aggregators [$/h] Avg. 49.004 111.484 26.866 62.378 48.504 63.166 77.043 147.260
Med. 47.409 101.634 26.492 61.087 48.996 64.636 78.349 147.850
SD 11.852 26.799 4.103 5.268 1.665 5.667 6.336 8.320
Obs. 291 97 208 76 344 390 400 432
Incentive payment to the customers [$/h] Avg. 20.092 51.281 12.105 29.448 22.627 28.039 30.510 67.382
Med. 20.115 52.360 10.606 29.282 20.901 27.432 29.660 67.446
SD 5.570 5.954 6.314 2.940 6.307 3.483 4.240 4.252
Obs. 291 97 208 76 344 390 400 432
Load reduction procured by the customers [kWh] Avg. 308.201 551.901 191.900 371.298 604.200 651.196 801.898 804.699
Med. 311.051 553.074 192.312 371.649 603.094 651.628 804.004 805.741
SD 9.814 11.053 5.593 11.579 6.587 11.687 9.678 6.922
Obs. 291 97 208 76 344 390 400 432
Cost of electricity imports [$/h] Avg. 8.611 15.237 3.985 7.907 5.531 8.402 9.004 17.516
Med. 9.044 15.780 4.238 7.974 5.406 7.941 8.172 17.049
SD 3.531 2.020 0.881 0.187 1.087 1.873 2.556 4.937
Obs. 291 97 208 76 344 390 400 432
Total operational cost of the MG [$/h] Avg. 57.615 126.721 30.851 70.285 54.035 71.568 86.047 164.776
Med. 56.453 117.414 30.730 69.061 54.402 72.577 86.521 164.899
SD 2.618 7.981 2.771 3.217 2.410 3.651 4.206 9.325
Obs. 291 97 208 76 344 390 400 432
S. Mohseni et al.](https://image.slidesharecdn.com/mohseni2021-211027145406/85/Mohseni2021-22-320.jpg)
![Applied Energy 287 (2021) 116563
23
coincidence of the residential load with the start of the business day,
but also the fact that non-dispatchable renewable power generation
from wind and hydro resources is considerably less during the
autumn and winter MP period than the corresponding EP period (see
Fig. 10). Crucially, the proposed non-cooperative game-theoretic DR
scheduling model has yielded reductions in load demand of, on
average, ~24% and ~22% respectively during the winter morning
and evening peak periods. This equates to an average hourly energy
reduction of ~802 kWh in the MP and ~805 kWh in the EP. In
summer, this percentage decreases to ~13% (192 kWh) in the MP,
and ~15% (371 kWh) in the EP period.
3. Defining the data skewness as (mean–median) / standard deviation,
it can be shown that the skewness values of the ‘cost of electricity
imports’ and the ‘incentive payments made by the utility to the
aggregators’ datasets have opposite signs at all the eight quarterly
time intervals. For example, the skewness values of the above
mentioned datasets for the winter MP period are 0.326 and − 0.206,
respectively. Accordingly, the mean of the former dataset is greater
than its median (i.e., the dataset distribution is positively skewed),
whereas the mean of the latter dataset is less than its median (i.e.,
negatively skewed). This suggests that the optimal trade-off between
imported power and utilised DR capacity tends to follow an
approximately consistent pattern during each period of peak elec
tricity use. This finding corroborates the robustness and validity of
the proposed non-cooperative game-theoretic DSM approach in
producing the best compromise between the imported power and
elicited DR capacity.
Moreover, Table 8 provides a statistical evaluation of the efficacy of
the proposed market-based integration (MBI) of responsive loads using
non-cooperative game theory as compared to the BAU model in the four
seasons. Note that, for reasons of space, the modelling results are not
broken down into the morning and evening peak periods.
From Table 8, emerge a number of key statistically valid evidence to
support the superiority of the proposed game-theoretic DR scheduling
model to the BAU interruptible programmes:
1. The proposed aggregator-mediated, market-based DR programme is
able to unlock new sources of economic value that are inaccessible by
the BAU-DR scheduling approach. This has resulted in a ~17%
(equating to ~$39 k) reduction in the operational cost of the MG in
the baseline year. In large part, this is because the proposed model
ensures a level playing field for all the DR providers and equitably
allocates the benefits of third-party DR aggregation, whilst addi
tionally providing a platform for the MG operator, DRAs, and end-
consumers to mutually optimise their portfolios and determine the
lowest operational costs.
2. A comparison of the seasonal performance of the proposed model
and the BAU approach reveals that, on average, the DR resources are
under-utilised in autumn and winter, whilst additionally the DR
providers are over-compensated in spring and summer in the BAU
approach. More specifically, in contrary to the obtained results from
the proposed model, where the distributions of the ‘incentive pay
ments to the aggregators’ and the ‘cost of electricity imports’ data are
oppositely skewed, they have similar skewness patterns in the BAU
approach. The BAU approach’s results indicate that both of the
above-mentioned distributions are skewed to the left (i.e., most of
the observations lie to the right of the mean) in spring and summer,
whereas they are both positively-skewed (i.e., most of the observa
tions lie to the left of the mean) in autumn and winter. A major
explanation for this is the BAU interruptible service approach’s
incapability to provide a more-targeted, non-prespecified incentive
price signal that fluctuates hourly reflecting changes in the wholesale
prices of electricity.
3. While the percentage of incentive payments to the customers to the
incentive prices received by the aggregators remains nearly the same
across the seasons in the proposed game-theoretic modelling results
(within the range of approximately 43–46%), the percentage varies
significantly from season to season if the problem is solved in a BAU
way. In particular, the BAU modelling results yield the highest utility
margin for the customers (with the customers’ share of the utility
incentives of ~53%) during the winter months (June to August)
when their use of electricity for heating contributes to high network
loads. On the other hand, the per-unit profit of the DRAs is largest
during the summer months (December to February) when electric
Table 8
Comparative statistical analysis of the proposed and BAU-DR scheduling models.
Variable Spring Summer Autumn Winter
BAU* MBI BAU* MBI BAU* MBI BAU* MBI
MG operator-offered incentive [$/kWh] Avg. 0.147 0.170 0.112 0.148 0.051 0.109 0.073 0.155
Med. 0.152 0.159 0.108 0.146 0.045 0.110 0.074 0.156
SD 0.017 0.033 0.017 0.029 0.025 0.031 0.030 0.028
Obs. 388 388 284 284 734 734 832 832
Incentive payment to the aggregators [$/h] Avg. 16.821 64.624 9.725 36.369 13.342 55.595 35.537 111.054
Med. 16.874 55.729 9.777 29.113 12.242 51.287 33.964 87.232
SD 1.670 31.920 0.863 5.482 4.387 8.279 5.125 35.922
Obs. 388 388 284 284 734 734 832 832
Incentive payment to the customers [$/h] Avg. 6.390 27.889 3.112 16.746 6.538 25.245 18.835 48.370
Med. 6.454 23.463 3.223 13.639 6.458 25.131 17.657 36.607
SD 0.574 14.624 0.279 9.624 1.837 5.914 10.386 18.948
Obs. 388 388 284 284 734 734 832 832
Load reduction procured by the customers [kWh] Avg. 104.263 369.126 80.929 239.908 240.608 629.171 465.260 803.352
Med. 104.979 314.339 81.951 194.866 238.388 512.418 463.248 714.444
SD 7.450 106.143 4.920 79.940 14.018 123.846 16.353 101.754
Obs. 388 388 284 284 734 734 832 832
Cost of electricity imports [$/h] Avg. 76.462 10.268 41.742 5.034 61.044 6.919 113.148 13.127
Med. 76.845 10.633 42.059 5.074 31.421 6.948 55.583 13.392
SD 1.902 4.274 2.314 0.148 6.255 2.935 5.206 3.129
Obs. 388 388 284 284 734 734 832 832
Total operational cost of the MG [$/h] Avg. 93.283 74.892 51.467 41.404 74.386 62.515 148.685 124.181
Med. 93.836 58.145 52.657 33.028 72.784 57.739 146.465 94.166
SD 1.766 30.248 3.311 17.753 7.011 9.224 7.104 40.021
Obs. 388 388 284 284 734 734 832 832
*
The BAU results represent the business-as-usual model’s best performance out of different daily utility-offered incentives ranging from $0.02/kWh to $0.32/kWh in
intervals of $0.02/kWh.
S. Mohseni et al.](https://image.slidesharecdn.com/mohseni2021-211027145406/85/Mohseni2021-23-320.jpg)
![Applied Energy 287 (2021) 116563
24
heating cannot provide DR, which reduces the customers’ share of
revenues to as low as ~32%. This indicates the BAU approach’s
failure to provide a fair division of the utility-offered financial in
centives between the DRAs and their corresponding customers,
which results in the overall DR underperformance.
As these observations are shown to remain valid for the year-round
operation of the system, their positive impact on the lifetime cost and
cost-effectiveness of the conceptualised system is discussed in the next
sub-section.
5.3. Optimal equipment capacity-planning results
To evaluate the effectiveness of the proposed DR scheduling frame
work in reducing the whole-life cost of MGs, the equipment capacity-
planning of the conceptualised MG was carried out under three cases:
(1) taking a BAU (static) interruptible load approach to managing the
smaller DR resources (as detailed in Section 5.2), (2) using the proposed
market-based (dynamic) integration of the aggregator-mediated inter
ruptible responsive loads (presented in Section 3), and (3) not imple
menting any DSM strategies. Tables 9 and 10 present the MG investment
planning model results under the above-mentioned three cases, which
are respectively denoted by ‘BAU-DR’, ‘MBI-DR’, and ‘NO-DR’. Specif
ically, Table 9 details a breakdown of the optimised cost components
included in the life-cycle analysis of the MG system (see Eq. (39)), while
Table 10 provides the optimum size of the MG equipment, which are the
main decision variables of the optimisation problem. Note that the
optimisation model results are adjusted for the value of biomass feed
stock. To this end, the total cost associated with the pelletisation of
blended biomass feedstocks–agricultural and woody biomass–was
considered to be $72/tonne of pellets [123]. The case study site’s nat
ural endowment of forest biomass together with its temperate climate
that is ideally suited to the agricultural activities, narrows, to a
considerable extent, the feedstock supply uncertainty bounds. This
provides strong support for taking an exogenous approach to account for
the biomass feedstock costs–in the post-optimisation phase.
It is also noteworthy that the results reported in the tables represent
the best-case performance of the MFOA out of 30 independent trials.
Moreover, to demonstrate the adequacy of the maximum number of it
erations, and the number of search agents considered, the convergence
curves of the MFOA in its best and worst overall performances for each
of the above-mentioned cases are shown in Fig. 14.
The comparative results presented in Table 9 reveal that the pro
posed market-based modelling of the interruptible DSM processes in the
planning phase of the conceptualised MG reduces the estimated whole-
life cost of the system by at least 21% and up to a maximum of 32% (with
an incentive resolution of $0.02/kWh), as compared to the BAU inter
ruptible DR-integrated and non-DR-integrated MG planning cases,
respectively.
Table 9
Breakdown of the total discounted system cost under different DR provision strategies.
Cost component Cost subcomponent Simulation case
MBI-DR BAU-
DR
NO-DR
Total discounted equipment-related costs ((
∑
c∈CNPCc
20− yr
) + NPCI
20− yr
) [$] 18.25 m 21.88 m 25.62 m
Total discounted MG operational costs (NPV(
20− yr
∑8760
t=1 OCMG(t))) Total discounted incentive payment to the aggregators
(NPV
20− yr
(
∑8760
t=1 IMGO(t)
∑
j∈JD
j
LA(t))) [$]
3.99 m 3.48 m −
Total discounted cost of electricity imports (NPV
20− yr
(
∑8760
t=1 costim(t))) [$] 0.46 m 2.76 m 7.46 m
Total discounted FCEV2G electricity provision costs
(NPV(
20− yr
∑8760
t=1 πFCEV2GPFCEV2G(t))) [$]
0.42 m 0.49 m 0.50 m
Total discounted operating costs of the biopower plant Total discounted emission credits (NPV
20− yr
(
∑8760
t=1 costem(t))) [$] 0.52 m 0.57 m 0.62 m
Total discounted biomass feedstock costs* (NPV
20− yr
(72 ×
∑8760
t=1 MBP(t)))
[$]
0.49 m 0.54 m 0.58 m
Total discounted income derived from electricity exports
(− NPV
20− yr
(
∑8760
t=1 incomeex(t))) [$]
− 2.41
m
− 2.42
m
− 2.97
m
Whole-life cost of the system (WLC) [$] 21.72 m 27.3 m 31.81 m
*
The total cost of the biomass feedstock is not systematically affected by changes in the endogenous variables of the model in this study. That is, the total cost
imposed by the biomass feedstock was calculated outside the optimisation model and the results were then corrected accordingly.
Table 10
Size of the MG equipment in the cost-minimal solution under different DR
provision strategies.
Component Simulation case
MBI-DR BAU-DR NO-DR
PV plant NPV [no.] 3,594 3,690 4,742
STDEC* [%] 3.54 3.04 3.33
Wind plant NWT [no.] 4 5 6
STDEC* [%] 24.11 26.35 26.73
Micro-hydro power plant NMH [no.] 6 6 6
STDEC* [%] 1.91 1.59 1.36
Biopower plant NBP [no.] 4 4 7
STDEC* [%] 0.77 0.64 0.96
Transformer NT [kVA] 310 320 329
STDEC* [%] 0.11 0.10 0.08
Hydrogen tank NHT [kg] 6,079 7,904 9,168
STDEC* [%] 16.93 18.11 18.16
Electrolyser NE [no.] 122 144 157
STDEC* [%] 4.14 4.08 3.80
Fuel cell NFC [no.] 238 378 440
STDEC* [%] 6.75 8.58 8.66
Battery bank N1600 [no.] 2 2 2
N400 [no.] 0 1 2
N100 [no.] 2 0 3
STDEC* [%] 17.49 15.06 16.00
Super-capacitor bank NSC [no.] 1,982 2,090 2,136
STDEC* [%] 14.53 12.61 11.01
FCEV2G setup NFCEV2G [kW] 504 573 608
STDEC* [%] 0.57 0.53 0.49
Hydrogen station NS [kg-H2/h] 6.14 7.94 9.15
STDEC* [%] 0.42 0.45 0.45
Inverter N900 [no.] 5 6 7
N115 [no.] 2 3 5
N33 [no.] 1 3 1
STDEC* [%] 8.73 8.86 8.97
*
STDEC stands for the share of the total discounted equipment-related costs,
which can be expressed explicitly in mathematical terms as ((
∑
c∈CNPCc
20− yr
) +
NPCI
20− yr
).
S. Mohseni et al.](https://image.slidesharecdn.com/mohseni2021-211027145406/85/Mohseni2021-24-320.jpg)
![Applied Energy 287 (2021) 116563
25
Furthermore, the results summarised in Tables 9 and 10 are indica
tive of the high efficiency of the proposed model for the aggregator-
activated, responsive load-aware MG capacity design in the following
ways:
1. While the total discounted equipment-related costs in the BAU case
are higher by ~20% than the MBI case, the total discounted income
derived from electricity exports has remained at nearly the same
level. This is because the majority of this extra cost is spent on the
backup power equipment, the energy output of which, according to
the MG’s hourly operational strategy in Fig. 5, cannot be sold to the
main grid–for energy efficiency considerations. To examine the
robustness of this assumption, a further unreported model was run in
both the MBI and BAU simulation cases, where the backup power
was allowed to be sold into the utility grid, while maintaining the
rest of the model unchanged. A comparative analysis of the results of
the two models for the investigated test-case is presented in Sup
plementary Material (Additional File 4: Table S1). The results show
that the relative difference of the total discounted equipment-related
costs in the MBI and BAU cases reduces to ~15%, from ~20% for the
base-case model, when the sale of backup power into the grid is not
prohibited. The key factor underpinning this change is that the un
reported optimisation model that supports the sale of backup power
to the main grid finds the opportunity to arbitrage intertemporal
differences in wholesale prices and buy-back rates. The unreported
model, therefore, increases the proportion of total nominal storage to
generation capacity in the optimal equipment capacity configuration
as compared to the base-case model. More specifically, the propor
tion of the share of the total back-up components’ capacity to the
share of the total primary generation components’ capacity in the
system’s whole-life cost increases from 1.97 and 1.85 to 3.51 and
3.22 in the MBI and BAU model realisations, respectively, at rela
tively modest extra total equipment-related costs–that is, ~9% and
~5%, respectively. This, however, increases the MG’s total net in
come from the exchange of energy with the utility grid by ~76% and
~429%, respectively, in the aforementioned two cases. As a conse
quence, the MG’s whole-life cost reduces by ~3% and ~5%,
respectively, in the two cases mentioned above–but at the cost of
higher total energy dissipated as a result of increased energy con
version rocesses.
2. The total discounted income derived from electricity exports in the
non-DR-integrated case is higher by ~23% in comparison with the
base case, which is mainly due to the increased excess of renewable
energy generation in low-demand periods. Note that the export of
energy is seen merely as a means to avoid spillage of non-
dispatchable renewable energy, and the low export tariff makes it
irrational for the solution algorithm to optimise the capacity of the
MG equipment for energy export purposes. It should not be over
looked, however, that energy export made a fair contribution to the
cost-efficiency of the proposed MG system in all of the cases studied.
It is also important to note that the solution algorithm, in the MBI
case, has almost always avoided buying and storing electricity from
the upstream grid at times of low demand, but the surplus renewable
energy is sold to the grid at these times due to: (1) the higher level of
feed-in-tariff than the system’s levelised cost of energy (LCOE) at
most of the off-peak times of the year, and (2) the fact that the battery
and SC banks soon reach their maximum capacity limits when the
MG system is lightly loaded. This is while the total discounted cost of
electricity imports occupies ~10% and ~24% of the total discounted
system costs in the BAU DR-integrated and non-DR-integrated cases,
respectively.
3. In all of the investigated cases, the optimised size of the electrolyser
unit is considerably lower compared to those in established size
combinations–of electrolyser to hydrogen reservoir to fuel cell–in the
literature (see, for example, [124,125,126]). This is due to the spe
cific conditions of the case study site, where load demand is subject
to a high degree of seasonality. Accordingly, an electrolyser of lower
capacity is sufficient for the purpose–since the hydrogen tank can be
filled gradually during the low season, from October through June.
That is also why the optimum capacity of the electrolyser experi
enced the least changes among the backup power equipment in the
three scenarios investigated.
4. As planned, the fuel cell generation using the stored hydrogen has
accounted for seasonal load levelling. The optimal capacity of the
fuel cell is more highly impacted by the proposed interruptible DR
implementation as compared to the battery and SC banks. This
observation implies that peak load shaving–fulfilled by exploiting
the responsive loads–has had a substantial role in smoothing out the
seasonal variation in load demand, and, in turn, improving the load
factor of the annual load power demand profile. In other words,
much of the suggested DR scheduling strategy’s positive impact on
the cost-efficiency of the conceptualised MG is derived from its
implementation in the winter high season. This also explains the
marked increase in the size of the WT, hydrogen tank, fuel cell, and
the electrical loads’ inverter–as the main drivers of increasing the
equipment-related costs–when the DR is implemented in a BAU
manner, or, more significantly, when no DR scheduling process is
implemented. To provide a clearer picture of the impact of the pro
posed DSM model on the load power demand data fed into the
optimal capacity planning algorithm, the monthly mean 24-h
Fig. 14. Convergence process of the MFOA in its best and worst runs throughout 30 simulation cases.
S. Mohseni et al.](https://image.slidesharecdn.com/mohseni2021-211027145406/85/Mohseni2021-25-320.jpg)
![Applied Energy 287 (2021) 116563
26
electricity consumption profile is presented in Fig. 15 for the simu
lation cases under study. According to the figure, realising the pro
posed DSM model under the BAU and MBI cases shaves ~24% and
~38% off the maximum peak power demand compared to the NO-
DR case, respectively. This, consequently, increases the load factor
from 0.25 in the NO-DR case to 0.31 and 0.35 in the BAU and MBI
simulation cases, respectively.
5. The relatively low share of biomass in the optimised energy resource
mix, in spite of its vast potential in the site under study (see Fig. 11),
is revealing in two ways: (1) the solution algorithm has succeeded in
restricting the bioenergy use to a sustainable level by imposing an
emission penalty and, more importantly, (2) it gives credence to the
idea that biomass resources need to be deployed in a way that con
tributes primarily to energy security–in favour of a deep green
approach to renewable energy system planning [127]. More specif
ically, the biopower plant in the conceptualised MG plays a critical
role in improving the system’s self-sufficiency, as it is the only dis
patchable power generation unit in the system.
5.3.1. Financial appraisal
To demonstrate the financial sustainability of the long-term invest
ment proposal, this sub-section compares the LCOE of the MG with the
existing retail electricity prices at the site, as well as the LCOE reported
in the literature for the most similar projects. More specifically, the
project was benchmarked against the studies in the literature that met
the following three criteria: (1) a self-sufficiency ratio of at least 85% if
the system is grid-connected, (2) powered by 100% renewable energy,
and (3) tailored to the electrification of small- to medium-scale
communities.
The LCOE, in the MG context, represents the average revenue per
unit of energy generated that would be required to recoup the lifetime
costs of the system. Accordingly, the LCOE [$/kWh] of the MG under
study can be calculated as follows [128]:
LCOE =
WLC
∑PL
n=1
(
∑8760
t=1
PL(t)+
∑8760
t=1
PS(t))Δt
(1+ir)n
, (49)
where PL represents the project lifetime [years], ir denotes the real in
terest rate per annum [%], and the terms
∑8760
t=1 PL(t) and
∑8760
t=1 PS(t)
respectively denote the total annual electric and hydrogen power de
mand on the MG, which are discounted to reflect the NPV of future
energy flows.
By solving Eq. (49), the LCOE of the proposed MG is found to be
$0.08/kWh, while the most recent yearly average retail price of elec
tricity is as high as $0.22/kWh at the studied site. That is, implementing
the proposed MG system is expected to realise savings of at least 64% in
the community’s energy costs if financed as a community-owned
renewable energy project. Note that the MG’s LCOE is calculated for
the case where the aggregator-mediated demand-side flexibility re
sources are scheduled using the proposed non-cooperative game-theo
retic DSM framework–and the whole-life cost of the MG is $21.72 m.
Table 11 benchmarks the conceptualised system in terms of the LCOE
with the most similar projects in the literature.
As can be seen from Table 11, the LCOE of the simulated MG is highly
competitive with that of the best value reported in the recent literature
for a community-scale, 100%-renewable electrification project. Add to
this the fact that a carbon-free, hydrogen-based, light-duty trans
portation fleet is integrated into the proposed MG, making it one of the
first of its kind. This provides additional support for the economic sus
tainability of the conceptualised community energy system.
Based on the above premises, the modelled MG provides an evidence
base to inform the energy sector and climate change policy, infrastruc
ture providers, and the wider modelling community of both the tech
nical feasibility and economic viability of leveraging the potential
synergies in the integration of energy networks for electricity, heating,
and transport to realise economy-wide deep decarbonisation.
6. Conclusions
The projections on the uptake of demand response programmes used
in the long-term capital infrastructure planning of sustainable energy
systems are substantially influenced by the biases and preferences of
end-consumers, which can be modelled in terms of the elasticity of the
customer supply of demand response capacity. This study, one of the
first to provide an understanding of end-consumer behavioural traits in
long-term demand-side management schemes, developed a comfort-
aware, demand response-integrated optimisation model for equipment
capacity-planning of renewable energy systems. To this end, the study
developed a two-stage, aggregator-mediated, non-cooperative game-
theoretic demand-side management market design to improve the ac
curacy of the long-term forecasts of end-users’ participation in
incentive-directed demand response programmes. The proposed model
provides an effective framework for improving the accuracy of
Fig. 15. Comparison of the monthly mean daily profile for load power demand in different simulation cases.
S. Mohseni et al.](https://image.slidesharecdn.com/mohseni2021-211027145406/85/Mohseni2021-26-320.jpg)
![Applied Energy 287 (2021) 116563
27
investment assessments made for demand response-aided energy sys
tems by adopting the endogenous Stackelberg leader–follower re
lationships in two stages, namely: first, for interactions between the
micro-grid operator and responsive load aggregators, and second, for
aggregator-customer exchanges. Moreover, the devised model success
fully generalised the long-term, community-level renewable energy
system design problem in the following four areas:
1. It guaranteed a level playing field for a variety of clean energy
technologies–in the interest of energy diversification–where the use
of biomass resources is limited to a sustainable level by imposing a
new constraint term.
2. It implemented the potential of cross-vector integration (in partic
ular, power-to-gas technology) in conjunction with the value of fuel
cell electric vehicles in vehicle-to-grid operation to improve the
flexibility of energy systems with deep penetration of renewables.
3. It allowed for a meta-heuristic solution algorithm based on the moth-
flame optimisation algorithm to find the cost-optimal mix of micro-
grid assets, whilst facilitating long-term decision-making on the de
livery of aggregator-mediated incentive-responsive loads using a
realistic example. The use of a case study illustrated the application
of the model in the town of Ohakune, New Zealand, demonstrating
that many of the challenges for integrating a 100%-renewable energy
system can be surmounted.
4. The suggested solution algorithm was also shown to be efficient in
nearing the formulated problem’s globally optimum solution. In
addition, a comparative analysis of the proposed market-driven and
business-as-usual realisations of the interruptible load programme
verified the validity of the proposed modelling framework as a de
cision support tool for utilities to make reliable forecasts about the
engagement of different classes of end-consumers in demand
response programmes. This is particularly important when designing
greenfield renewable energy systems, or as micro-grids are used to
increase the penetration of responsive loads.
The numeric results obtained from the model’s application to the
test-case system of Ohakune have revealed two novel insights:
1. The use of the proposed two-stage demand-side management market
design for the projection of flexible demand resources brings higher-
order information about micro-grid operator-aggregators-customers
interactions into the analysis, which can be leveraged towards
improving the economic viability of renewable energy systems.
Notably, as compared to the case where demand-side resources are
managed using a business-as-usual interruptible load approach, the
model results have indicated that a cost saving of at least 21%
(equating to approximately $5.5 m) can be generated for the simu
lated micro-grid in Ohakune, while imposing the same discomfort
cost on end-users.
2. The large-scale supply of demand-side flexibility resources, enabled
by demand response aggregators, has great potential in reducing the
estimated life-cycle cost of sustainable energy systems. Specifically,
the evidence from this study demonstrates that assisting the con
ceptualised micro-grid with incentive-driven, market-directed
demand-side management processes reduces the total discounted
system costs by circa 32% (equating to around $10 m in this case
study). In this light, a thorough analysis of the value of lost load to
the target customers–in the interest of improving the accuracy of the
forecasted willingness of the end-users to deliver their demand
response resources–is of paramount importance in the design phase
of all-renewable micro-grids. This is especially true for the devel
opment of first-access energy systems in remote areas where the
values of unserved energy are expected to be lower than those esti
mated for urban and industrial customers.
In conclusion, this paper has shown that capturing the flexible de
mand potential of small- to medium-scale customers during the planning
phases of a hydrogen-based grid-connected micro-grid system can pave
the way toward achieving greater energy independence, energy de
mocracy, and energy security in rural and semi-urban areas in a cost-
effective and environmentally efficient way.
CRediT authorship contribution statement
Soheil Mohseni: Conceptualization, Methodology, Data curation,
Formal analysis, Investigation, Resources, Software, Validation, Visu
alization, Writing - original draft. Alan C. Brent: Supervision, Project
administration, Writing - review & editing. Scott Kelly: Supervision,
Writing - review & editing. Will N. Browne: Supervision, Writing - re
view & editing. Daniel Burmester: Supervision, Writing - review &
editing.
Declaration of Competing Interest
The authors declare that they have no known competing financial
interests or personal relationships that could have appeared to influence
the work reported in this paper.
Table 11
Comparative evaluation of the proposed MG’s LCOE against those of the comparable schemes.
Reference Renewable energy system architecture Case study site(s) Climatic conditions Unsubsidised LCOE
[$/kWh]*, †
Hosseinalizadeh et al.,
2016 [129]
An on-grid PV/WT/BESS/FC MG Four villages in Iran, namely Moaleman,
Ghadamgah, Marvdasht, and Nikouyeh
Diverse climatic
conditions
0.54–1.60
Shang et al., 2016 [130] An insular PV/WT/BESS/DG MG An unnamed island near Singapore Tropical/equatorial 0.14
Chauhan and Saini, 2017
[28]
A stand-alone PV/WT/BESS/DG/BP/
MHPP MG
Chamoli district, Uttarakhand state, India Warm temperate 0.07–0.10
Fu et al., 2018 [131] Stand-alone solar PV systems U.S.-wide Diverse climatic
conditions
0.13–0.16
Li, 2019 [132] A grid-independent PV/BESS/FC MG A community centre in Kunming, China Humid subtropical 1.55
Rezk et al., 2019 [133] A grid-independent PV/FC hybrid
renewable energy system
The city of Minya, Egypt Mediterranean 0.06
This study A grid-tied PV/WT/MHPP/BP/FC/
BESS/SC MG
The town of Ohakune, New Zealand Temperate 0.08
Key: BESS = Battery Energy Storage System, BP = Biopower Plant, DG = Diesel Generator, FC = Fuel Cell, LCOE = Levelised Cost of Energy, MG = Micro-Grid, MHPP
= Micro-Hydro Power Plant, PV = Photovoltaic, SC = Super-Capacitor, WT = Wind turbine.
†
Where appropriate, the LCOE values were adjusted to 2019 U.S. dollars.
*
For cases where different configurations of the proposed system are investigated, or the conceptualised system is optimised under different climatic conditions, or
the optimisation process is carried out in a multi-objective search space or in a stochastic way, the value of LCOE is reported as a range, rather than a certain value.
S. Mohseni et al.](https://image.slidesharecdn.com/mohseni2021-211027145406/85/Mohseni2021-27-320.jpg)
![Applied Energy 287 (2021) 116563
28
Acknowledgements
The authors would like to acknowledge excellent comments and
suggestions on the categorisation of load demand from the audience of
the 2020 17th
International Conference on the European Energy Market
(EEM) in Stockholm, Sweden, as well as constructive comments from the
anonymous reviewers.
Appendix A. Supplementary material
Supplementary data to this article can be found online at https://doi.
org/10.1016/j.apenergy.2021.116563.
References
[1] Hui H, Ding Y, Shi Q, Li F, Song Y, Yan J. 5G network-based Internet of Things for
demand response in smart grid: a survey on application potential. Appl Energy
2020;257:113972. https://doi.org/10.1016/j.apenergy.2019.113972.
[2] Mohseni S, Moghaddas-Tafreshi SM. A multi-agent system for optimal sizing of a
cooperative self-sustainable multi-carrier microgrid. Sustain Cities Soc 2018;38:
452–65. https://doi.org/10.1016/j.scs.2018.01.016.
[3] Niu J, Tian Z, Zhu J, Yue L. Implementation of a price-driven demand response in
a distributed energy system with multi-energy flexibility measures. Energy
Convers Manag 2020;208:112575. https://doi.org/10.1016/j.
enconman.2020.112575.
[4] Ihsan A, Jeppesen M, Brear MJ. Impact of demand response on the optimal,
techno-economic performance of a hybrid, renewable energy power plant. Appl
Energy 2019;238:972–84. https://doi.org/10.1016/j.apenergy.2019.01.090.
[5] Modarresi Ghazvini A, Olamaei J. Optimal sizing of autonomous hybrid PV
system with considerations for V2G parking lot as controllable load based on a
heuristic optimization algorithm. Sol Energy 2019;184:30–9. https://doi.org/
10.1016/j.solener.2019.03.087.
[6] Jing Z, Zhu J, Hu R. Sizing optimization for island microgrid with pumped
storage system considering demand response. J Mod Power Syst Clean Energy
2018;6(4):791–801. https://doi.org/10.1007/s40565-017-0349-1.
[7] Kim H, Bae J, Baek S, Nam D, Cho H, Chang HJ. Comparative analysis between
the Government Micro-Grid Plan and computer simulation results Based on real
data: the practical case for a South Korean Island. Sustainability 2017;9(2):197.
https://doi.org/10.3390/su9020197.
[8] Amir V, Azimian M. Dynamic multi-carrier microgrid deployment under
uncertainty. Appl Energy 2020;260:114293. https://doi.org/10.1016/j.
apenergy.2019.114293.
[9] Gelazanskas L, Gamage KAA. Demand side management in smart grid: a review
and proposals for future direction. Sustain Cities Soc 2014;11:22–30. https://doi.
org/10.1016/j.scs.2013.11.001.
[10] Haider HT, See OH, Elmenreich W. A review of residential demand response of
smart grid. Renew Sustain Energy Rev 2016;59:166–78. https://doi.org/
10.1016/j.rser.2016.01.016.
[11] Esther BP, Kumar KS. A survey on residential Demand Side Management
architecture, approaches, optimization models and methods. Renew Sustain
Energy Rev 2016;59:342–51. https://doi.org/10.1016/j.rser.2015.12.282.
[12] Wang J, Zhong H, Ma Z, Xia Q, Kang C. Review and prospect of integrated
demand response in the multi-energy system. Appl Energy 2017;202:772–82.
https://doi.org/10.1016/j.apenergy.2017.05.150.
[13] Robert FC, Sisodia GS, Gopalan S. A critical review on the utilization of storage
and demand response for the implementation of renewable energy microgrids.
Sustain Cities Soc 2018;40:735–45. https://doi.org/10.1016/j.scs.2018.04.008.
[14] Jordehi AR. Optimisation of demand response in electric power systems, a
review. Renew Sustain Energy Rev 2019;103:308–19. https://doi.org/10.1016/j.
rser.2018.12.054.
[15] Behrangrad M. A review of demand side management business models in the
electricity market. Renew Sustain Energy Rev 2015;47:270–83. https://doi.org/
10.1016/j.rser.2015.03.033.
[16] Jin X, Wu Q, Jia H. Local flexibility markets: literature review on concepts,
models and clearing methods. Appl Energy 2020;261:114384. https://doi.org/
10.1016/j.apenergy.2019.114387.
[17] Kahrobaee S, Asgarpoor S, Qiao W. Optimum sizing of distributed generation and
storage capacity in smart households. IEEE Trans Smart Grid 2013;4(4):
1791–801. https://doi.org/10.1109/TSG.2013.2278783.
[18] Yu L, Li YP, Huang GH, An CJ. A robust flexible-probabilistic programming
method for planning municipal energy system with considering peak-electricity
price and electric vehicle. Energy Convers Manag 2017;137:97–112. https://doi.
org/10.1016/j.enconman.2017.01.028.
[19] Varasteh F, Nazar MS, Heidari A, Shafie-khah M, Catalão JPS. Distributed energy
resource and network expansion planning of a CCHP based active microgrid
considering demand response programs. Energy 2019;172:79–105. https://doi.
org/10.1016/j.energy.2019.01.015.
[20] Cardoso G, Stadler M, Bozchalui MC, Sharma R, Marnay C, Barbosa-Póvoa A,
et al. Optimal investment and scheduling of distributed energy resources with
uncertainty in electric vehicle driving schedules. Energy 2014;64:17–30. https://
doi.org/10.1016/j.energy.2013.10.092.
[21] Hosseinnia H, Nazarpour D, Talavat V. Multi-objective optimization framework
for optimal planning of the microgrid (MG) under employing demand response
program (DRP). J Ambient Intell Humaniz Comput 2019;10(7):2709–30. https://
doi.org/10.1007/s12652-018-0977-y.
[22] Moghaddas-Tafreshi SM, Jafari M, Mohseni S, Kelly S. Optimal operation of an
energy hub considering the uncertainty associated with the power consumption
of plug-in hybrid electric vehicles using information gap decision theory. Int J
Electr Power Energy Syst 2019;112:92–108. https://doi.org/10.1016/j.
ijepes.2019.04.040.
[23] Martins VF, Borges CLT. Active distribution network integrated planning
incorporating distributed generation and load response uncertainties. IEEE Trans
Power Syst 2011;26(4):2164–72. https://doi.org/10.1109/
TPWRS.2011.2122347.
[24] Zhu L, Yan Z, Lee WJ, Yang X, Fu Y, Cao W. Direct load control in microgrids to
enhance the performance of integrated resources planning. IEEE Trans Ind Appl
2015;51(5):3553–60. https://doi.org/10.1109/TIA.2015.2413960.
[25] Atia R, Yamada N. Sizing and analysis of renewable energy and battery systems in
residential microgrids. IEEE Trans Smart Grid 2016;7(3):1204–13. https://doi.
org/10.1109/TSG.2016.2519541.
[26] Pazouki S, Haghifam MR. Optimal planning and scheduling of energy hub in
presence of wind, storage and demand response under uncertainty. Int J Electr
Power Energy Syst 2016;80:219–39. https://doi.org/10.1016/j.
ijepes.2016.01.044.
[27] Schachter JA, Mancarella P, Moriarty J, Shaw R. Flexible investment under
uncertainty in smart distribution networks with demand side response:
assessment framework and practical implementation. Energy Policy 2016;97:
439–49. https://doi.org/10.1016/j.enpol.2016.07.038.
[28] Chauhan A, Saini RP. Size optimization and demand response of a stand-alone
integrated renewable energy system. Energy 2017;124:59–73. https://doi.org/
10.1016/j.energy.2017.02.049.
[29] Nojavan S, Majidi M, Esfetanaj NN. An efficient cost-reliability optimization
model for optimal siting and sizing of energy storage system in a microgrid in the
presence of responsible load management. Energy 2017;139:89–97. https://doi.
org/10.1016/j.energy.2017.07.148.
[30] Amrollahi MH, Bathaee SMT. Techno-economic optimization of hybrid
photovoltaic/wind generation together with energy storage system in a stand-
alone micro-grid subjected to demand response. Appl Energy 2017;202:66–77.
https://doi.org/10.1016/j.apenergy.2017.05.116.
[31] Chen J, Zhang W, Li J, Zhang W, Liu Y, Zhao B, et al. Optimal sizing for grid-tied
microgrids with consideration of joint optimization of planning and operation.
IEEE Trans Sustain Energy 2018;9(1):237–48. https://doi.org/10.1109/
TSTE.2017.2724583.
[32] Zheng Y, Jenkins BM, Kornbluth K, Kendall A, Træholt C. Optimization of a
biomass-integrated renewable energy microgrid with demand side management
under uncertainty. Appl Energy 2018;230:836–44. https://doi.org/10.1016/j.
apenergy.2018.09.015.
[33] Xiao H, Pei W, Dong Z, Kong L. Bi-level planning for integrated energy systems
incorporating demand response and energy storage under uncertain
environments using novel metamodel. CSEE J Power Energy Syst 2018;4(2):
155–67. https://doi.org/10.17775/CSEEJPES.2017.01260.
[34] Husein M, Chung I-Y. Optimal design and financial feasibility of a university
campus microgrid considering renewable energy incentives. Appl Energy 2018;
225:273–89. https://doi.org/10.1016/j.apenergy.2018.05.036.
[35] Gazijahani FS, Salehi J. Reliability constrained two-stage optimization of multiple
renewable-based microgrids incorporating critical energy peak pricing demand
response program using robust optimization approach. Energy 2018;161:
999–1015. https://doi.org/10.1016/j.energy.2018.07.191.
[36] Amir V, Jadid S, Ehsan M. Optimal planning of a multi-carrier microgrid (MCMG)
considering demand-side management. Int J Renew Energy Res 2018;8(1):
238–49.
[37] Mohseni S, Brent A, Burmester D, Chatterjee A. Optimal sizing of an islanded
micro-grid using meta-heuristic optimization algorithms considering demand-
side management. In: Proceedings of the 2018 Australas. Univ. Power Eng. Conf.,
Auckland, New Zealand, 27–30 Nov. 2018, p. 1–6. https://doi.org/10.1109
/AUPEC.2018.8757882.
[38] Nazari A, Keypour R. A two-stage stochastic model for energy storage planning in
a microgrid incorporating bilateral contracts and demand response program.
J Energy Storage 2019;21:281–94. https://doi.org/10.1016/j.est.2018.12.002.
[39] Prathapaneni DR, Detroja KP. An integrated framework for optimal planning and
operation schedule of microgrid under uncertainty. Sustain Energy, Grids
Networks 2019;19:100232. https://doi.org/10.1016/j.segan.2019.100232.
[40] Bhamidi L, Sivasubramani S. Optimal planning and operational strategy of a
residential microgrid with demand side management. IEEE Syst J, Early Access
2019:1–9. https://doi.org/10.1109/JSYST.2019.2918410.
[41] Mohseni S, Brent AC, Burmester D. A demand response-centred approach to the
long-term equipment capacity planning of grid-independent micro-grids
optimized by the moth-flame optimization algorithm. Energy Convers Manag
2019;200:112105. https://doi.org/10.1016/j.enconman.2019.112105.
[42] Salyani P, Abapour M, Zare K, Babri T. Optimal stochastic planning of DERs in a
game theory framework considering demand response and pollution issues. In:
Demand Response Application in Smart Grids: Concepts and Planning Issues,
Volume 1. Cham: Springer International Publishing; 2020. p. 193–214. https://
doi.org/10.1007/978-3-030-31399-9_8.
[43] Burger S, Chaves-Avila JP, Batlle C, Perez-Arriaga IJ. The Value of Aggregators in
Electricity Systems. CEERP WP 2016-001. MIT Center for Energy and
Environmental Policy Research; 2016. Available: https://energy.mit.edu/
S. Mohseni et al.](https://image.slidesharecdn.com/mohseni2021-211027145406/85/Mohseni2021-28-320.jpg)
![Applied Energy 287 (2021) 116563
29
wp-content/uploads/2016/01/CEEPR_WP_2016-001.pdf [Accessed: 18-Oct.-
2020].
[44] Stevenson T, Batstone S, Reeve D, Poynton M, Comendant C. Transitioning to zero
net emissions by 2050: moving to a very low-emissions electricity system in New
Zealand. SAPERE Research Group – Prepared for the New Zealand Productivity
Commission; 2018. Available: https://www.productivity.govt.nz/assets/Docum
ents/3374eca8c4/Transitioning-to-zero-net-emissions-by-2050.pdf [Accessed:
18-Oct.2020].
[45] Okur Ö, Voulis N, Heijnen P, Lukszo Z. Aggregator-mediated demand response:
minimizing imbalances caused by uncertainty of solar generation. Appl Energy
2019;247:426–37. https://doi.org/10.1016/j.apenergy.2019.04.035.
[46] Nash J. Non-cooperative games. Ann Math 1951;54(2):286–95. https://doi.org/
10.2307/1969529.
[47] Basar T, Olsder GJ. Dynamic noncooperative game theory, vol. 23. Siam; 1999.
[48] Canadian Solar Inc. CS6K-270|275|280P; Dec. 2017, PV Module Product
Datasheet V5.552_EN. Available: https://www.collectiu-solar.cat/pdf/2-Panel-Ca
nadian_Solar-Datasheet-CS6K.pdf [Accessed: 4-Mar.-2020].
[49] Patel MR. Wind and solar power systems: design, analysis, and operation. CRC
Press; 2005.
[50] Murugaperumal K, Raj PADV. Feasibility design and techno-economic analysis of
hybrid renewable energy system for rural electrification. Sol Energy 2019;188:
1068–83. https://doi.org/10.1016/j.solener.2019.07.008.
[51] Meteonorm. Bern, Switzerland: Genossenschaft Meteotest; 2018. Version 7.3.0.
[52] Alstom. ecotècnia ECO 48/750; Aug. 2014, ecotècnia ECO 48/750 Datasheet.
[Online]. Available: https://en.wind-turbine-models.
com/turbines/791-ecot-cnia-eco-48-750#datasheet/ [Accessed: 4-Mar.-2020].
[53] Wagner R, Antoniou I, Pedersen SM, Courtney MS, Jørgensen HE. The influence
of the wind speed profile on wind turbine performance measurements. Wind
Energy Int J Prog Appl Wind Power Convers Technol 2009;12(4):348–62.
[54] Gipe P. Wind power. Wind Eng 2004;28(5):629–31. https://doi.org/10.1260/
0309524043028145.
[55] Sunceo Hydro. XJ50-100SCTF6-Z; 2015. [Online]. Available: https://www.micro
-hydro-power.com/100kw-hydro-turbine-generator/ [Accessed: 4-Mar.-2020].
[56] Khan MRB, Jidin R, Pasupuleti J. Multi-agent based distributed control
architecture for microgrid energy management and optimization. Energy Conserv
Manag 2016;112:288–307. https://doi.org/10.1016/j.enconman.2016.01.011.
[57] Bosona T, Gebresenbet G. Modeling hydropower plant system to improve its
reservoir operation. Int J Water Resour Environ Eng 2 (4):87–94. [Online].
Available: https://academicjournals.org/journal/IJWREE/article-full-text-pdf/34
24C7E1713 [Accessed: 19-Dec.-2020].
[58] All Power Labs. Old PP20 vs New PP30 Cogen-CS: Technical Specifications &
Comparisons; Jan. 2019, Rev 03. Available: http://www.allpowerlabs.com/wp-
content/uploads/2019/07/PP30-vs-PP20-Spec-Sheet-Public-2019-Rev-03-cu
rrent-July-2019.pdf [Accessed: 4-Mar.-2020].
[59] Singh S, Singh M, Kaushik SC. Feasibility study of an islanded microgrid in rural
area consisting of PV, wind, biomass and battery energy storage system. Energy
Convers Manag 2016;128:178–90. https://doi.org/10.1016/j.
enconman.2016.09.046.
[60] Ahmed OY, Ries MJ, Northrop WF. Emissions factors from distributed, small-scale
biomass gasification power generation: Comparison to open burning and large-
scale biomass power generation. Atmos Environ 2019;200:221–7. https://doi.
org/10.1016/j.atmosenv.2018.12.024.
[61] Moghaddas-Tafreshi SM, Mohseni S, Karami ME, Kelly S. Optimal energy
management of a grid-connected multiple energy carrier micro-grid. Appl Therm
Eng 2019;152:796–806. https://doi.org/10.1016/j.applthermaleng.2019.02.113.
[62] Qin H, Kimball JW. Solid-state transformer architecture using AC-AC dual-active-
bridge converter. IEEE Trans Ind Electron 2013;60(9):3720–30. https://doi.org/
10.1109/TIE.2012.2204710.
[63] Leonics Co. Apollo GTP-500; 2018, P.LEN.BRO.INV.156 Rev. 12.00. Available:
http://www.leonics.com/product/renewable/inverter/dl/GTP-500-156.pdf
[Accessed: 4-Mar.-2020].
[64] Akram U, Khalid M, Shafiq S. An innovative hybrid wind-solar and battery-
supercapacitor microgrid system—development and optimization. IEEE Access
2017;5:25897–912. https://doi.org/10.1109/ACCESS.2017.2767618.
[65] Eaton Corporation. XLR-48 Supercapacitor; Feb. 2019, Technical Data 10510.
Available: https://datasheet.octopart.com/XLR-48R6167-R-Eaton-datasheet-
130052459.pdf [Accessed: 4-Mar.-2020].
[66] CellCube Energy Storage Systems Inc. Use your own power grid: Intelligent
storage systems based on vanadium redox flow technology; 2018, D6144/
0714ND3. Available: https://static1.squarespace.com/static/5b1198ada2772c6
585959926/t/5b57363f88251b71261fc4a1/1532442177499/CellCube+-+Use+
Your+Own+Power+Grid.pdf [Accessed: 4-Mar.-2020].
[67] H-TEC Systems. The core of electrolysis: PEM electrolyser stacks – Designed for
ideal integration in systems; 2019, H-TEC Series-S: S 30/30. Available: https
://www.h-tec-systems.com/fileadmin/Content/PDFs/19022019/H-TEC_SYS
TEMS_Datenblatt_Stacks_SE30_30_EN.PDF [Accessed: 4-Mar.-2020].
[68] Ballard Power Systems Inc. FCgen–1020ACS; May 2015, SPC5101559-0H.
Available: https://www.ballard.com/docs/default-source/spec-sheets/fcgen-102
0-acs-v2.pdf?sfvrsn=c3ebc380_4 [Accessed: 19-Dec.-2020].
[69] Graham JDT, Mulvenna AJ, Mufford WE, Borck JG, Ko J, Harper MAM. Hydrogen
fueling station. U.S. Patent 6,810,925, issued 2 Nov. 2004.
[70] Pure Energy Centre Limited. Hydrogen fueling station; 2019. [Online]. Available:
https://pureenergycentre.com/hydrogen-fueling-station/ [Accessed: 4-Mar.-
2020].
[71] Takagi H. Queueing analysis: a foundation of performance evaluation.
Amsterdam: North-Holland; 1991.
[72] Robledo CB, Oldenbroek V, Abbruzzese F, van Wijk AJM. Integrating a hydrogen
fuel cell electric vehicle with vehicle-to-grid technology, photovoltaic power and
a residential building. Appl Energy 2018;215:615–29. https://doi.org/10.1016/j.
apenergy.2018.02.038.
[73] Ahmad MS, Mehmood MA, Luo H, Shen B, Latif M, Ghani WAWAK, et al.
Pyrolysis and thermogravimetric study to elucidate the bioenergy potential of
novel feedstock produced on poor soils while keeping the environmental
sustainability intact. Sustain 2019;11(13). https://doi.org/10.3390/su11133592.
[74] Winter C-J, Sizmann RL, Vant-Hull LL, editors. Solar power plants: fundamentals,
technology, systems, economics. Springer Science & Business Media; 2012.
[75] My Solar Quotes. Solar Power Buy-Back Rates; 2020. [Online]. Available: https
://www.mysolarquotes.co.nz/about-solar-power/residential/solar-power-buy-ba
ck-rates-nz/ [Accessed: 4-Mar.-2020].
[76] [dataset] CliFlo: New Zealand’s National Climate Database. [Online]. Available:
http://cliflo.niwa.co.nz/ [Retrieved: 10-Oct.-2019].
[77] Elert G. Energy density of hydrogen. Energy Density of Hydrogen− The Physics
Factbook 2017;22.
[78] XiaoW, Dunford WG, Capel A. A novel modeling method for photovoltaic cells. In:
Proceedings of the 2004 IEEE 35th Annual Power Electronics Specialists
Conference (IEEE Cat. No.04CH37551), Aachen, Germany, 20–25 Jun. 2004, pp.
1950–56. https://doi.org/10.1109/PESC.2004.1355416.
[79] Engineering ToolBox. Wind Shear; 2008. [Online]. Available: https://www.
engineeringtoolbox.com/wind-shear-d_1215.html/ [Accessed: 4-Mar.-2020].
[80] Interagency Working Group on Social Cost of Greenhouse Gases, United States
Government. Technical support document: technical update of the social cost of
carbon for regulatory impact analysis under executive order 12866. The August
2016 revision of the 2013 Technical Support Document on the Social Cost of
Carbon. Available: https://www.epa.gov/sites/production/files/2016-12/do
cuments/sc_co2_tsd_august_2016.pdf [Accessed: 4-Mar.-2020].
[81] Nerini FF, Broad O, Mentis D, Welsch M, Bazilian M, Howells M. A cost
comparison of technology approaches for improving access to electricity services.
Energy 2016;95:255–65. https://doi.org/10.1016/j.energy.2015.11.068.
[82] Thomas D, Deblecker O, Ioakimidis CS. Optimal design and techno-economic
analysis of an autonomous small isolated microgrid aiming at high RES
penetration. Energy 2016;116:364–79. https://doi.org/10.1016/j.
energy.2016.09.119.
[83] Duman AC, Güler Ö. Techno-economic analysis of off-grid PV/wind/fuel cell
hybrid system combinations with a comparison of regularly and seasonally
occupied households. Sustain Cities Soc 2018;42:107–26. https://doi.org/
10.1016/j.scs.2018.06.029.
[84] Mohseni S, Brent AC. Economic viability assessment of sustainable hydrogen
production, storage, and utilisation technologies integrated into on- and off-grid
micro-grids: a performance comparison of different meta-heuristics. Int J
Hydrogen Energy 2019. https://doi.org/10.1016/j.ijhydene.2019.11.079.
[85] Oldenbroek V, Smink G, Salet T, van Wijk AJM. Fuel cell electric vehicle as a
power plant: Techno-economic scenario analysis of a renewable integrated
transportation and energy system for smart cities in two climates. Appl Sci 2020;
10(1):143. https://doi.org/10.3390/app10010143.
[86] Oldenbroek V, Hamoen V, Alva S, Robledo CB, Verhoef LA, van Wijk AJM. Fuel
cell electric vehicle-to-grid: experimental feasibility and operational performance
as balancing power plant. Fuel Cells 2018;18(5):649–62. https://doi.org/
10.1002/fuce.201700192.
[87] Farahani SS, van der Veen R, Oldenbroek V, Alavi F, Lee EHP, van de Wouw N,
et al. A hydrogen-based integrated energy and transport system: the design and
analysis of the car as power plant concept. IEEE Trans Syst Man Cybern Mag
2019;5(1):37–50. https://doi.org/10.1109/MSMC.2018.2873408.
[88] Steward DM. Critical Elements of Vehicle-to-Grid (V2G) Economics. No. NREL/
TP-5400-69017, Golden, CO, USA: National Renewable Energy Lab (NREL);
2017.
[89] Soshinskaya M, Crijns-Graus WHJ, van der Meer J, Guerrero JM. Application of a
microgrid with renewables for a water treatment plant. Appl Energy 2014;134:
20–34. https://doi.org/10.1016/j.apenergy.2014.07.097.
[90] Khan MRB, Jidin R, Pasupuleti J, Shaaya SA. Optimal combinations of PV, wind,
micro-hydro and diesel systems for a seasonal load demand. In: Proceedings of
the 2014 IEEE International Conference on Power and Energy (PECon), Kuching,
Malaysia, 1–3 Dec. 2014, pp. 171–76. https://doi.org/10.1109/PECON.201
4.7062435.
[91] Chauhan A, Saini RP. Techno-economic optimization based approach for energy
management of a stand-alone integrated renewable energy system for remote
areas of India. Energy 2016;94:138–56. https://doi.org/10.1016/j.
energy.2015.10.136.
[92] Naderi M, Bahramara S, Khayat Y, Bevrani H. Optimal planning in a developing
industrial microgrid with sensitive loads. Energy Rep 2017;3:124–34. https://doi.
org/10.1016/j.egyr.2017.08.004.
[93] Mohseni S, Moghaddas-Tafreshi SM. Development of a multi-agent system for
optimal sizing of a commercial complex microgrid. arXiv preprint, arXiv:
1811.12553; 2018.
[94] Mohseni S, Moghaddas-Tafreshi SM. A multi-agent approach to optimal sizing of
a combined heating and power microgrid. arXiv preprint, arXiv:1812.11076;
2018.
[95] Mohseni S, Brent AC, Burmester D. Community resilience-oriented optimal micro-
grid capacity expansion planning: The case of Totarabank eco-village, New
Zealand. Energies 13 (15):3970. https://doi.org/10.3390/en13153970.
[96] Mohseni S, Brent AC, Kelly S. A hierarchical, market-based, non-cooperative
game-theoretic approach to projecting flexible demand-side resources: towards
more realistic demand response-integrated, long-term energy planning models.
S. Mohseni et al.](https://image.slidesharecdn.com/mohseni2021-211027145406/85/Mohseni2021-29-320.jpg)
![Applied Energy 287 (2021) 116563
30
In: Proceedings of the 2020 17th International Conference on the European
Energy Market (EEM), Stockholm, Sweden, 16–18 Sep. 2020, pp. 1–6. https://
doi.org/10.1109/EEM49802.2020.9221977.
[97] Yu M, Hong SH. Incentive-based demand response considering hierarchical
electricity market: a Stackelberg game approach. Appl Energy 2017;203:267–79.
https://doi.org/10.1016/j.apenergy.2017.06.010.
[98] Fahrioglu M, Alvarado L. Designing incentive compatible contracts for effective
demand management. IEEE Trans Power Syst 2000;15(4):1255–60. https://doi.
org/10.1109/59.898098.
[99] Fahrioglu M, Alvarado L. Using utility information to calibrate customer demand
management behavior models. IEEE Trans Power Syst 2001;16(2):317–22.
https://doi.org/10.1109/59.918305.
[100] Mirjalili S. Moth-flame optimization algorithm: a novel nature-inspired heuristic
paradigm. Knowledge-Based Syst 2015;89:228–49. https://doi.org/10.1016/j.
knosys.2015.07.006.
[101] Goldberg DE, Holland JH. Genetic algorithms and machine learning. Mach Learn
1988;3:95–9. https://doi.org/10.1023/A:1022602019183.
[102] Kennedy J, Eberhart R. Particle swarm optimization. In: Proceedings of the 1995
International Conference on Neural Networks, Perth, WA, Australia, 27 Nov.–1
Dec. 1995, p. 1942–48. https://doi.org/10.1109/ICNN.1995.488968.
[103] Mohseni S, Brent AC, Burmester D. A comparison of metaheuristics for the
optimal capacity planning of an isolated, battery-less, hydrogen-based micro-grid.
Appl Energy 2020;259:114224. https://doi.org/10.1016/j.
apenergy.2019.114224.
[104] Mohseni S, Brent AC, Burmester D. A sustainable energy investment planning
model based on the micro-grid concept using recent metaheuristic optimization
algorithms. In: Proceedings of the 2019 IEEE Congress on Evolutionary
Computation (CEC), Wellington, New Zealand, 10–13 Jun. 2019, pp. 219–26.
https://doi.org/10.1109/CEC.2019.8790007.
[105] Mohseni S, Brent AC, Burmester D, Chatterjee A. Stochastic optimal sizing of
micro-grids using the moth-flame optimization algorithm. In: Proceedings of the
2019 IEEE Power & Energy Society General Meeting (PESGM), Atlanta, GA, USA,
4–8 Aug. 2019, pp. 1–5. https://doi.org/10.1109/PESGM40551.2019.8973570.
[106] Trading Economics. New Zealand − Real Interest Rate. [Online]. Available: htt
ps://tradingeconomics.com/new-zealand/real-interest-rate-percent-wb-data.ht
ml/ [Retrieved: 5-Feb.-2020].
[107] Hübert T, Boon-Brett L, Buttner W. Sensors for safety and process control in
hydrogen technologies. CRC Press; 2016.
[108] Chen Q, Xia M, Zhou Y, Cai H, Wu J, Zhang H. Optimal planning for partially self-
sufficient microgrid with limited annual electricity exchange with distribution
grid. IEEE Access 2019;7:123505–20. https://doi.org/10.1109/
ACCESS.2019.2936762.
[109] Chowdhury S, Zhang J, Messac A. Avoiding premature convergence in a mixed-
discrete particle swarm optimization (MDPSO) algorithm. In: Proceedings of the
53rd AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and Materials
Conference, Honolulu, Hawaii, USA, 23–26 Apr. 2012, p. 1678. https://doi.org/
10.2514/6.2012-1678.
[110] Khan B, Singh P. Selecting a meta-heuristic technique for smart micro-grid
optimization problem: a comprehensive analysis. IEEE Access 2017;5:13951–77.
https://doi.org/10.1109/ACCESS.2017.2728683.
[111] MATLAB. Natick, MA: The MathWorks Inc; 2018. R2018b, Version 9.5.
[112] Anonymous. Ohakune Useful Information; 2011. [Online]. Available: http://
www.visitohakune.co.nz/page/ohakune-useful-information/12/ [Accessed: 4-
Mar.-2020].
[113] Environmental Protection Authority. Summary and Analysis – Carrots and
Parsnips. No. APP201045, 2012. Available: https://www.epa.govt.nz/assets/F
ileAPI/hsno-ar/APP201045/fb604d3064/APP201045-Summary-and-Ana
lysis-Carrot-and-Parsnip.pdf [Accessed: 4-Mar.-2020].
[114] Hall P, Jack M. Bioenergy options for New Zealand − Analysis of large-scale
bioenergy from forestry. No. 1124-2019-3124, Scion − Next Generation
Biomaterials, 2009. Available: https://niwa.co.nz/sites/niwa.co.nz/files/impor
ted/__data/assets/pdf_file/0007/95668/Large-scale-forestry-for-bioenergy.pdf
[Accessed: 4-Mar.-2020].
[115] [dataset] Anderson B, Eyers D, Ford R, Ocampo DG, Peniamina R, Stephenson J,
et al. New Zealand GREEN Grid Household Electricity Demand Study 2014–2018.
Colchester, Essex: UK Data Service. https://doi.org/10.5255/UKDA-SN-853334.
[116] Anderson J. Pulling the plug on network congestion. [B.A. dissertation]. Dunedin,
New Zealand: Department of Economics, University of Otago; 2009.
[117] Tayan O, Alginahi YM, Kabir MN, Al Binali AM. Analysis of a transportation
system with correlated network intersections: a case study for a central urban city
with high seasonal fluctuation trends. IEEE Access 2017;5:7619–35. https://doi.
org/10.1109/ACCESS.2017.2695159.
[118] The Electricity Market Information: The New Zealand Electricity Authority’s
wholesale database. [Online]. Available: https://www.emi.ea.govt.nz/Wholesale
/Reports/ [Retrieved: 12-Feb.-2020].
[119] Concept Economics Pty Ltd. Investigation of the value of unserved energy − Stage
1. No. ABN 73 129 990 530, Prepared for Electricity Commission; 2008.
Available: https://www.researchgate.net/profile/Deb_Chattopadhyay/publicati
on/322255908_Investigation_of_the_Value_of_Unserved_Energy/links/5a4e8
ff5458515e71b085a3f/Investigation-of-the-Value-of-Unserved-Energy.pdf
[Accessed: 19-Dec.-2020].
[120] Xi X, Sioshansi R. Using price-based signals to control plug-in electric vehicle fleet
charging. IEEE Trans Smart Grid 2014;5(3):1451–64. https://doi.org/10.1109/
TSG.2014.2301931.
[121] Nwulu NI, Xia X. Multi-objective dynamic economic emission dispatch of electric
power generation integrated with game theory based demand response programs.
Energy Convers Manag 2015;89:963–74. https://doi.org/10.1016/j.
enconman.2014.11.001.
[122] Nwulu NI, Xia X. Implementing a model predictive control strategy on the
dynamic economic emission dispatch problem with game theory based demand
response programs. Energy 2015;91:404–19. https://doi.org/10.1016/j.
energy.2015.08.042.
[123] Mani S, Sokhansanj S, Bi X, Turhollow A. Economics of producing fuel pellets
from biomass. Appl Eng Agric 2006;22(3):421–6. https://doi.org/10.13031/2
013.20447.
[124] Baghaee HR, Mirsalim M, Gharehpetian GB. Multi-objective optimal power
management and sizing of a reliable wind/PV microgrid with hydrogen energy
storage using MOPSO. J Intell Fuzzy Syst 2017;32(3):1753–73. https://doi.org/
10.3233/JIFS-152372.
[125] Li B, Roche R, Paire D, Miraoui A. Sizing of a stand-alone microgrid considering
electric power, cooling/heating, hydrogen loads and hydrogen storage
degradation. Appl Energy 2017;205:1244–59. https://doi.org/10.1016/j.
apenergy.2017.08.142.
[126] Mukherjee U, Maroufmashat A, Ranisau J, Barbouti M, Trainor A, Juthani N, et al.
Techno-economic, environmental, and safety assessment of hydrogen powered
community microgrids; case study in Canada. Int J Hydrogen Energy 2017;42
(20):14333–49. https://doi.org/10.1016/j.ijhydene.2017.03.083.
[127] Kampman B, Bergsma G, Schepers B, Croezen H, Fritsche UR, Henneberg K, et al.
BUBE: better use of biomass for energy - background report to the position paper
of IEA RETD and IEA bioenergy. IEA RETD and IEA Bioenergy. No. 10.3844.56;
2010. Available: https://www.ieabioenergy.com/wp-content/uploads/2013/1
0/Better-Use-of-Biomass-for-Energy-Background-Report.pdf [Accessed: 4-Mar.-
2020].
[128] Lai CS, Locatelli G, Pimm A, Wu X, Lai LL. A review on long-term electrical power
system modeling with energy storage. J Cleaner Prod 2021;280:124298. https://
doi.org/10.1016/j.jclepro.2020.124298.
[129] Hosseinalizadeh R, Shakouri GH, Amalnick MS, Taghipour P. Economic sizing of a
hybrid (PV–WT–FC) renewable energy system (HRES) for stand-alone usages by
an optimization-simulation model: Case study of Iran. Renew Sustain Energy Rev
2016;54:139–50. https://doi.org/10.1016/j.rser.2015.09.046.
[130] Shang C, Srinivasan D, Reindl T. An improved particle swarm optimisation
algorithm applied to battery sizing for stand-alone hybrid power systems. Int J
Electr Power Energy Syst 2016;74:104–17. https://doi.org/10.1016/j.
ijepes.2015.07.009.
[131] Fu R, Feldman DJ, Margolis RM. US solar photovoltaic system cost benchmark:
Q1 2018. No. NREL/TP-6A20-72399. Golden, CO, USA: National Renewable
Energy Lab (NREL); 2018.
[132] Li C. Techno-economic study of off-grid hybrid photovoltaic/battery and
photovoltaic/battery/fuel cell power systems in Kunming, China. Energy Sources
Part A Recover Util Environ Eff 2019;41(13):1588–604. https://doi.org/10.1080/
15567036.2018.1549134.
[133] Rezk H, Sayed ET, Al-Dhaifallah M, Obaid M, El-Sayed AHM, Abdelkareem MA,
et al. Fuel cell as an effective energy storage in reverse osmosis desalination plant
powered by photovoltaic system. Energy 2019;175:423–33. https://doi.org/
10.1016/j.energy.2019.02.167.
S. Mohseni et al.](https://image.slidesharecdn.com/mohseni2021-211027145406/85/Mohseni2021-30-320.jpg)
Mohseni2021
- 1. Applied Energy 287 (2021) 116563 Available online 9 February 2021 0306-2619/© 2021 Elsevier Ltd. All rights reserved. Strategic design optimisation of multi-energy-storage-technology micro-grids considering a two-stage game-theoretic market for demand response aggregation Soheil Mohseni a,* , Alan C. Brent a,b , Scott Kelly c , Will N. Browne a , Daniel Burmester a a Sustainable Energy Systems, School of Engineering and Computer Science, Faculty of Engineering, Victoria University of Wellington, PO Box 600, Wellington 6140, New Zealand b Department of Industrial Engineering and the Centre for Renewable and Sustainable Energy Studies, Stellenbosch University, Stellenbosch 7600, South Africa c Institute for Sustainable Futures, University of Technology Sydney, Sydney, NSW 2007, Australia H I G H L I G H T S G R A P H I C A L A B S T R A C T • A market-driven model is devised for long-term projections of incentive- aware loads. • Responsive loads are integrated through dedicated aggregators for improved accuracy. • A level playing field is provided for fuel cell electric vehicle-to-grid technology. • An energy filter-based approach is employed to allocate various storage technologies. • The model’s potential in cutting a test micro-grid’s lifetime costs by 21% is shown. A R T I C L E I N F O Keywords: Sustainable energy systems Demand-side management Strategic energy planning Optimal investment planning Demand response aggregator Game theory A B S T R A C T While industrial demand response programmes have long been valued to support the power grid, recent advances in information and communications technology have enabled new opportunities to leverage the potential of responsive loads in less energy-dense end-use sectors. This brings to light the importance of accurately projecting flexible demand-side resources in the long-term investment planning process of micro-grids. This paper in troduces a customer comfort-aware, demand response-integrated long-term micro-grid planning optimisation model. The model (1) draws on non-cooperative game theory and the Stackelberg leadership principles to un derstand and reflect the strategic behaviour of energy utilities, demand response aggregators, and end- consumers, (2) produces optimal trade-offs between power imported from the main grid and available de mand response resources, (3) determines the cost-optimal resource allocation for energy infrastructure, including multiple energy storage systems, and (4) provides a level playing field for emerging technologies, such as power- to-gas and vehicle-to-grid interventions. The multi-energy-storage-technology test-case was effectively applied to achieve 100%-renewable energy generation for the town of Ohakune, New Zealand. Numerical simulation results suggest that the proposed incentive-compatible demand-side management market-clearing mechanism is able to estimate the cost-optimal solution for the provision of renewable energy during the planning phase. The cost- optimal system saves ~21% (equating to around US$5.5 m) compared to a business-as-usual approach, where * Corresponding author. E-mail address: soheil.mohseni@ecs.vuw.ac.nz (S. Mohseni). Contents lists available at ScienceDirect Applied Energy journal homepage: www.elsevier.com/locate/apenergy https://doi.org/10.1016/j.apenergy.2021.116563 Received 13 August 2020; Received in revised form 20 January 2021; Accepted 23 January 2021
- 2. Applied Energy 287 (2021) 116563 2 the participation of end-users in demand response programmes is projected by running uniform price demand response auctions. The most salient distinction of the proposed two-stage (wholesale and retail) demand-side management market model is the continual process of trading, with incentive prices unique to each transaction. Nomenclature Indices and sets c ∈ C = {PV,WT,MH,T,E,FC,HT,BP,B,SC,S,FCEV2G} micro-grid components, the optimal size of which is under investigation: photovoltaic panels (PV), wind turbines (WT), micro-hydro turbines (MH), transformer (T), electrolyser (E), fuel cell (FC), hydrogen tank (HT), biopower plant (BP), battery bank (B), super-capacitor bank (SC), hydrogen station (S), and fuel cell electric vehicle-to-grid unit (FCEV2G) d ∈ D = {1,2,⋯,365} day of the year-round micro-grid operation D* = {d* 1,d* 2,⋯,d* K} set of best-response load reductions contributed by all the customers D j,* LA = {dk,j,* ,d− k,j,* } set of best-response strategies of all the customers signed up with the j-th aggregator es ∈ ES = {B,SC,HT,FCEV} energy storage media: battery bank (B), super-capacitor bank (SC), hydrogen tank (HT), and aggregated fuel cell electric vehicles’ tanks (FCEV) I* LA = {I1,* LA ,I2,* LA ,I3,* LA ,I4,* LA ,I5,* LA } set of best-response incentive payments of aggregators j ∈ J responsive load aggregators k ∈ NJ customers enrolled with aggregator j K set of all the micro-grid’s customers p ∈ Pd⊂T day-specific peak consumption hour t ∈ T = {1,2,⋯,8760} time-step of the year-round micro-grid operation Parameters c k,j 1 discomfort tolerance coefficient of customer k of aggregator j [$/kWh2 ] c k,j 2 discomfort tolerance coefficient of customer k of aggregator j [$/kWh] C900, C115, C33 nameplate capacities of inverters [kW] CV gross calorific value of biomass feedstock [kWh/kg] δj load type-dependent demand response procurement factor; sectoral elasticity of customer-supplied demand response capacity Δt time-step length [h] d k,j cr (t), d k,j ncr(t) critical/non-critical portion of the load power demanded by the k-th customer subscribed to aggregator j at time-step t [kWh] d k,j full(t) full load power demanded by the k-th customer subscribed to aggregator j at time-step t [kWh] disk,j,min , disk,j,max lower/upper limit of the discomfort cost imposed on the k-th customer of aggregator j [$] DF derating factor of the PV module [%] ξCO2 social cost of CO2 emissions [$/tCO2] ηes round-trip efficiency of storage medium es [%] ηFCEV2G efficiency of the operation of the fuel cell electric vehicles in the vehicle-to-grid mode [%] ηPV, ηMH, ηBP, ηT, ηI, ηSC, ηB, ηE, ηHT, ηFC, ηS efficiency of the PV plant/ micro-hydro plant/biopower plant/transformer/inverter/ super-capacitor/battery/electrolyser/hydrogen tank/fuel cell/hydrogen station [%] ηPV,DC/DC, ηMH,AC/DC, ηBP,AC/DC PV plant’s DC/DC converter efficiency, micro-hydro power plant’s AC/DC converter efficiency, biopower plant’s AC/DC converter efficiency [%] ECO2 CO2 emission factor of the biopower plant [kg-CO2/kg- feedstock] Ees,min, Ees,max lower/upper capacity limit of storage medium es [kWh] F(t) river streamflow rate at time-step t [m3 /s] g acceleration of gravity [m/s2 ] h wind turbine hub height [m] hg micro-hydro turbine gross head [m] href reference height of wind speed records [m/s] HHVH2 higher heating value of hydrogen [kWh/kg] iMGO step size for the micro-grid operator-determined incentive [$/kWh] IG(t) global solar irradiance on the horizontal surface at time- step t [kW/m2 ] I j,min LA , I j,max LA lower/upper limit of the incentives determined by aggregator j [$/kWh] Imin MGO, Imax MGO lower/upper limit of the micro-grid operator-offered incentives [$/kWh] Iref reference solar irradiance [kW/m2 ] Itermax maximum number of iterations K DC gain of the transfer function Kp PV module’s temperature coefficient [%/◦ C] LPSPmax e , LPSPmax H2 maximum allowable loss of power supply probability in supplying electricity/hydrogen [%] MBD(t) biomass feedstock mass consumption rate at time-step t [kg/h] N j cust number of customers enrolled with aggregator j Nmax c upper limit of the size (capacity/quantity) of component c NSA number of search agents of the optimisation algorithm NMOT nominal PV module operating temperature [◦ C] ρ water density [kg/m3 ] Pch,max es , Pdch,max es upper limit of the charging/discharging rate of storage medium es [kW] Pch,min es , Pdch,min es lower limit of the charging/discharging rate of storage medium es [kW] Pmax FCEV2G(t) maximum V2G power at time-step t [kW] PFC,r,PE,r rated capacity of each fuel cell/electrolyser stack [kW] PL(t) load power demand at time-step t [kW] PL,max maximum electrical load on the micro-grid [kW] PMH,r, PBP,r rated capacity of each micro-hydro turbine/biopower plant [kW] PPV(t), PWT(t), PMH(t), PBP(t) power output from the PV/wind turbine/micro-hydro/biopower plant at time-step t [kW] PPV,r rated capacity of the PV module under standard test conditions [kW] PS(t) hydrogen power demand of the station at time-step t [kW] penconst penalty term added to the life-cycle cost function where constraints are not met [$] πex, πim(t) per-unit income from electrical energy exports [$/kWh], per-unit cost of electrical energy imports at time-step t [$/kWh] πFCEV2G per-unit premium tariff rate for V2G power [$/kWh] Q quality factor of the low-pass energy filter tup minimum up-time of the electrolyser, fuel cell, and biopower plant [h] Ta(t) ambient temperature at time-step t [◦ C] Tm(t) PV module temperature at time-step t [◦ C] S. Mohseni et al.
- 3. Applied Energy 287 (2021) 116563 3 1. Introduction One of the principal advantages of making the electricity grid “smart” is that it enables consumers to proactively engage in electricity markets and benefit from demand-side management (DSM) schemes designed and incentivised by utilities to curtail/interrupt or shift a proportion of electricity demand, and thereby flatten the load power profile–and improve the load factor. While demand response (DR) pro grammes have been in use to improve the energy efficiency of industrial consumers for years, the expansion of the concept to include less energy- TSTC PV module temperature under standard test conditions [◦ C] Vh normalised wind speed profile to the wind turbine hub height [m/s] Vref reference wind speed profile [m/s] ω0 cut-off frequency [dB] γ wind shear exponent Variables costem(t) total penalties imposed for emissions at time-step t [$] costFCEV2G(t) cost associated with the FCEV2G operations at time- step t [$] costim(t) cost of electricity import at time-step t [$] dk,j (t) load reduction contributed by the k-th customer of aggregator j at time-step t [kWh] dk,j,* (t) best-response strategy taken by the k-th customer subscribed to the j-th aggregator for load reduction at time- step t [kWh] Ddef (t) capacity deficit to meet the loads at time-step t [kWh] D j LA(t) load reduction contributed by aggregator j at time-step t [kWh] disk,j (t) discomfort cost imposed on the k-th customer subscribed to aggregator j at time-step t [$] Ees(t) energy content of storage medium es at time-step t [kWh] ESC(t), EB(t), EHT(t) energy content of the super-capacitor/battery bank/hydrogen tank at time-step t [kWh] I j LA(t) incentive payment offered by aggregator j for load reduction at time-step t [$/kWh] I j,* LA(t) best-response incentive payment for load reduction offered by aggregator j at time-step t [$/kWh] IMGO(t) rate of micro-grid operator-posted incentive payments for load reduction at time-step t [$/kWh] I* MGO(t) globally-optimum incentive payment for load reduction offered by the MG operator at time-step t [$/kWh] incomeex(t) income from electricity export at time-step t [$] LPSPe, LPSPH2 loss of power supply probability in supplying electricity/hydrogen [%] mHT(t) mass of hydrogen stored in the tank at time-step t [kg] NB optimal capacity of the overall battery bank [kWh] NFCEV2G optimal capacity of the fuel cell electric vehicle-to-grid system [kW] NHT optimal capacity of the hydrogen tank [kg] NI optimal capacity of the electrical loads’ overall power inversion system [kW] NPV, NWT, NMH, NBP, NE, NFC, NSC optimal quantity of PV modules/ wind turbines/micro-hydro turbines/biopower units/ electrolyser stacks/fuel cell stacks/super-capacitor modules NS optimal capacity of the hydrogen refuelling station [kg-H2/ h] NT optimal capacity of the transformer [kVA] N900, N115, N33 optimal quantity of 900-kW/115-kW/33-kW inverters N1600, N400, N100 optimal quantity of 1600-kWh/400-kWh/100- kWh battery packs NPCc net present cost of component c [$] NPCI net present cost of the inverter [$] OCMG(t) operational cost of offsetting power deficit at time-step t [$] OC* MG(t) globally-optimum operational cost of the micro-grid to address the shortage of power generation capacity at time- step t [$] Pch(t), Pdch(t) total charging/discharging power of the hybrid battery/super-capacitor storage system at time-step t [kW] Pch,HF2, Pdch,HF2 charging/discharging power of the super-capacitor bank [kW] Pch,LF2, Pdch,LF2 charging/discharging power of the battery bank [kW] PE(t) power consumed by the electrolyser at time-step t [kW] PE− HT(t) hydrogen power directed from the electrolyser to the hydrogen tank at time-step t [kW] Pch es (t), Pdch es (t) charging/discharging rate of energy storage medium es at time-step t [kW] PFC(t) power generated by the fuel cell at time-step t [kW] PFCEV2G(t) aggregated vehicle-to-grid power provided by fuel cell electric vehicles at time-step t [kW] PHT− FC(t) hydrogen power directed from the hydrogen tank to the fuel cell at time-step t [kW] PHT− S(t) hydrogen power directed from the hydrogen tank to the station at time-step t [kW] Pim(t), Pex(t) imported/exported electricity at time-step t [kW] PSH(t), PEX(t) shortage/excess of renewable power generation at time-step t [kW] PSH− LF1(t), PSH− HF1(t) low-/high-frequency component of the renewable power shortage signal at the first low-pass filter output at time-step t [kW] PSH− LF2(t), PSH− HF2(t) low-/high-frequency component of the renewable power shortage signal at the second low-pass filter output at time-step t [kW] PEX− LF1(t), PEX− HF1(t) low-/high-frequency component of the renewable power excess signal at the first low-pass filter output at time-step t [kW] PEX− LF2(t), PEX− HF2(t) low-/high-frequency component of the renewable power excess signal at the second low-pass filter output at time-step t [kW] Pr j LA(t) profit gained by aggregator j at time-step t [$] QL(t), QH2 (t) unmet electrical/hydrogen load demand at time-step t [kW] Uk,j (t) utility of the customer k serviced by aggregator j at time- step t [$] Functions H(s) low-pass energy filter transfer function NPV 20− yr (z) net present value of cost component z over the 20-year life of the project [$] NPCc 20− yr net present cost of micro-grid component c over the 20- year life of the project [$] NPCI 20− yr net present cost of the overall power inverter over the 20- year life of the project [$] OCMG hourly operational cost function of the micro-grid [$] ⌊⋅⌋ floor function ⌈⋅⌉ ceiling function S. Mohseni et al.
- 4. Applied Energy 287 (2021) 116563 4 dense demand sectors, namely the residential, agricultural, and com mercial sectors, as well as electrified transport, is enabled by recent advancements in information and communications technology (ICT), which have substantially contributed to the development of advanced metering infrastructure [1,2]. Recent studies have revealed that the consideration of DSM strategies in the optimum investment planning phase of renewable and sustainable energy systems (RSESs) for domestic applications can offer cost savings of about 15% to nearly 35% (depending on the participation rate of end-users in the DR pro grammes), whilst preserving consumer comfort standards [3,4,5]. That is, the proper integration of DR programmes into RSESs would result in a win–win–win situation–the third winner being the environment, as they will accelerate the transition to a low-carbon energy economy and a world run on green energy. 1.1. Long-term, demand response-integrated micro-grid infrastructure planning background A reformed formulation of the micro-grid (MG) equipment capacity- planning problem is required to make effective use of the economic opportunities offered by DSM processes to support decision-making in developing cost-effective RSESs [6]. A solution to the optimal DR- integrated MG design problem identifies the least-cost combination of the size of the components of the system over a decades-long–often spanning 20–30 years–investment planning horizon to meet the pro jected demand for energy, while leveraging the potential of responsive loads [7,8]. Recent review studies have focused on discussing methods and trends for harvesting the potential of the demand-side flexibility to contribute significantly to energy affordability in energy networks with a high penetration of distributed renewables. Gelazanskas and Gamage [9], Haider et al. [10], Esther and Kumar [11], Wang et al. [12], Robert et al. [13], as well as, more recently, Jordehi [14] have scrutinised various approaches to implementing DR arrangements, while optimally designing RSESs, with a particular focus on residential DR resources. Moreover, various types of DSM strategies have been incorporated in the formulation of the MG capacity-optimisation models. This implies that DR programmes are well-analysed for the planning of RSESs, a state ment that has likewise been made in the context of different DSM business models in electricity markets [15,16], as well as for the optimal operational scheduling (energy management) of RSESs [12]. There have also been attempts to exploit other types of DR structures for the optimal capacity planning of RSESs. For instance, Kahrobaee et al. [17] devised a particle swarm optimisation (PSO)-based planning model for a smart home nano-grid that utilises the real-time pricing (RTP) scheme, which allows for leveraging the historical records of the price elasticity of demand for personalised dynamic pricing. In another instance, Yu et al. [18] proposed a robust flexible-programming approach for the integration of renewables into a municipal energy system, which runs a critical peak pricing (CPP) rate structure. More over, Varasteh et al. [19] employ a hybrid direct load control-time-of- use (DLC-ToU) DR framework to drive down the whole-life cost of a grid-tied combined heat and power (CHP) MG. In addition, some studies have explored the potential of vehicle-to- grid (V2G) technologies and electric vehicle (EV) charging/discharg ing coordination through DSM mechanisms in driving economic sus tainability improvement for renewable energy development projects. For instance, Cardoso et al. [20] have proposed a DLC decision model for the aggregated energy scheduling of EVs and demonstrated its distinc tive contribution to reducing the lifetime cost of a multiple energy carrier MG, while considering the uncertainty associated with the EV driving schedules. In another instance, Hosseinnia et al. [21] have provided further evidence of the utility and economic benefits of EV fleet trip level energy management and V2G connectivity in the context of sustainable energy system design and planning. Moghaddas-Tafreshi et al. [22] have also underlined the potential of optimal charging/ discharging scheduling of plug-in hybrid EVs in improving the profit ability of an energy hub and reaping cost-savings for vehicle owners, while addressing the uncertainty associated with the power consump tion of vehicles during trips. Table 1 summarises the most rigorous studies carried out to date on the integration of demand-side resources (for the strategic planning of energy demand) in the long-term capacity optimisation models of RSESs (listed in ascending order of publication date), whilst additionally situating this study in the context of the existing literature. 1.2. Demand response-integrated life-cycle planning of micro-grids: knowledge gaps and proposition As Table 1 indicates, there is a growing body of literature lending support to the integration of DSM frameworks into the design phase of RSESs. However, as far as can be ascertained, no single study has eval uated the attitude of neither end-users nor electricity providers in relation to adopting these practices during the optimal design and planning process of RSESs. Accordingly, oversimplified assumptions have commonly been made in the literature regarding the available capacity of responsive loads, which have substantially reduced the ac curacy of projections. That is, many hypotheses regarding the degree of end-users’ participation in the DR schemes are not well-grounded. To aid the associated asset-allocation decision-making procedure, a long- term, DR-integrated MG investment planning approach needs to model the involvement of aggregator-mediated customers in the DR programmes in a systematic, market-driven approach. The market- driven approach needs to capture the dynamic nature of strategic in teractions between rational, utility-maximising active economic agents in an aggregator-mediated DSM market. More specifically, the approach needs to identify the reaction and commitment of different classes of customers mediated by third-party demand response aggregators (DRAs), when exposed to variations in the economic incentives for load curtailment/shifting. In this context, the DRAs round up parcels of interruptible loads to enable them to reach the sufficient scale required for selling services to the system operator(s) [43,44,45]. In addition, more work is needed to evaluate the effect of different levels of discomfort experienced by different customer classes on the economic feasibility of renewable energy projects as the characterisation of aggregator-mediated customer comfort constraints during the planning phases of RSESs is less well explored. To assist decision-makers in designing cost-optimal sustainable energy systems consistent with the expectations of their customers, it is critically important to devise ac curate models aimed at reflecting user values and preferences (which furnish the basis for service flexibility) in the design of MG projects. This brings to light the need for an investment decision-making framework that accommodates end-users’ preferences (which could be derived from their energy service needs and the relative values they place on them) within the long-term MG capital-investment plans. 1.3. Objective The main objective of the paper is to demonstrate the potential of aggregator-mediated, incentive-based, market-driven DSM programmes tailored to small- to medium-scale end-consumers in improving the economic viability of community-scale MG systems. Accordingly, the paper expands the boundaries of knowledge and understanding of the positive impacts of altering energy consumption behaviour of different types of electrical loads–through effective incentive-based DR pro grammes–on the cost-optimal design of MGs. Also, a secondary objective of the paper is to ascertain the technological competence and cost- competitiveness of utilising hydrogen as an energy vector in community-scale MGs for niche applications–inter-seasonal energy storage to meet seasonal demand, and hydrogen mobility to decarbonise the transport sector. More specifically, the paper contributes to the trend of the S. Mohseni et al.
- 5. Applied Energy 287 (2021) 116563 5 Table 1 Summary of the studies on the provision of the DSM procurements in the long-term investment planning of the RSESs. Reference Test-case system configuration DSM scheme Flexible loads V2G capabilities DR-inherent uncertainties Multi-temporal reserve procurement Aggregator-mediated customer comfort characterisation Modelling approach Objective(s) Solution algorithm Case study area Martins and Borges, 2011 [23] A typical active distribution grid with a high share of renewables ICSs Unspecified × × × × Stochastic LCCM GA A typical distribution network in Brazil Kahrobaee et al., 2013 [17] A grid-connected WT/ BESS nano-grid RTP SRAs × × × × Stochastic LCCM PSO A typical house in the U.S. Cardoso et al., 2014 [20] A grid-connected PV/ ST/ICE/MT/GT/FC/ BESS/AC MG DLC EV-charging √ × × × Stochastic LCCM DER-CAM tool San Francisco, CA, U.S. Zhu et al., 2015 [24] An off-grid PV/WT/ BESS/DG MG DLC HVAC × × × × Deterministic LCCM NP Shanghai, China Atia and Yamada, 2016 [25] A grid-tied PV/WT/ BESS MG DLC SRAs and EV-charging × × × × Stochastic LCCM MILP Okinawa, Japan Pazouki and Haghifam, 2016 [26] A WT/MCHP/boiler/ BESS/TESS energy hub DLC Unspecified × × × × Stochastic LCCM MILP Unspecified Schachter et al., 2016 [27] A typical smart distribution grid with deep renewable penetration DLC Unspecified × √ × × Stochastic LCCM SDM Unspecified Yu et al., 2017 [18] A WT/PV/BP/coal/ gas municipal energy system CPP Unspecified × × × × Stochastic LCCM RFP Qingdao, China Chauhan and Saini, 2017 [28] A stand-alone PV/ WT/BESS/DG/BP/ MHPP MG DLC Smart appliances of the residential, commercial, agricultural, and community sectors × × × × Deterministic LCCM DHS Chamoli, India Nojavan et al., 2017 [29] The standard IEEE 33- bus distribution network ToU Unspecified × × × × Stochastic LCCM and reliability maximisation MINLP Unspecified Amrollahi and Bathaee, 2017 [30] A grid-connected PV/ WT/BESS DLC Unspecified × × × × Deterministic LCCM MILP An unnamed forestry camp, northwest of Iran Chen et al., 2018 [31] A grid-tied PV/WT/ BESS MG DLC SRAs and HVAC × × × × Stochastic LCCM and reliability maximisation MILP Unspecified Zheng et al., 2018 [32] A grid-tied PV/BP/ boiler MG ToU Unspecified × × × × Stochastic LCCM LP Davis, CA, U.S. Xiao et al., 2018 [33] A modified IEEE 33- bus distribution network with deep penetration of renewables Hybrid DLC- ICSs Unspecified thermal and electrical loads × × × × Stochastic LCCM MBGO Unspecified Husein and Chung, 2018 [34] An on-grid PV/WT/ MHPP/GPP/BP MG ToU Unspecified × √ × × Stochastic LCCM ESM Seoul, South Korea Gazijahani and Salehi, 2018 [35] A modified IEEE 33- bus distribution network with high penetration of renewables CPP Unspecified √ × × × Deterministic LCCM RMILP Unspecified (continued on next page) S. Mohseni et al.
- 6. Applied Energy 287 (2021) 116563 6 Table 1 (continued) Reference Test-case system configuration DSM scheme Flexible loads V2G capabilities DR-inherent uncertainties Multi-temporal reserve procurement Aggregator-mediated customer comfort characterisation Modelling approach Objective(s) Solution algorithm Case study area Amir et al., 2018 [36] A grid-connected PV/ BESS/boiler/TESS/ MCHP MG ToU Unspecified × × × × Deterministic LCCM GA Unspecified Mohseni et al., 2018 [37] An off-grid PV/WT/ battery MG DLC EV-charging × × × × Deterministic LCCM GA Kish Island, Iran Nazari and Keypour, 2019 [38] An on-grid PV/WT/ BESS/MT MG DLC SRAs, PWS, and HVAC × × × × Stochastic LCCM MILP Unspecified Prathapaneni and Detroja, 2019 [39] A stand-alone PV/ BESS/DG MG DLC EV-charging and PWS × × × × Stochastic LCCM MINLP Hyderabad, India Hosseinnia et al., 2019 [21] An on-grid PV/WT/ BESS/boiler/MCHP/ TESS MG ToU SRAs √ × × × Stochastic LCCM and GHGEM TSA Unspecified Bhamidi and Sivasubramani, 2019 [40] A grid-connected PV/ WT/BESS/MT/DG MG ToU SRAs and EV-charging × × × × Deterministic LCCM and GHGEM NSGA-II San Angelo, TX, U.S. Varasteh et al., 2019 [19] A multi-carrier PV/ WT/CCHP/boiler/ BESS MG Hybrid DLC- ToU SRAs × × × × Deterministic LCCM MINLP Unspecified Mohseni et al., 2019 [41] A grid-independent PV/WT/BESS MG DLC SRAs and EV-charging × × × × Deterministic LCCM MFOA Hengam Island, Iran Amir and Azimian, 2020 [8] A grid-connected PV/ MCHP/BESS/TESS multiple energy carrier MG Hybrid DLC- ToU Unspecified × × × × Stochastic LCCM GA-MINLP Unspecified Salyani et al., 2020 [42] The standard IEEE 33- bus distribution network RTP EV-charging √ √ × × Stochastic LCCM and GHGEM MINLP Unspecified This study A grid-tied PV/WT/ MHPP/BP/FC/BESS/ SC MG ICSs SRAs and FCEVs √ √ √ √ Deterministic LCCM MFOA Ohakune, New Zealand Key: AC = Absorption Chiller, BESS = Battery Energy Storage System, BP = Biopower Plant, CA = California state, CPP = Critical Peak Pricing, CCHP = Combined Cooling, Heating, and Power, DER-CAM = Distributed Energy Resources-Customer Adoption Model, DG = Diesel Generator, DHS = Discrete Harmony Search, DLC = Direct Load Control, ESM = Enumeration Search Method, EV = Electric Vehicle, FC = Fuel Cell, FCEV = Fuel Cell Electric Vehicle, GA = Genetic Algorithm, GHGEM = Greenhouse Gas Emissions Minimisation, GPP = Geothermal Power Plant, GT = Gas Turbine, HVAC = Heating, Ventilation, and Air Conditioning, ICE = Internal Combustion Engine, ICSs = Interruptible/Curtailable Services, LCCM = Life-Cycle Cost Minimisation, LP = Linear Programming, MBGO = Metamodel-Based Global Optimisation, MCHP = Micro-Combined Heat and Power, MG = Micro-Grid, MHPP = Micro-Hydro Power Plant, MILP = Mixed-Integer Linear Programming, MINLP = Mixed-Integer Nonlinear Programming, MFOA = Moth-Flame Optimisation Algorithm, MT = Micro- Turbine, NP = Nonlinear Programming, NSGA-II = Non-dominated Sorting Genetic Algorithm II, PSO = Particle Swarm Optimisation, PV = Photovoltaic, PWS = Pumped Water Storage, RFP = Robust Flexible Pro gramming, RMILP = Robust MILP, RTP = Real-Time Pricing, SC = Super-Capacitor, SDM = Supply-Demand Matching, SRAs = Smart Residential Appliances, ST = Solar Thermal, TESS = Thermal Energy Storage System, ToU = Time-of-Use, TSA = Tabu Search Algorithm, TX = Texas state, WT = Wind Turbine. S. Mohseni et al.
- 7. Applied Energy 287 (2021) 116563 7 conservation of energy through procuring DSM provisions for the stra tegic decision-making related to the optimal mix of distributed energy resources (DERs) to be integrated into RSESs–which is discussed in the literature review in Section 1.1. Accordingly, this study puts forward a novel long-term, comfort-preserving MG equipment capacity-planning decision-making framework that offers a new solution to fill the litera ture voids identified in Section 1.2. Notably, this paper makes the following key contributions: • The strategic interactions between the MG operator (utility), mo nopoly DRAs, and end-consumers are characterised using an equi table market model for DR aggregation in community-scale renewable energy projects using tools borrowed from non- cooperative game theory [46] and the endogenous Stackelberg leader–follower relationships1 [47]. The proposed DSM market model is designed on the basis of interruptible DR programmes and accounts for the elasticity of customer-supplied DR capacity (load type-dependent DR procurement factor). • The proposed DSM market design is integrated into a standard model of long-term, meta-heuristic-based capacity planning of grid- connected MGs to elucidate the contributions of more accurate DR resource projections in improving the economic viability of MG development projects. • A novel hydrogen-based MG system is conceptualised, which is the first to capture the potential of the fuel cell electric vehicles in vehicle-to-grid operation (FCEV2G) technology in improving the dispatchability of 100%-renewable MG systems and, in turn, ensuring the economic sustainability of strategic MG investment planning decisions. • The application of the energy filter-based approach to scheduling energy storage infrastructure is expanded to multiple energy storage technologies, namely: hydrogen storage, vanadium redox flow bat teries, and super-capacitors (SCs). This provides a platform to more efficiently address the intermittency of renewables by economically dispatching different backup systems running at various temporal resolutions, namely: seasonal, inter- and intra-day, and transient. 1.4. Structure of paper The rest of this paper is organised as follows. Section 2 mathemati cally defines the conceptualised stand-alone, multi-energy-storage- technology MG architecture employed as a test-case to evaluate the utility and effectiveness of the proposed two-stage market-driven DSM business model. The proposed interruptible DR scheduling framework is presented in Section 3. Section 4 integrates the proposed DSM frame work into a standard meta-heuristic-based MG capacity planning model. A case study analysis is carried out in Section 5. Finally, conclusions are made in Section 6. A schematic outline of the paper, which illustrates the steps followed in this study and their interconnectedness, is set out in Fig. 1. 2. Test-case micro-grid system The conceptualised grid-connected, DC-coupled, multiple energy carrier MG test-case system (see Fig. 2) is envisioned to supply green power and transportation fuel to communities residing in the vicinity of, or within relatively short distances from, the main power grids. Also, it serves five different categories of energy demand: (1) residential, (2) agricultural, (3) commercial, and (4) industrial load power demands, as well as (5) the demand for hydrogen (through dedicated hydrogen refuelling infrastructure) from fuel cell electric vehicles (FCEVs). The test-case is used to verify the effectiveness of the proposed DR-integrated energy planning optimisation model. 2.1. Micro-grid equipment For the purposes of this study, the leading brands of equipment in New Zealand’s renewable energy asset market were chosen based on the first author’s judgement of prevalence. The following sub-sections mathematically model the system equipment. 2.1.1. Photovoltaic plant Canadian Solar’s CS6K-280P poly-crystalline photovoltaic (PV) modules [48], which have a nominal power of 280 W are employed in this study for PV power generation. The power output from the PV plant at each time-step, PPV(t) [kW], can be estimated as follows [25,49,50]: Tm(t) = Ta(t) + IG(t) × NMOT − 20 0.8 , (1) PPV (t) = NPV × PPV,r × ηPV,DC/DC × DF × IG(t) ISTC × ( 1 − Kp 100 × (Tm(t) − TSTC ) ) , (2) where NPV is the optimum quantity of the modules; PPV,r is the rated capacity of the module under the standard test conditions (STC); ηPV,DC/DC is the PV plant’s DC/DC converter efficiency; Kp is the tem perature coefficient of the module; Tm, Ta, and TSTC respectively repre sent the PV module temperature, ambient temperature, and the module temperature at the STC; IG and ISTC respectively denote the global solar irradiance on the horizontal surface and the solar irradiance at the STC; and NMOT and DF respectively stand for the nominal module operating temperature and derating factor. The tilt angle is assumed as 30◦ and the Meteonorm software [51] is used to normalise the values of IG to this tilt angle. Also, the numeric values 20 and 0.8 respectively represent the ambient temperature [◦ C] and solar irradiance [kW/m2 ] at which the NMOT is defined. 2.1.2. Wind plant The wind turbine (WT) ECO 48/750, which has a rated power of 750 kW is considered for wind power generation [52]. The turbine’s manufacturer-provided characteristic power-wind speed curve is shown in Fig. 3. The wind plant’s output power at each time-step, PWT(t) [kW], can be obtained by multiplying the optimal quantity of the WTs, NWT, by each turbine’s output power estimated from the power curve presented in Fig. 3. Also, since the power curve of the WT is characterised for its hub height wind speed, Eq. (3) can be used to normalise the wind speed data measured at other heights to the turbine’s hub height [53]. Vh = Vref ×( h href )γ , (3) where Vref denotes the reference wind speed collected at the height of href and γ ∈ [0.1, 0.25] is the wind shear exponent, which varies with respect to the terrain [54]. 2.1.3. Micro-hydro plant Suneco Hydro’s XJ50-100SCTF6-Z 100-kW micro-hydro turbines are selected to be integrated into the run-of-the-river plant as part of the MG system [55]. The power output from the plant at each time-step [kW] can be estimated from Eq. (4) [56,57]. PMH(t) = NMH × ηMH,AC/DC × ηMH × ρ × g × hg × F(t) 1000 , (4) where NMH denotes the optimum quantity of turbines, ηMH is the total efficiency of the plant (including the turbine, generator, and water 1 In game theory, a Stackelberg duopoly is a non-symmetric, strategic, sequential game with one party, or a group of parties, taking over the leading position and the other(s) acting as follower(s). S. Mohseni et al.
- 8. Applied Energy 287 (2021) 116563 8 wheel efficiency), ηMH,AC/DC is the efficiency of the plant’s AC/DC con verter, ρ represents the density of water, g is the acceleration due to gravity, hg is the gross head (which is defined as the difference between the head race and tail race levels when water is not flowing), F(t) is the flow rate at time-step t [m3 /s], while the numeric value of 1000 converts the unit of measurement from Wh to kWh. 2.1.4. Biomass plant The integrated biomass gasifier-generator system PP30 Cogen-CS manufactured by All Power Labs [58] is utilised in this study. The plant, the flow diagram of which is shown in Fig. 4, is a commercially available, off-the-shelf component with an electrical rated power of 25 kW. The power output from the biomass plant at each time-step [kW] can be calculated from Eq. (5) [59]. PBP(t) = NBP × ηBP,AC/DC × ηBP × CV × MBP(t), (5) where NBP represents the optimal quantity of the considered biopower units, ηBP,AC/DC is the efficiency of the plant’s AC/DC converter, ηBP is the overall bio-electricity generation efficiency of the system, CV stands for the gross calorific value of the biomass feedstock, and MBP(t) denotes the feedstock mass consumption rate at time-step t [kg/h].2 Furthermore, the system is characterised with a carbon emission factor of 1.53 kg-CO2 per kg of feedstock used [60]. Accordingly, the social cost of the carbon emissions needs to be factored into the decision- making–for an eco-design of the MG system. The following equation can be used to calculate the life-cycle penalty imposed on the MG for CO2 emissions: Fig. 1. Overview of the section-wise modelling procedure employed in this paper for the aggregator-mediated, market-driven integration of flexible demand re sources in the long-term planning of MGs. 2 Note that the rated powers of micro-hydro turbines and biopower plants, are incorporated into the model and the decision-making process in an indirect manner using the power rating-dependent parameters–hg in the case of micro- hydro turbines, and MBP in the case of biopower units–as well as specifically developed terminal constraints (refer to Section 4.2.6 for more details). S. Mohseni et al.
- 9. Applied Energy 287 (2021) 116563 9 costem = ξCO2 1000 × ECO2 × ∑ T t=1 MBP(t), (6) where ξCO2 [$/tCO2] denotes the social cost of CO2 emissions used as a reference to account for life-cycle GHG impacts of the biopower plant in the model, and ECO2 represents the CO2 emission factor of the plant [kg- CO2/kg-feedstock]. 2.1.5. Upstream power grid The MG system is tied to the upstream electricity network through a dedicated bidirectional MV/LV transformer, the optimal capacity of which is under investigation. The cost imposed by purchasing electricity from the grid at each time-step could be represented by Eq. (7), while the income generated by the MG’s electricity exports is obtained from Eq. (8) [61]. costim(t) = πim(t) × Pim(t) × Δt, (7) incomeex(t) = πex × Pex(t) × Δt, (8) where πim(t) represents the (time-varying) wholesale electricity market price at time-step t [$/kWh], πex is the utility grid’s single-tier (flat) buy- back rate [$/kWh], Pim(t) is the amount of power imported from the utility grid at time-step t, Pex(t) is the amount of power exported to the Fig. 2. Micro-grid system architecture and streams of energy driven by renewables and the upstream grid. Fig. 3. Power curve of the ECO 48/750. Data Source: [52]. Fig. 4. Schematic diagram of the considered integrated biomass gasifier- generator system. Source: [60]. S. Mohseni et al.
- 10. Applied Energy 287 (2021) 116563 10 main grid at time-step t, and Δt represents the length of each time-step. The power exchange is expected to adhere to the following con straints: Pim(t)/ηT ≤ NT , (9) Pex(t)/ηT ≤ NT , (10) where ηT denotes the transformer’s efficiency and NT represents the rated capacity of the transformer, which is to be optimised. The generic solid-state transformer, designed by Qin and Kimball [62], is used in this study. The size of the transformer is characterised by the apparent power [kVA] and, as a simplifying assumption, the power factor is assumed to be 0.95. 2.1.6. Power conversion apparatuses As shown in Fig. 2, the MG system is equipped with several con verters to serve the purpose of coupling the equipment to a common DC busbar. For electrical loads, Leonics’ GTP-519-S 900-kW, GTP-506 115- kW, and GTP-501 33-kW inverters are considered in this study [63]. To calculate the size of the electrical loads’ inverters, first, the following equation is used to determine the nominal power of the overall power inversion system: NI = PL,max ηI , (11) where PL,max represents the demanded annual peak electrical loads and ηI identifies the power inversion equipment’s efficiency. Then, NI is rounded up to the next integer and the number of each inverter model is identified by the following equations, which give priority to higher-rated inverters as they carry a lower per-unit cost: N900 = ⌊ NI C900 ⌋ , (12) N115 = ⌊ NI − (N900C900) C115 ⌋ , (13) N33 = ⌈ NI − (N900C900)− (N115C115) C33 ⌉ , (14) where N900, N115, and N33 respectively denote the quantity of the 900- kW, 115-kW, and 33-kW inverters, while C900, C115, and C33 indicate their respective rated capacities. 2.1.7. Internal backup energy storage The proposed system leverages the temporal characteristics of various DERs providing backup power, or energy storage. To this end, this study expands on the idea proposed by Akram et al. [64] that low- pass energy filters could be used to calculate the share of each energy storage medium in supplying load power demand on a representative MG. Accordingly, the power mismatch signal is first broken down into the low- and high-frequency components using a low-pass filter with a transfer function given in Eq. (15). H(s) = Kω2 0 s2 + (ω0/Q)s+ω2 0 , (15) where ω0 denotes the cut-off frequency, K represents the DC gain, and Q = 1/2ξ identifies the quality with ξ indicating the damping factor. Then, the low-frequency signal is directed to the hydrogen system (including the electrolyser, hydrogen tank, and the fuel cell), while the high-frequency signal is transferred to the hybrid battery-SC system. Subsequently, another low-pass filter with a lower cut-off frequency identifies the contribution of the battery and SC banks in serving loads or storing surplus power. The technical rationale underlying this power allocation approach is the longer cycle life, higher round-trip efficiency, and more rapid response capability of SCs (batteries) to balance out generation-demand mismatches than batteries (the hydrogen system). That is, the shortest and longest periods of surplus or shortage of electricity are addressed using the SC bank and hydrogen system, respectively, while the battery bank bridges the gap between these two storage media.3 2.1.7.1. Super-capacitor bank. Eaton’s 48-V, 166-F XLR-48R6167-R SC modules [65], which are of the type electrochemical double-layer capacitor (EDLC), are used to address short-term renewable power and load demand mismatches–and improve the MG’s dynamic response and overall efficiency. The SC bank’s energy content at each hour of the MG operation can be calculated as follows: ESC(t) = ESC(t − 1) + ( Pch,HF2 − ( Pdch,HF2 ηSC ) ) × Δt, (16) where ηSC represents the SC’s round-trip efficiency, while Pch,HF2 and Pdch,HF2 are the high-frequency components of the outputs of the second- stage filtered charging and discharging signals, respectively. 2.1.7.2. Battery bank. CellCube’s vanadium redox flow-based battery bank [66] is used in the conceptualised MG. Likewise to the inverter system, three different battery product models are selected and the same procedure is followed to apportioning the total optimal size of the bat tery bank to different model types, following the same logic. The battery product models are: FB 10–100 (100 kWh), FB 200–400 (400 kWh), and FB 400–1600 (1600 kWh). The battery bank’s energy content at each hour can be obtained as follows: EB(t) = EB(t − 1) + ( Pch,LF2 − ( Pdch,LF2 ηB ) ) × Δt, (17) where ηB is the battery bank’s round-trip efficiency, while Pch,LF2 and Pdch,LF2 denote the low-frequency components of the outputs of the second-stage filtered charging and discharging signals, respectively. 2.1.7.3. Hydrogen storage. The hydrogen-based storage system mainly includes polymer electrolyte membrane (PEM) electrolyser stacks, a medium-pressure (20 bar) hydrogen reservoir, and stationary PEM fuel cell stacks. H-TEC Systems’ S 30/50 5-kW electrolyser stacks [67], a generic hydrogen reservoir (which needs to be customised), and Bal lard’s FCgen-1020ACS 3.3-kW fuel cell stacks [68] are used as part of the hydrogen storage system. The hydrogen power directed from the elec trolyser outlet to the reservoir at time-step t can be obtained as follows: PE− HT (t) = PE(t) × ηE, (18) where PE is the electrolyser’s consumed power, which is controlled by the low-frequency component of the output of the first-stage filtered charging signal, while ηE denotes the electrolyser’s efficiency. The mass of hydrogen, mHT [kg], stored in the reservoir at each time- step can be calculated as follows: EHT (t) = EHT (t − 1) + ( PE− HT (t) − (PHT− FC(t) + PHT− S(t)) ηHT ) × Δt, (19) mHT (t) = EHT (t) HHVH2 , (20) where EHT represents the reservoir’s energy level, PE− HT is the directed 3 Note that the backup power allocation strategy employed in this study is tailored towards long-term capacity planning, at which stage long-term fore casted data are available. A forward-looking predictive modelling approach (using a critic network, for example) is indispensable for the real-time operation phase. S. Mohseni et al.
- 11. Applied Energy 287 (2021) 116563 11 hydrogen power from the electrolyser to the reservoir, PHT− FC and PHT− S respectively denote the hydrogen power consumption of the fuel cell and the FCEV parking lot, ηHT represents the tank’s round-trip efficiency, and HHVH2 stands for the higher heating value of hydrogen. The electric power output from the high-energy-density fuel cell at time-step t, which is controlled by the low-frequency component of the output of the first-stage filtered discharging signal, can be obtained using Eq. (21). PFC(t) = PHT− FC(t) × ηFC, (21) where PHT− FC represents the fuel cell’s consumed hydrogen power and ηFC denotes its electrical efficiency, which is defined as the ratio between the electricity generated and the hydrogen consumed. 2.1.8. Fuel cell electric vehicle parking lot The hydrogen refuelling infrastructure of the parking lot mainly consists of a medium-pressure (20/350 bar) compressor, a buffer stor age, a cryogenic pump, as well as a vaporiser, a refrigeration unit, and some dispensers to deliver gaseous hydrogen fuel to FCEVs [69]. The refuelling infrastructure is modelled by its overall efficiency, which is denoted by ηS. To this end, the Pure Energy Centre’s customised hydrogen refilling station [70] is considered for integration into the proposed MG. 2.1.8.1. Selected fuel cell electric vehicles. A fleet of ultra-light-duty personal passenger vehicles is planned for integration into the envi sioned system through the coordinated use of the refuelling infrastruc ture. Accordingly, vehicles are assumed to be refuelled on a first-come/ first-served basis using the multi-server Erlang-C queuing model [71], where C identifies the optimal number of dispensers. Also, FCEVs are assumed to be of the model Riversimple Rasa. 2.1.8.2. Fuel cell electric vehicles in vehicle-to-grid operation. To provide a platform for exploiting the V2G capabilities of the FCEVs, the FCEV2G setup designed in [72] is used in this study. The setup enables the conversion of the DC power of the vehicle’s fuel cell engine into AC that can be directed to the input port of the electrical loads’ inverter after frequency synchronisation, with an overall efficiency of ηFCEV2G. Accordingly, modulation of the power output from each FCEV, the owner of which aspires to participate in the V2G operations, can be made from 0 to 8.5 kW DC–in compliance with the nominal capacity of Rasa’s built-in fuel cell. This means the costs arising from payments made to FCEV owners to provide V2G power at each time-step–under a feed-in-tariff style programme–can be calculated by the following equation: costFCEV2G(t) = πFCEV2G × ηFCEV2G × PFCEV2G(t) × Δt, (22) where πFCEV2G represents the per-unit premium tariff rate for V2G power [$/kWh] and PFCEV2G(t) denotes the amount of V2G power used for operational scheduling at time-step t. For the sake of simplification, it was assumed that at each time-step of the MG operation, the maximum amount of available V2G power that can be provided by the station at each time-step, Pmax FCEV2G(t), equals 25% of the load reduction potential of the station at that time-step. 2.1.9. Data: Selected product models The values of the underlying system scalars, defined above, are presented in Table 2. Also, the techno-economic specifications of the MG equipment, namely the capital, replacement, and operation and main tenance (O&M) costs, as well as the estimated service life and efficiency of the equipment are summarised in Table 3. 2.2. Operational strategy A rule-based, hourly-basis, cycle-charging operational strategy is adopted in this study for the dispatch of energy within the MG system, which is illustrated by the flowchart in Fig. 5. In the devised energy scheduling plan, (1) energy storage devices and FCEVs are recharged/ refilled using only the surplus non-dispatchable renewable power, (2) non-dispatchable renewable power and electrical loads are partitioned into the ultra-high, high, and low-frequency components and then stored/supplied within/using the SC bank, battery bank, and the hydrogen tank/fuel cell, respectively, (3) the dispatchable biopower plant can only be operated during the time-slots stamped as peak hours to partially or wholly offset the lack of sufficient fuel cell power,4 (4) the upstream grid serves as the ultimate guarantor of the perfect satisfaction of the electric load demand, and (5) the FCEV2G capability is considered as a resource to compensate for at least part of the electricity left un served by the fuel cell and the biopower plant, or the shortage of battery and SC capacity to meet the load power demand. To this end, morning and evening peak demand were assumed to occur between the hours of 6 a.m. to 10 a.m. and 5 p.m. to 9 p.m., respectively–in compliance with historical records of electricity consumption in New Zealand. Moreover, the key assumptions made in conceptualising the pro posed MG system and conducting the life-cycle analysis are listed in Supplementary Material (Additional File 1: Key assumptions underlying the conceptualised micro-grid model). 3. Aggregator-mediated, incentive-based demand-side management market design This section presents a mathematical formulation of a two-stage, aggregator-mediated, incentive-based DSM market model specifically developed for integration into standard MG capacity planning ap proaches. Building on the interruptible load programmes, the model is designed specifically to improve the accuracy of projections of the small- to medium-scale DR resource availability across different end-use sec tors–residential, commercial, industrial, agricultural, and electrified transportation. More specifically, it characterises the interactions be tween a MG operator, DRAs, and end-consumers. To this end, the model consistently treats these three sets of actors as rational, utility- Table 2 Data values and sources for the proposed micro-grid system scalars. Scalar Value Source Scalar Value Source CV 5.07 kWh/kg [73] Kp − 0.40%/◦ C [48] Δt 1 h (this paper) NMOT 43 ◦ C [48] ηPV,DC/DC 95% [34] ρ 1000 kg/m3 − ηMH,AC/DC, ηBP,AC/DC 95% [34] PBP,r 25 kW [58] DF 85% [74] πex $0.05/kWh [75] ECO2 1.53 kg-CO2/ kg-feedstock [60] πFCEV2G $0.05/kWh (this paper) g 9.81 m/s2 − PL,max 7.31 MW (this paper) h 55 m [52] PMH,r 100 kW [55] hg 10 m [55] PPV,r 0.28 kW [48] href 10 m [76] TSTC 25 ◦ C [78] HHVH2 39.7 kWh/kg [77] γ 0.15 [79] ISTC 1 kW/m2 [78] ξCO2 $42/tCO2, $50/tCO2* [80] * A central value of $42/tCO2 is applied for the first 10-year planning horizon (covering the years 2020 to 2030), which rises to $50/tCO2 for the second half of the projected lifespan of the project in accordance with the Obama adminis tration’s central estimates [80]. 4 This assumption can be explained by the relatively long cold start-up time of the biopower plant (i.e. ~10–15 min) and the inefficiency of leaving the bio power plant on standby at all times. S. Mohseni et al.
- 12. Applied Energy 287 (2021) 116563 12 maximising (self-interested), active economic agents. The proposed market design provides a forum for these economic agents to negotiate on how to mutually optimise their objective functions in non- cooperative (strategic) game settings under the Stackelberg competi tion. It also identifies the minimum operational MG costs based on hourly priced DR products and the wholesale power price. In this way, the model enables all the active agents within the MG to be involved in co-designing a business model for more independent energy procure ment. Fig. 6 displays a schematic of the overall structure of the model with the sequence of incentive price/DR supply communications be tween the market participants overlaid [96]. As Fig. 6 shows, the market-based, aggregator-mediated DSM strat egy is modelled as an interactive hierarchical decision-making process, which consists of two levels of leader–follower relationships, namely between the MG operator and the DRAs (wholesale DSM market), and between the DRAs and their customers (retail DSM market). Although the DSM market participants are hierarchically related with respect to DR service, each has an independent viewpoint on the problem, which is modelled by specific objective functions in the following sub-sections. 3.1. Micro-grid operator It is assumed that the conceptualised MG, laid out in Section 2, runs on an energy-as-a-service business model in that not only does a third- party (private company) own the MG, but it also provides an over arching framework for energy management (through effective incentive Table 3 Data values and sources for techno-economic specifications of the conceptualised system’s components. Component Manufacturer part number Nameplate rating Capital cost* Replacement cost† Operation and maintenance cost† Expected service life Nominal efficiency Source Per unit Per standard unit of generation/ storage/ conversion capacity Notation Value [%] PV module CS6K-280P 280 W $210/ unit $750/kW $200/unit $1/unit/year 25 years ηPV 17.11 [48] Wind turbine ECO 48/750 750 kW $1.096 m/unit $1.46 k/kW $0.822 m/ unit $21 k/unit/ year 20 years N/A‡ N/A‡ [52] Micro-hydro turbine XJ50-100SCTF6- Z 100 kW $56 k/ unit $560/kW $17 k/unit $2 k/unit/year 25 years ηMH 78 [55] Biopower unit§ PP30 Cogen-CS 25 kW $32 k/ unit $1.28 k/kW $23 k/unit $0.01/unit/ hour 10 k hours ηBP 23 [58] Transformer Generic − − $65/kVA $55/kVA $2/kVA/year 30 years ηT 93 [62,81] Inverter GTP-501 33 kW $12 k/ unit $364/kW $12 k/unit $85/unit/year 15 years ηI 96 [63] GTP-506 115 kW $38 k/ unit $330/kW $38 k/unit $250/unit/ year GTP-519-S 900 kW $270 k/ unit $300/kW $270 k/unit $1.9 k/unit/ year Super- capacitor module XLR-48R6167-R 166F, 48 V ≡ 0.054 kWh $1.3 k/ unit $24.1 k/kWh $1.3 k/unit $13/unit/year 1 m cycles ηSC 97 [65] Battery pack FB 10–100 100 kWh $110 k/ unit $1.1 k/kWh $110 k/unit $220/unit/ year 20 years with unlimited cycles ηB 80 [66,82] FB 200–400 400 kWh $400 k/ unit $1 k/kWh $400 k/unit $840/unit/ year FB 400–1600 1600 kWh $1.442 m/unit $901/kWh $1.442 m/ unit $4 k/unit/year Electrolyser stack S 30/50 5 kW $6 k/ unit $1.2 k/kW $6 k/unit $120/unit/ year 20 years ηE 75 [67] Hydrogen tank Generic − − $500/kg $500/kg $1/kg/year 20 years ηHT 95 [83] Fuel cell stack FCgen-1020ACS 3.3 kW $5 k/ unit $1.52 k/kW $5 k/unit $0.02/unit/ hour 10 k hours ηFC 40 [68] Hydrogen station Generic (Pure Energy Centre) − − $10 k/(kg-H2/h) $5 k/(kg-H2/ h) $350/(kg-H2/ h)/year 20 years ηS 95 [69,70,84] Generic (The Energy Technology Section, TU Delft)¶| − − $155/kW $95/kW $32/kW/year 20 years ηFCEV2G 44# [85,86,87,88] ¶ In view of the assumption that the DC power provided by the FCEVs is fed into the electrical loads’ inverter, the costs associated with the FCEV2G technology only include the costs of modifying the vehicles with a V2G DC outlet plug. * All of the reported capital costs represent the actual cost of buying the selected components in New Zealand’s energy asset market as of October 2019–which were adjusted to 2019 U.S. dollars. In October 2019, US$1 = NZ$1.56. † All of the replacement and O&M costs were calibrated in accordance with the component-specific ratios of capital to replacement and O&M costs reported in [82,83,89,90,91,92,93,94,95]. ‡ Not applicable as the wind turbine efficiency is reflected in its power curve shown in Fig. 3. § To value the positive impact of the biopower plant on the internal dispatchability of the MG, the total discounted cost of pellet feedstock was considered to be an exogenous variable, which is determined outside the model based on the imposed emission credits (see Eq. (6)) with respect to the total discounted energy output of the plant (see Eq. (5)). # The V2G infrastructure’s efficiency in this paper represents a tank-to-DC-bus efficiency (units converted based on the higher heating value of hydrogen). S. Mohseni et al.
- 13. Applied Energy 287 (2021) 116563 13 arrangements reflective of wholesale market prices) tailored to the needs of the MG. Specifically, on a 24-h day-ahead basis, the MG operator predicts the net energy deficit of the MG, which needs to be procured by a combi nation of imported power and customer-supplied DR units. Accordingly, it sends an incentive payment signal to the aggregators to induce lower energy use at times of high wholesale power prices, when the total power output from the renewable power generation technologies is low, or during periods when reserve shortfalls arise. Equation (23) expresses the objective function of the MG operator, which needs to be minimised for each critical hour of the next day (t ∈ Pd⊂T = {1,2,⋯,8760}) subject to the constraints in Eqs. (24) and (25): OCMG(t) = costim(t) + IMGO(t) × ∑ j∈J Dj LA(t)∀t, (23) Imin MGO ≤ IMGO(t) ≤ Imax MGO∀t, (24) Ddef (t) = Pim(t) + ∑ j∈J Dj LA(t)∀t, (25) where OCMG is the MG’s operational cost defined based on the cost of the imported power, costim, and the total incentive payments for load Fig. 5. Flowchart of the MG’s energy management scheme, consisting of a set of pre-defined control logics. Fig. 6. General architecture of the proposed two-stage, aggregator-mediated, incentive-based DSM market design. S. Mohseni et al.
- 14. Applied Energy 287 (2021) 116563 14 reduction, ∑ j∈JD j LA; IMGO is the MG operator-posted incentive price signal to the wholesale DSM market, with superscripts min and max representing its lower and upper limits, respectively; Ddef denotes the net energy deficit of the system; and D j LA is the total load reduction procured by DRA j ∈ J. 3.2. Demand response aggregators The DRAs serve as a go-between, interfacing with the smaller DR providers and the broader MG system operator so as to maintain the visibility of the small-scale DR products. The independence of the DRAs is fully preserved in the proposed model as they are precluded from ownership of the energy infrastructure. Precisely, third-party aggre gators enlist end-consumers to take part in interruptible load pro grammes. To this end, they take a percentage of the MG operator-offered incentive as compensation, passing the rest on to their customers. More specifically, the DRAs aim to maximise the objective (profit) function in Eq. (26) subject to the constraints in Eqs. (27) and (28) [97]: Prj LA(t) = ( IMGO(t) − Ij LA(t) ) × Dj LA(t)∀j, t, (26) Ij,min LA ≤ Ij LA(t) ≤ Ij,max LA ∀j, t, (27) Dj LA(t) = ∑ k∈NJ dk,j (t)∀j, t, (28) where I j LA is the incentive rate posted by the j-th aggregator to the retail DSM market; dk,j denotes the capacity of DR product supplied by customer k subscribed to aggregator j; NJ is the set of customers serviced by aggregator j, which is a proper subset of set of all the customers within the MG system’s operational territory, K; and I j,min LA and I j,max LA respectively represent the lower and upper bounds of the incentive payments offered by aggregator j. 3.3. End-consumers End-consumers, who are activated by third-party DRAs, have the opportunity to take full advantage of their flexibility potential, whilst adhering to a set of discomfort cost constraints. To this end, the end- consumers determine the optimum supply of their DR resources with respect to the DRA-offered incentive prices by maximising the utility function expressed in Eq. (29) subject to Eqs. (30) and (31). Uk,j (t) = dk,j (t) × Ij LA(t) − disk,j (t)∀k, t, (29) 0 ≤ dk,j (t) ≤ dk,j ncr(t)∀k, t, (30) dk,j full(t) = dk,j cr (t) + dk,j ncr(t)∀k, t, (31) where disk,j denotes the cost of discomfort (inconvenience) associated with load reductions as a measure of the value of electricity, which can be obtained from Eq. (32)5 [98,99], and must lie within a certain range, as constrained by Eq. (33); d k,j full is the full (original) load demanded by customer k of aggregator j; d k,j cr is the critical portion of the original load, any shedding of which results in impaired reliability; and d k,j ncr is the non- critical (dispatchable) demand, which can be interrupted by making effective incentive payments to customers for curtailing load. disk,j = ck,j 1 (dk,j )2 + ck,j 2 (1− δj)dk,j ∀k, t, (32) disk,j,min ≤ disk,j ≤ disk,j,max ∀k, t, (33) where c k,j 1 and c k,j 2 are positive individual-level parameters specified by end-consumers that characterise their sensitivity to load reductions, for customers indifferent to incentive payment, c k,j 1 ,c k,j 2 →∞; 0 ≤ δj ≤ 1 is the sector-level elasticity of customer-supplied DR capacity, for a hypo thetical completely inelastic customer category, δj→0; while disk,j,min and disk,j,max respectively denote the minimum and maximum allowable Fig. 7. Sequence diagram of implementing the proposed DSM model in the context of the conceptualised MG system. 5 The customer discomfort cost function can be viewed as the second-order best-fit equation to individual-level, user-specified data points representing ordered pairs of DR capacity supply and the associated discomfort cost incurred. S. Mohseni et al.
- 15. Applied Energy 287 (2021) 116563 15 limits for customer-specific discomfort costs. Incorporating the term ( − c k,j 2 δjdk,j ) in Eq. (32) ensures that the market equilibrium of the two-stage aggregator-mediated DSM game is aware of the marginal values the end-users across different sectors place on an uninterrupted power supply–that is, the value to consumers of the last (incremental) unit of DR capacity supply. It should be noted that this analysis does not account for the supply elasticity of inframarginal DR capacity. 3.4. Solution algorithm To solve the proposed two-stage, aggregator-mediated, incentive- based DSM market model, a distributed algorithm approach, which dynamically updates the MG operator-offered incentive price, is implemented. The idea is to update the MG operator-posted incentive from Imin MGO to Imax MGO with an increment size of iMGO and determine the hourly operational cost of the MG as a function of the wholesale power price and contributed load reductions. The model is solved repeatedly for different values of the MG operator-offered incentive prices until no further improvement (reduction) in the MG operational cost occurs (terminating condition). Algorithm 1 presents the distributed algorithm developed to quantify the optimal trade-off between the imported power and dispatched load reduction during the critical hours of MG operation in terms of on-site resource adequacy. The superscript “*” in the algo rithm denotes the global optimality. Algorithm 1. (Proposed distributed algorithm to produce the optimal day- ahead trade-offs between imported power and exploited DR resources dur ing the critical peak hours of MG operation) 1: Initialise: I* MGO = 0 and OC* MG(t) = costim(Ddef ) 2: for IMGO ranging from Imin MGO to Imax MGO at steps of iMGO do 3: Submit the incentive price signal IMGO to the wholesale DSM market 4: for each DRA j ∈ J do 5: Determine the best-strategy incentive rate to be offered to the end-users, Ij,* LA, by setting the first-order derivative of the DRA’s profit function in Eq. (26), in which dk,j is substituted with the best-response strategy of the corresponding customers derived by setting the first-order derivative of their utility functions in Eq. (29) equal to zero 6: Send the incentive price signal Ij,* LA to the corresponding customers 7: for each customer k ∈ NJ do 8: Derive the customer’s best-response strategy by setting the first-order derivative of its utility function given in Eq. (29) equal to zero 9: Calculate the best-response load reduction with respect to the financial incentive offered by the DRA it has subscribed to, using the customer-specific best-response strategy profile derived above 10: Send the amount of load curtailment contributed by the customer to the corresponding DRA 11: end for 12: Aggregate the load reductions supplied by the end-users 13: Send the total load reduction procured by the DRA to the MG operator 14: end for 15: Update the hourly operating cost of the MG as: OCMG = costim ( Ddef − ∑ j∈JD j LA ) + IMGO × ∑ j∈JD j LA 16: if (OCMG < OC* MG) then 17: Update the optimal MG operator-posted incentive and the MG’s operating cost as: I* MGO = IMGO and OC* MG = OCMG 18: end if 19: end for 20: Return the set (I* MGO,I j,* LA,dk,j,* ) as the unique, globally-optimum equilibrium solution for each hour of the coming day Algorithm 1 determines the unique, pure-strategy Nash equilibrium of the game, which identifies the best-response strategies of the DRAs and end-consumers by setting the first-order derivatives of their objec tive functions equal to zero. To prove that doing so maximises the cus tomers’ utility functions and the aggregators’ profit functions (and yields the unique, globally-optimum solutions), one must show the concavity or convexity of these payoff functions. Taking the second-order derivative of Uk,j given in Eq. (29) with respect to the customer-supplied DR capacity yields: ∂Uk,j ∂dk,j = Ij LA − ( 2ck,j 1 dk,j +ck,j 2 ( 1 − δj ) ) , (34) ∂2 Uk,j ∂(dk,j)2 = − 2ck,j 1 . (35) Substituting the best-response strategies of end-consumers–obtained by setting the first-order derivative of their utility function, derived in Eq. (34), equal to zero–into the profit function of the DRAs given in Eq. (26), yields: Prj LA = ( IMGO − Ij LA ) × ∑ k∈NJ Ij LA− ck,j 2 ( 1 − δj ) 2ck,j 1 = − ( Ij LA )2∑ k∈NJ 1 2ck,j 1 + Ij LA( ∑ k∈NJ ck,j 2 ( 1 − δj ) 2ck,j 1 + ∑ k∈NJ IMGO 2ck,j 1 ) + IMGO ∑ k∈NJ − ck,j 2 ( 1 − δj ) 2ck,j 1 . (36) Then, the second-order derivative of Prj LA, re-written in Eq. (36), with respect to the aggregator-offered incentive payments can be obtained as follows: ∂Prj LA ∂Ij LA = − Ij LA ∑ k∈NJ 1 ck,j 1 + ( ∑ k∈NJ ck,j 2 ( 1 − δj ) 2ck,j 1 + ∑ k∈NJ IMGO 2ck,j 1 ), (37) ∂2 Prj LA ∂(Ij LA) 2 = − ∑ k∈NJ 1 ck,j 1 . (38) Given the positive value of ck,j 1 , the second-order derivatives of Prj LA and Uk,j are strictly negative. This implies that Prj LA and Uk,j are strictly concave over the feasible regions of Ij LA and dk,j , respectively. Accord ingly, this proves that setting the first-order derivatives of the aggre gators’ and end-consumers’ objective functions equal to zero is guaranteed to yield the unique, globally-optimum solutions. 3.5. Communication sequence Furthermore, to help visualise the sequence of actions and reactions required to execute the proposed interruptible DR market design, the application-driven sequence diagram of Algorithm 1 is presented in Fig. 7 for the conceptualised MG, laid out in Section 2. As illustrated in the figure, the process starts by communicating the day-ahead state estimates of non-controllable renewables and energy reserves from one utility-owned entity, the MG asset manager, to another utility-owned entity, the MG operator. After receiving a response to its enquiry regarding the availability of biomass resources from the MG asset manager, the MG operator sends financial incentive signals to the DR aggregators and asks about the amount of available interruptible loads at each hour of the upcoming day. To this end, a two-stage iterative Stackelberg incentive price game is run in accordance with Algorithm 1, which enables decentralised decision-making. Specifically, at the top level (wholesale market), the MG operator is the leader and the DRAs are the followers. The DRAs are, at the same time, the leading players at the bottom level (retail market), where end-consumers serve as final fol lowers. Note that the MG operator calls a DR event and sends the incentive price signals to the aggregators for the time-steps at which a net energy deficit is predicted. The proposed DR scheduling framework, shown in Fig. 7, forms part of the input to the hourly energy management strategy of the proposed equipment capacity-planning method, the flowchart of which is pro vided in Fig. 5. That is, the energy demand data input to the flowchart is aware of the interruptible demand resources–or, better put, both the S. Mohseni et al.
- 16. Applied Energy 287 (2021) 116563 16 power and hydrogen demand on the system are scaled-down (modified) through running the proposed DR scheduling framework for the specific peak hours of each day of the representative year before being fed to the hourly operational scheduling strategy outlined in Fig. 5. The process continues by transmitting the aggregator’s incentives for load reduction to their corresponding customers, and completes by clearing the DSM markets respectively at the local (retail) and wholesale levels. As mentioned above, this procedure is repeated for each hour of a repre sentative hourly-basis, one-year operational timeframe. To this end, the year-long demand profiles are decomposed into daily profiles so as to be used in the day-ahead DR management plan of the MG (see Fig. 7), the DR-adjusted values of which are then used in the course of the hourly energy management of the system (see Fig. 5). 4. Micro-grid capacity-optimisation model This section explains the deterministically estimated life-cycle cost of the conceptualised MG system before describing how the proposed non- cooperative game-theoretic DR management scheme is integrated into the MG sizing model. The MG capacity-optimisation model consists of three key elements: (1) the net present cost (NPC) and net present value (NPV) methods utilised to formulate the total discounted system cost function, (2) the loss of power supply probability (LPSP) technique to quantify the reliability of the system in servicing the electrical and hydrogen load demands, and (3) the moth-flame optimisation algorithm (MFOA) [100] as a single-objective meta-heuristic optimisation algo rithm to find the globally optimum solution to the problem by mini mising the life-cycle cost of the MG, whilst adhering to the technical, reliability, and systemic constraints (see Supplementary Material (Additional File 2: Techniques used in the micro-grid capacity-optimi sation model) for details). The superiority of the single-objective MFOA to the well-established meta-heuristics in the MG investment planning literature–for instance, the genetic algorithm (GA) [101] and the PSO [102]–as well as to a wide variety of state-of-the-art meta-heuristics in terms of nearing the globally optimum solution is supported in previous studies [41,84,103,104,105]. 4.1. Objective function A static analysis of expected future cash flows for the underlying project lays the basis for the mathematical formulation of the objective function. The whole-life cost of the MG based on the NPC and NPV calculations, which is to be minimised, can be expressed as follows: minWLC =( ∑ c∈C NPCc) + NPCI + NPV ( ∑ 8760 t=1 OCMG(t) ) + NPV ( ∑ 8760 t=1 costem(t) ) + NPV ( ∑ 8760 t=1 costFCEV2G(t) ) − NPV ( ∑ 8760 t=1 incomeex(t) ) + penconst, (39) Where NPCc represents the NPC of the components, the optimal size of which is under investigation and are indexed by c ∈ C = {PV,WT,MH,T, E, FC, HT, BP, B, SC, S, FCEV2G}; NPCI denotes the NPC incurred by the inverter; OCMG is the operational cost of the MG to serve the unmet loads, either by paying incentives for load reduction or purchasing power from the upstream grid, as defined in Eq. (7); costem is the cost imposed on the system for buying emission credits on account for running the biopower plant, as given in Eq. (6); costFCEV2G denotes the cost resulting from the provision of FCEV2G services, as expressed in Eq. (22); incomeex is the income generated by selling the surplus power to the main grid, as expressed in Eq. (8); while the term penconst enforces the solutions to meet the constraints set out in Section 4.2. In this context, the useful life of the project was considered to be 20 years and the real interest rate was set to 3.7%. The real interest rate was projected by taking the mean of the historical records in New Zealand over a 10-year period, between 2010 and 2019 [106]. 4.2. Problem constraints The objective function presented above is subject to various sets of constraints along the following lines. 4.2.1. System reliability The LPSP reliability metric is employed to characterise the system performance over its projected 20-year life span. To this end, two separate LPSP indices are used to evaluate the reliability of electricity and hydrogen supply, which are constrained by Eqs. (40) and (41), respectively. LPSPe ≤ LPSPmax e , (40) LPSPH2 ≤ LPSPmax H2 , (41) where LPSPmax e and LPSPmax H2 denote the imposed upper bounds on LPSPe and LPSPH2 , respectively. 4.2.2. System-wide power balance According to Eq. (42), at each time-step of the system operation, the sum of all of the internally generated energy components, energy re leases from the storage media, energy imports from the main grid, and any unmet load must be equal to the sum of the total energy consumed within the MG (to serve the loads or to charge the energy storage de vices) and any power sold to the upstream grid. PPV (t) + PWT (t) + PMH(t) + PBP(t) + Pdch(t) + PFC(t) + Pim(t) + PFCEV2G(t) + QL(t) ηI + QH2 (t) ηS = Pch(t) + PE(t) + Pex(t) + PL(t) ηI + PS(t) ηS ∀t, (42) where QL(t) and QH2 (t) respectively represent the unmet electrical and hydrogen demands at time-step t, which are used in the LPSP calculations. 4.2.3. Demand response scheduling As mentioned previously, under equilibrium conditions of the pro posed two-stage, aggregator-mediated, market-driven DR arrangement, the constraints in Eqs. (24), (25), (27), (28), (30), (31), (33) must be relaxed. 4.2.4. Energy storage systems and fuel cell electric vehicles The optimisation of the MG equipment capacity must additionally adhere to some constraints in terms of charge/discharge rate limits of the energy storage media and FCEVs, bounding the state of charge/ hydrogen of the storage systems and vehicles, as well as the state of energy reserves in the first and last operating hours, which could be expressed mathematically as: Ees,min ≤ Ees(t) ≤ Ees,max∀t, es, (43) Pch,min es ≤ Pch es (t) ≤ Pch,max es ∀t, es, (44) Pdch,min es ≤ Pdch es (t) ≤ Pdch,max es ∀t, es, (45) Ees− {FCEV}(0) = 0.5 × Ees− {FCEV},max∀es, (46) Ees− {FCEV}(8760) ≥ Ees− {FCEV}(0)∀es, (47) where Ees(t) is the energy content of the energy storage technology es ∈ ES = {B, SC, HT, FCEV} at time-step t; Ees,min and Ees,max respectively denote the minimum and maximum allowable energy content of energy storage technology es; Pch es (t) and Pdch es (t) respectively represent the S. Mohseni et al.
- 17. Applied Energy 287 (2021) 116563 17 charging and discharging rates of storage technology es at time-step t; Pch,max es and Pdch,max es are the maximum charging and discharging rates of storage technology es, respectively; and Pch,min es and Pdch,min es are the min imum charging and discharging rates of storage technology es, respec tively. The maximum allowable energy content of the battery bank, SC bank, and hydrogen tank are defined by their optimised capacity at each iteration of the optimisation process, whereas the maximum total energy content of the releasable hydrogen stored in the FCEVs’ tanks (max(Pmax FCEV2G(t)Δt) where t ∈ T) is limited by the maximum (optimal) capacity of the FCEV2G setup (as part of the hydrogen station) in addition to the stored hydrogen in the vehicles’ tanks at time-step t. That is, the variables Ees,max, es ∈ ES are treated as endogenous variables in the model. Also, the same principle holds for the variables Pch,max es and Pdch,max es . Moreover, in the interest of preventing the performance degrada tion–and mitigating the energy losses–during the start-up and shut- down cycles of the electrolyser, fuel cell, and biopower plant, a spe cific constraint preserves the durability of their operation. To this end, when the electrolyser, fuel cell, and biopower plant are started up, they are constrained to continue to run for at least tup time-steps–as a mini mum up-time constraint–at operating points equal to, or greater than the initially adjusted operating points. Accordingly, the power output from the fuel cell and biopower plant are treated as negative loads in the course of the MG operation on the extra hours mentioned above, whilst also being allowed to take higher operating point values where appro priate. In addition, to avoid severe pressure drops in the hydrogen tank, complete releases of hydrogen are prevented by enforcing EHT,max not to fall short of 5% of the optimised capacity of the tank. Also, to ensure that the design pressure of the tank is not exceeded, the upper limit on the energy content of the tank is set as 95% of its optimum capacity [107]. 4.2.5. Energy exchange The MG’s transactions of energy with the upstream power network is constrained by Eqs. (9) and (10) to adhere to the optimal size of the transformer connecting the MG system with the upstream grid at the point of common coupling (PCC). 4.2.6. Decision variables Specific upper bounds are set for maximum values the non-negative design variables can take, as represented in Eq. (48). These bounds are adjusted commensurate with the practical feasibility of implementing the conceptualised MG system in the considered area. For example, land limitations, characteristics of the catchment sites, available biomass as a feedstock, and acceptable emissions limits (from the biopower plant) could constrain the feasible solution space. Nc ≤ Nmax c ∀c, (48) where subscript c ∈ C indicates the MG components, the optimal size of which is under investigation, while the superscript max denotes the maximum permissible value of the optimum quantity/capacity of the equipment (Nc).6 4.3. Meta-heuristic optimisation algorithm Mathematically, the underlying MG capacity-planning model is a nonlinear, non-convex, non-deterministic polynomial time-hard (NP- hard) decision problem at its core, as indicated by Chen et al. [108]. Consequently, it cannot be solved exactly or by enumerating the entire search space explicitly or implicitly, but meta-heuristic techniques could be used effectively to solve the problem. As noted earlier, the MFOA is employed to optimise a solution to the formulated MG capacity-optimisation problem on account of its well- proven superior performance to a wide range of both the well- established and state-of-the-art meta-heuristics in the MG planning context. Furthermore, owing to the mixed-discrete-continuous structure of the formulated problem, the technique proposed by Chowdhury et al. [109] is employed to modify the original continuous MFOA to make it applicable to the problem at hand. Moreover, the control parameters of the algorithm were adjusted as suggested by its developer [100], while the number of search agents, NSA, and the maximum number of itera tions, Itermax, were set based on the findings of Khan and Singh [110] on the appropriate values to ensure the convergence of a broad spectrum of meta-heuristic optimisation algorithms–including both the well- established and state-of-the-art meta-heuristics–in the context of MG design optimisation and capacity planning. 4.4. Data: Adjusted demand-response integrated micro-grid equipment capacity planning model parameters Table 4 lists the chosen data values for the parameters used to build the proposed DR-integrated MG equipment capacity-planning model. 4.5. Overview of the proposed solution algorithm The flowchart of the proposed MG equipment capacity-planning model, which uses the proposed two-stage, aggregator-mediated market-driven DR model to realistically project the customer engage ment in incentive-based DR programmes–based on an economically stable allocation of the profits generated from interruptible load pro grammes between the sole energy service provider, DSM aggregators, and end-users–is presented in Fig. 8. As can be seen from the figure, the solution algorithm integrates the proposed DR provision framework (the yellow block) and applies the developed rule-based hourly-basis oper ational scheduling strategy (the light coral block), while taking an iterative approach to optimise the discounted MG investment cost with which to determine the respective size of the equipment (the blue blocks). Table 4 Data values for the demand response-integrated micro-grid equipment capacity planning model parameters. Scalar Value Scalar Value Ees,max (endogenous variable) Nmax FCEV2G 5,000 kW Ees− {HT},min 0 kWh* Nmax HT 50,000 kg EHT,min (endogenous variable) Nmax MH 30 iMGO $0.02/kWh Nmax PV 20,000 Imin MGO $0.02/kWh NSA 100 Imax MGO $0.32/kWh Nmax S 100 kg-H2/h Ij,min LA $(0.02− ε† )/kWh Nmax SC 10,000 I j,max LA $(0.32− ε† )/kWh Nmax T 8,000 kVA Itermax 500 Nmax WT 15 LPSPmax e 0% Pch,max es (endogenous variable) LPSPmax H2 5% Pch,min es ε† kW Nmax B 20,000 kWh Pdch,max es (endogenous variable) Nmax BP 50 Pdch,min es ε† kW Nmax E 1,000 penconst (1/ε† ) Nmax FC 2,000 tup 3 h * Note that the depth of discharge capability of the vanadium redox flow battery is 100% and the total energy content of the FCEVs’ tanks is assumed to be aware of the specific level of hydrogen expected (desired) by each FCEV owner at the scheduled departure time. † The symbol ε denotes a small positive infinitesimal quantity. 6 The maximum permissible values of the design variables are aware of the rated powers of the corresponding components. S. Mohseni et al.
- 18. Applied Energy 287 (2021) 116563 18 Fig. 8. Flowchart of the proposed non-cooperative game-theoretic DR-integrated approach for the optimal capacity planning of MGs. Fig. 9. Diagrammatic representation of the step-wise procedure for implementing the proposed optimal MG planning framework. S. Mohseni et al.
- 19. Applied Energy 287 (2021) 116563 19 Furthermore, the step-wise representation of the integrated simula tion platform to optimally design the conceptualised MG, while man aging the DR resources using the proposed DR scheduling approach is summarised in Fig. 9. After the procurement and pre-processing of the input data, the model is built up in a multi-layered structure, which consists of (from bottom to top): (1) a rule-based hourly energy sched uling strategy, (2) a two-stage, aggregator-mediated, DSM market design to arrange the delivery of the DR resources on a day-ahead basis, (3) various constraints the objective function is subjected to, and (4) the derived fitness function representing the whole-life cost of the system, which is to be optimised using the MFOA. 5. Case study To confirm the proposition put forward in Section 1 on the effec tiveness of integrating the proposed DR management framework into the standard meta-heuristic-based MG capacity planning approach, as well as the viability of the conceptual test-case MG system, laid out in Section 2, this section presents the results of the case study analysis conducted for the town of Ohakune, New Zealand. To this end, first, the validity of the model is confirmed through a direct comparison of the extreme-case model results with those of a business-as-usual (BAU), non- game-theoretic interruptible DR scheduling framework. Then, the eco nomic viability of integrating the developed DSM strategies into the long-term MG investment decision-making processes is benchmarked against two cases where: (1) the DSM market is cleared without employing ideas from non-cooperative game theory for interactive decision-making regarding the practical capacity of DR resources, and (2) no provision is made to employ the responsive loads as a backup resource in the proposed MG system. Finally, a financial appraisal assessment demonstrates the economic sustainability of the proposed renewable energy project. Numerical simulations were carried out using the MATLAB software (version 9.5, R2018b) [111]. 5.1. Case study site: The town of Ohakune, New Zealand The notional MG system proposed in this study is envisioned to decarbonise the energy economy of the town of Ohakune, which is sit uated in the central part of the North Island of New Zealand–latitude 39.4180◦ S, longitude 175.3985◦ E [112]. The forecasted hourly-basis, year-long climatic input data streams, are presented as monthly mean 24-h profiles in 3D plots in Fig. 10 [76]. Also, the forecasted monthly averaged profile for biomass availability is shown in Fig. 11, assuming that the amount of monthly available biomass is evenly distributed over the days of the months [113,114]. The forecasted one-year load power demand on the system, which is represented as a monthly mean 24-h profile for greater clarity, is shown in a 3D plot in Fig. 12 (a) [115,116]. Also, the forecasted monthly mean Fig. 10. CliFlo-compliant forecasted meteorological input data (at an hourly resolution) for Ohakune, New Zealand: (a) solar irradiance; (b) ambient temperature; (c) wind speed; and (d) streamflow. S. Mohseni et al.
- 20. Applied Energy 287 (2021) 116563 20 24-h profile for the hydrogen demand of the refuelling sta tion–considering the seasonal variation in demands for transportation fuel as suggested in [117]–is shown in a 3D plot in Fig. 12 (b). The forecasted hourly-basis, year-long wholesale electricity price input data stream, πim(t), obtained using the weighted rolling average method, is shown as a monthly averaged daily profile in Fig. 13 [118]. More details of the case study site and the complete details on how the forecasted one-year profiles for climatological, load demand, and wholesale electricity price data are derived, can be found in Supple mentary Material (Additional File 3: Case study details). Table 5 presents the data values and sources for all parameters of the proposed two-stage, aggregator-mediated, incentive-based DSM frame work. In addition to the values of the model parameters defined in Section 3, Table 5 presents the number of customers signed up with each aggregator, which is denoted by N j cust. Moreover, given the New Zealand government’s aspirations of electrifying transport to help meet its target of net-zero greenhouse gas emissions by 2050, as well as the recent government-funded ‘Warmer Kiwi Homes’ programme offering up to 90% heat pump grants to low- income home owners, the penetration levels of light-duty FCEVs and heat pumps were assumed to be 40% and 60%, respectively at the time of commitment. Accordingly, smart charging of FCEVs and control of heat pump demand is of utmost importance to smooth and manage the overall load during peak periods. 5.2. Validation of the proposed demand-side management market To validate the effectiveness of the proposed two-stage aggregator- mediated DSM market model, two instances of day-ahead energy man agement analysis are conducted and the obtained results are compared with the case where the aggregator-mediated interruptible/curtailable DR resources are scheduled in a BAU way. Accordingly, the non-market- driven (BAU) procurement of aggregator-activated interruptible/cur tailable responsive loads excludes the ability to adaptively update the incentives offered by the MG operator, based on which the aggregators post their incentives to the retail DSM market, and subsequently the end- consumers select their participation rate in load reduction programmes. More specifically, the MG operator offers a fixed, day-specific rate of incentive to the aggregators, who also offer fixed levels of incentives to their customers–for load reduction during the peak hours of electricity consumption. Subsequently, the end-users and aggregators respond to the aggregator-determined and MG operator-offered incentive rates, respectively. In this way, the retail and wholesale DSM markets are sequentially cleared for the day-specific incentives by stacking the cus tomers’ and aggregators’ bids, low to high, and allocating demand reduction schedules to the customers and aggregators in the merit order irrespective of whether the power shortage is addressed with the best compromise between load reduction and imported electricity for each hourly period. Expectedly, as there exists no mechanism to update the Fig. 11. Monthly mean profile for the estimated total biomass available per month at the site: Ohakune, New Zealand. Fig. 12. Forecasted monthly mean 24-h profiles for the energy demand of the town Ohakune: (a) load power demand; and (b) hydrogen demand. Fig. 13. Forecasted monthly mean 24-h profile for the wholesale power price. S. Mohseni et al.
- 21. Applied Energy 287 (2021) 116563 21 initial strategy of the MG operator, the efficiency of such a framework is particularly sensitive to the choice of the MG operator-offered incentive rate. Hence, the model response is determined for various day-specific MG operator-offered incentive rates. Accordingly, Table 6 summarises the results obtained by simulating the above-described BAU interrupt ible DR mechanism when applied to the DR provision problem at hand in two extreme scenarios with the MG operator-offered incentive payment ranging from $0.02/kWh to $0.32/kWh in intervals of $0.02/kWh. Specifically, the two days that represent the most intense peak and trough on the year-round, mean daily load profile (consisting of the mean of the load power demand forecasts for 24 equidistant times in the course of each continuous 24-hour period of the representative year), namely July 21st and February 14th, were chosen for scenario analysis. The table, furthermore, presents the results of the suggested market- driven interruptible DR model for the extreme days considered. The following observations can be made from a comparative analysis of the proposed model and BAU model results presented in Table 6: 1. The systematic updating of the MG operator-offered incentive for load reduction–for the time-steps at which the system is predicted to be under stress–using an aggregator-mediated, market-driven DSM market model, can play a pivotal role in unlocking the full potential of demand-side resources by finding the economically efficient DR allocation solutions. In other words, the lack of a systematic Table 5 Data values and assumption sources for the two-stage, aggregator-mediated, incentive-based demand-side management framework. Parameter Aggregator Residential Commercial Industrial Agricultural FCEV-refuelling δj* Value 0.48 0.51 0.57 0.63 0.76 Source [119] [119] [119] [119] [120] c k,j 1 [$/kWh2 ] Range [1.08 × 10-3 , 1.15 × 10-3 ] [1.04 × 10-3 , 1.07 × 10-3 ] [0.99 × 10-3 , 1.03 × 10-3 ] [0.95 × 10-3 , 0.98 × 10-3 ] [0.91 × 10-3 , 0.94 × 10-3 ] Source† [121,122] [121,122] [121,122] [121,122] [121,122] ck,j 2 [$/kWh] Range [11.49 × 10-3 , 11.70 × 10-3 ] [11.31 × 10-3 , 11.48 × 10-3 ] [11.71 × 10-3 , 11.86 × 10-3 ] [11.25 × 10-3 , 11.30 × 10-3 ] [11.40 × 10-3 , 11.57 × 10-3 ] Source† [121,122] [121,122] [121,122] [121,122] [121,122] d k,j full [kWh] Range [8, 30] [20, 100] [100, 200] [30, 65] [5, 30] Source (this paper) (this paper) (this paper) (this paper) (this paper) d k,j ncr [kWh] Range [2.5, 16.5] [5, 60] [20, 84] [10, 46.2] [4, 25.5] Source (this paper) (this paper) (this paper) (this paper) (this paper) N j cust Value(s) 250 65 10 55 {1, 2, …, 150}‡ Source (this paper) (this paper) (this paper) (this paper) (this paper) * The load type-dependent DR procurement factor (sectoral elasticity of customer-supplied DR capacity) for the residential, commercial, industrial, and agricultural loads (normalised to the range [0, 1]) were adjusted in proportion with the weighted average values of unserved energy for various durations of interruption in a New Zealand context [119], while this factor for the FCEV-refuelling load was adjusted based on the plug-in EVs’ value of lost load reported in [120]. † The range of values the discomfort tolerance coefficients of customers can take were arbitrarily selected. The chosen values were guided by those used in [121,122] for the customer outage cost function coefficients for the relevant customer categories. Additionally, the range of sector-wide customer discomfort tolerance co efficients was normalised with respect to the corresponding load type-dependent DR procurement factor (in an inversely proportional manner). ‡ Since the number of FCEVs that utilise the filling station’s equipment varies with the time of day, it was modelled as a range of possible scenarios, i.e. the number of vehicles. Table 6 Comparative analysis of the proposed and BAU realisations of the interruptible DR programme on the extreme days: February 14th and July 21st. MG operator-offered incentive* (IMGO) [$/kWh] Total daily incentive payment to the aggregators (Ip MGO( ∑ p∈Pd ∑ j∈JD j,p LA)) [$/d] Total daily incentive payment to the customers ( ∑ p∈Pd ∑ j∈JI j,p LAD j,p LA) [$/d] Total daily load reduction procured by the customers ( ∑ p∈Pd ∑ j∈J ∑ k∈NJ dk,j,p ) [kWh/d] Total daily cost of electricity imports ( ∑ p∈Pd costp im) [$/d] Total daily operational cost of the MG ( ∑ p∈Pd OCp MG) ∑ p∈Pd OCp MG) [$/d] Feb. 14th Jul. 21st Feb. 14th Jul. 21st Feb. 14th Jul. 21st Feb. 14th Jul. 21st Feb. 14th Jul. 21st Feb. 14th Jul. 21st Business-as-usual interruptible DR scheduling approach 0.02 0.02 10.5 43.3 3.9 15.6 525.0 2165.0 870.8 1997.7 881.3 2041.0 0.04 0.04 22.5 102.1 9.2 41.9 562.5 2552.5 863.2 1916.1 885.7 2018.2 0.06 0.06 44.8 162.8 18.8 70.0 746.7 2713.3 824.3 1882.4 869.1 2045.2 0.08 0.08 81.7 390.4 35.9 171.8 1021.3 4880.0 766.6 1427.4 848.3 1817.8 0.1 0.1 180.9 528.5 83.2 232.5 1809.0 5285.0 601.2 1342.3 782.1 1870.8 0.12 0.12 217.1 634.2 91.8 310.8 1809.0 5285.0 601.2 1342.3 818.3 1976.5 0.14 0.14 264.7 787.8 105.9 409.7 1890.7 5627.1 584.1 1270.5 848.8 2058.3 0.16 0.16 302.5 900.3 115.8 459.2 1890.7 5627.1 584.1 1270.5 886.6 2170.8 0.18 0.18 377.3 1013.7 188.7 547.4 2096.1 5631.7 540.9 1269.5 918.2 2283.2 0.2 0.2 421.5 1341.0 219.2 643.7 2107.5 6705.0 538.5 1044.1 960.0 2385.1 0.22 0.22 486.0 1611.0 233.3 757.2 2209.1 7322.7 517.3 914.4 1003.3 2525.4 0.24 0.24 551.8 2093.1 253.8 879.1 2299.2 8721.3 498.3 620.7 1050.1 2713.8 0.26 0.26 708.6 2593.3 311.8 959.5 2725.4 9974.2 408.9 357.6 1117.5 2950.9 0.28 0.28 1040.2 2906.8 436.9 1133.6 3714.9 10381.4 201.2 272.1 1241.4 3178.9 0.3 0.3 1401.8 3503.2 560.7 1191.1 4672.7 11677.3 0 0 1401.8 3503.2 0.32 0.32 1495.2 3736.7 586.2 1195.7 4672.7 11677.3 0 0 1495.2 3736.7 Proposed market-driven interruptible DR scheduling approach 0.17 0.15 566.3 1327.8 230.7 488.6 3253.8 8155.2 50.8 76.9 617.1 1404.7 Values in bold indicate the total daily operational cost of the MG in the best performance of the BAU interruptible DR management framework. * Given the variability of the best-strategy incentive offered by the MG operator at different peak hours of the day in the proposed market-driven model, the mean daily value of the optimal incentive rate offered by the MG operator (Ip,* MGO) is presented for the proposed model. S. Mohseni et al.
- 22. Applied Energy 287 (2021) 116563 22 framework to enable the DR programme administrator to vary the rate of incentive payment to increase or decrease the supply of DR capacity, either results in an overpayment for access to the DR re sources, or leads to the under-trading of the responsive loads. More specifically, the proposed model has outperformed the BAU model by at least ~21.1% (equating to a saving of $165) and ~22.7% (equating to a saving of $413.1) in terms of the daily operational cost of the MG ( ∑ p∈Pd OCp MG) respectively for the February 14th and July 21st scenarios. 2. The BAU realisation of the interruptible DR programme has failed to exploit the full potential of the demand-side flexibility resources available. The most crucial factor underpinning this under- utilisation of the responsive loads in this model is the lack of inter action between the MG operator and responsive load aggregators, as well as between aggregators and end-consumers to dynamically re- render the incentives for load reduction at different times of the day. This is evident from Table 6, where increasing the MG operator- posted incentive rate from $0.1/kWh to $0.12/kWh, and also from $0.14/kWh to $0.16/kWh, has only led to an increase in the total daily payment to the aggregators despite no increase in the net load reduction in both the scenarios considered. 3. In contrast to the proposed model where the daily operational cost of the system strictly decreases as the MG operator-offered incentive rate increases up to a saturation point, the BAU model’s response to variations in the MG operator-offered incentive rate does not tend to follow a particular pattern. For example, increasing the MG operator- offered incentive rate from $0.02/kWh to $0.04/kWh in the case of July 21st has resulted in a reduction of the daily operational cost of the MG by ~1.1%, then increasing the incentive rate from $0.04/ kWh to $0.06/kWh has increased the daily operational cost of the MG by ~1.3%, and then increasing the incentive rate from $0.06/ kWh to $0.08/kWh has substantially driven down the daily opera tional cost of the MG system–to the globally optimal level. Much of the reason for such an erratic behaviour of the BAU model lies in the fact that the participation of the aggregators depends on meeting certain threshold levels of profits. Put differently, increasing the rate of incentive payments leads to a worthless overpayment unless it triggers the participation of a further MG customer, provided that a lower incentive rate than the per-unit cost of electricity import is deemed sufficient by the customer. However, the interactive DSM market-clearing mechanism embedded in the proposed DSM market model–implemented using the proposed interactive value iteration solution approach (refer to Algorithm 1)–has addressed such a source of unreliability. To evaluate the weather-sensitivity of each model, the analysis is expanded to include all the days in which the interruptible DR pro gramme is executed. Table 7 summarises the descriptive statistics for the DR scheduling variables during the hours of peak demand for which a net energy deficit is predicted. Note the change in temporal resolutions of the dependent variables compared to Table 6. Specifically, the results are presented for the morning peak (MP) and evening peak (EP) hours across the seasons to provide insight into the temporal distribution of utilising the DR resources. The table is revealing in the following ways: 1. The DR events occur more frequently in autumn (734 times) and winter (832 times) than in spring (388 times) and summer (284 times). A comparison of the total number of DR event observations during the morning and evening peak periods across different sea sons offers the following insights: (i) two distinctive daily periods of positive net load demand–the total electric demand on the system minus local generation–can be identified for autumn and winter; while (ii) the net load demand in spring and summer is characterised by one period, namely the MP. This change in the capacity deficit pattern is mainly driven by weather conditions; the warmer months reduce the necessity of utilising electric space heating systems. Other seasonal covariates, including daylight saving and longer daylight hours in spring and summer, which lead to both lower lighting use and higher solar PV generation in the early evening, also contribute to this variation, albeit to a lesser degree. 2. Although the number of DR events that occurred during the MP period is lower than the corresponding EP period in the colder months, the average hourly load reduction procured is nearly the same for the morning and evening peak periods in autumn and winter. This implies that the profile of the net load demand has a shorter, sharper peak in the morning, but a longer, flatter peak in the evening in autumn and winter. This is not only due to the Table 7 Summary statistics for the DR scheduling variables. Variable Spring Summer Autumn Winter MP EP MP EP MP EP MP EP MG operator-offered incentive [$/kWh] Avg. 0.159 0.202 0.140 0.168 0.120 0.097 0.128 0.183 Med. 0.160 0.200 0.140 0.173 0.120 0.097 0.120 0.189 SD 0.031 0.034 0.026 0.027 0.015 0.038 0.039 0.029 Obs. 291 97 208 76 344 390 400 432 Incentive payment to the aggregators [$/h] Avg. 49.004 111.484 26.866 62.378 48.504 63.166 77.043 147.260 Med. 47.409 101.634 26.492 61.087 48.996 64.636 78.349 147.850 SD 11.852 26.799 4.103 5.268 1.665 5.667 6.336 8.320 Obs. 291 97 208 76 344 390 400 432 Incentive payment to the customers [$/h] Avg. 20.092 51.281 12.105 29.448 22.627 28.039 30.510 67.382 Med. 20.115 52.360 10.606 29.282 20.901 27.432 29.660 67.446 SD 5.570 5.954 6.314 2.940 6.307 3.483 4.240 4.252 Obs. 291 97 208 76 344 390 400 432 Load reduction procured by the customers [kWh] Avg. 308.201 551.901 191.900 371.298 604.200 651.196 801.898 804.699 Med. 311.051 553.074 192.312 371.649 603.094 651.628 804.004 805.741 SD 9.814 11.053 5.593 11.579 6.587 11.687 9.678 6.922 Obs. 291 97 208 76 344 390 400 432 Cost of electricity imports [$/h] Avg. 8.611 15.237 3.985 7.907 5.531 8.402 9.004 17.516 Med. 9.044 15.780 4.238 7.974 5.406 7.941 8.172 17.049 SD 3.531 2.020 0.881 0.187 1.087 1.873 2.556 4.937 Obs. 291 97 208 76 344 390 400 432 Total operational cost of the MG [$/h] Avg. 57.615 126.721 30.851 70.285 54.035 71.568 86.047 164.776 Med. 56.453 117.414 30.730 69.061 54.402 72.577 86.521 164.899 SD 2.618 7.981 2.771 3.217 2.410 3.651 4.206 9.325 Obs. 291 97 208 76 344 390 400 432 S. Mohseni et al.
- 23. Applied Energy 287 (2021) 116563 23 coincidence of the residential load with the start of the business day, but also the fact that non-dispatchable renewable power generation from wind and hydro resources is considerably less during the autumn and winter MP period than the corresponding EP period (see Fig. 10). Crucially, the proposed non-cooperative game-theoretic DR scheduling model has yielded reductions in load demand of, on average, ~24% and ~22% respectively during the winter morning and evening peak periods. This equates to an average hourly energy reduction of ~802 kWh in the MP and ~805 kWh in the EP. In summer, this percentage decreases to ~13% (192 kWh) in the MP, and ~15% (371 kWh) in the EP period. 3. Defining the data skewness as (mean–median) / standard deviation, it can be shown that the skewness values of the ‘cost of electricity imports’ and the ‘incentive payments made by the utility to the aggregators’ datasets have opposite signs at all the eight quarterly time intervals. For example, the skewness values of the above mentioned datasets for the winter MP period are 0.326 and − 0.206, respectively. Accordingly, the mean of the former dataset is greater than its median (i.e., the dataset distribution is positively skewed), whereas the mean of the latter dataset is less than its median (i.e., negatively skewed). This suggests that the optimal trade-off between imported power and utilised DR capacity tends to follow an approximately consistent pattern during each period of peak elec tricity use. This finding corroborates the robustness and validity of the proposed non-cooperative game-theoretic DSM approach in producing the best compromise between the imported power and elicited DR capacity. Moreover, Table 8 provides a statistical evaluation of the efficacy of the proposed market-based integration (MBI) of responsive loads using non-cooperative game theory as compared to the BAU model in the four seasons. Note that, for reasons of space, the modelling results are not broken down into the morning and evening peak periods. From Table 8, emerge a number of key statistically valid evidence to support the superiority of the proposed game-theoretic DR scheduling model to the BAU interruptible programmes: 1. The proposed aggregator-mediated, market-based DR programme is able to unlock new sources of economic value that are inaccessible by the BAU-DR scheduling approach. This has resulted in a ~17% (equating to ~$39 k) reduction in the operational cost of the MG in the baseline year. In large part, this is because the proposed model ensures a level playing field for all the DR providers and equitably allocates the benefits of third-party DR aggregation, whilst addi tionally providing a platform for the MG operator, DRAs, and end- consumers to mutually optimise their portfolios and determine the lowest operational costs. 2. A comparison of the seasonal performance of the proposed model and the BAU approach reveals that, on average, the DR resources are under-utilised in autumn and winter, whilst additionally the DR providers are over-compensated in spring and summer in the BAU approach. More specifically, in contrary to the obtained results from the proposed model, where the distributions of the ‘incentive pay ments to the aggregators’ and the ‘cost of electricity imports’ data are oppositely skewed, they have similar skewness patterns in the BAU approach. The BAU approach’s results indicate that both of the above-mentioned distributions are skewed to the left (i.e., most of the observations lie to the right of the mean) in spring and summer, whereas they are both positively-skewed (i.e., most of the observa tions lie to the left of the mean) in autumn and winter. A major explanation for this is the BAU interruptible service approach’s incapability to provide a more-targeted, non-prespecified incentive price signal that fluctuates hourly reflecting changes in the wholesale prices of electricity. 3. While the percentage of incentive payments to the customers to the incentive prices received by the aggregators remains nearly the same across the seasons in the proposed game-theoretic modelling results (within the range of approximately 43–46%), the percentage varies significantly from season to season if the problem is solved in a BAU way. In particular, the BAU modelling results yield the highest utility margin for the customers (with the customers’ share of the utility incentives of ~53%) during the winter months (June to August) when their use of electricity for heating contributes to high network loads. On the other hand, the per-unit profit of the DRAs is largest during the summer months (December to February) when electric Table 8 Comparative statistical analysis of the proposed and BAU-DR scheduling models. Variable Spring Summer Autumn Winter BAU* MBI BAU* MBI BAU* MBI BAU* MBI MG operator-offered incentive [$/kWh] Avg. 0.147 0.170 0.112 0.148 0.051 0.109 0.073 0.155 Med. 0.152 0.159 0.108 0.146 0.045 0.110 0.074 0.156 SD 0.017 0.033 0.017 0.029 0.025 0.031 0.030 0.028 Obs. 388 388 284 284 734 734 832 832 Incentive payment to the aggregators [$/h] Avg. 16.821 64.624 9.725 36.369 13.342 55.595 35.537 111.054 Med. 16.874 55.729 9.777 29.113 12.242 51.287 33.964 87.232 SD 1.670 31.920 0.863 5.482 4.387 8.279 5.125 35.922 Obs. 388 388 284 284 734 734 832 832 Incentive payment to the customers [$/h] Avg. 6.390 27.889 3.112 16.746 6.538 25.245 18.835 48.370 Med. 6.454 23.463 3.223 13.639 6.458 25.131 17.657 36.607 SD 0.574 14.624 0.279 9.624 1.837 5.914 10.386 18.948 Obs. 388 388 284 284 734 734 832 832 Load reduction procured by the customers [kWh] Avg. 104.263 369.126 80.929 239.908 240.608 629.171 465.260 803.352 Med. 104.979 314.339 81.951 194.866 238.388 512.418 463.248 714.444 SD 7.450 106.143 4.920 79.940 14.018 123.846 16.353 101.754 Obs. 388 388 284 284 734 734 832 832 Cost of electricity imports [$/h] Avg. 76.462 10.268 41.742 5.034 61.044 6.919 113.148 13.127 Med. 76.845 10.633 42.059 5.074 31.421 6.948 55.583 13.392 SD 1.902 4.274 2.314 0.148 6.255 2.935 5.206 3.129 Obs. 388 388 284 284 734 734 832 832 Total operational cost of the MG [$/h] Avg. 93.283 74.892 51.467 41.404 74.386 62.515 148.685 124.181 Med. 93.836 58.145 52.657 33.028 72.784 57.739 146.465 94.166 SD 1.766 30.248 3.311 17.753 7.011 9.224 7.104 40.021 Obs. 388 388 284 284 734 734 832 832 * The BAU results represent the business-as-usual model’s best performance out of different daily utility-offered incentives ranging from $0.02/kWh to $0.32/kWh in intervals of $0.02/kWh. S. Mohseni et al.
- 24. Applied Energy 287 (2021) 116563 24 heating cannot provide DR, which reduces the customers’ share of revenues to as low as ~32%. This indicates the BAU approach’s failure to provide a fair division of the utility-offered financial in centives between the DRAs and their corresponding customers, which results in the overall DR underperformance. As these observations are shown to remain valid for the year-round operation of the system, their positive impact on the lifetime cost and cost-effectiveness of the conceptualised system is discussed in the next sub-section. 5.3. Optimal equipment capacity-planning results To evaluate the effectiveness of the proposed DR scheduling frame work in reducing the whole-life cost of MGs, the equipment capacity- planning of the conceptualised MG was carried out under three cases: (1) taking a BAU (static) interruptible load approach to managing the smaller DR resources (as detailed in Section 5.2), (2) using the proposed market-based (dynamic) integration of the aggregator-mediated inter ruptible responsive loads (presented in Section 3), and (3) not imple menting any DSM strategies. Tables 9 and 10 present the MG investment planning model results under the above-mentioned three cases, which are respectively denoted by ‘BAU-DR’, ‘MBI-DR’, and ‘NO-DR’. Specif ically, Table 9 details a breakdown of the optimised cost components included in the life-cycle analysis of the MG system (see Eq. (39)), while Table 10 provides the optimum size of the MG equipment, which are the main decision variables of the optimisation problem. Note that the optimisation model results are adjusted for the value of biomass feed stock. To this end, the total cost associated with the pelletisation of blended biomass feedstocks–agricultural and woody biomass–was considered to be $72/tonne of pellets [123]. The case study site’s nat ural endowment of forest biomass together with its temperate climate that is ideally suited to the agricultural activities, narrows, to a considerable extent, the feedstock supply uncertainty bounds. This provides strong support for taking an exogenous approach to account for the biomass feedstock costs–in the post-optimisation phase. It is also noteworthy that the results reported in the tables represent the best-case performance of the MFOA out of 30 independent trials. Moreover, to demonstrate the adequacy of the maximum number of it erations, and the number of search agents considered, the convergence curves of the MFOA in its best and worst overall performances for each of the above-mentioned cases are shown in Fig. 14. The comparative results presented in Table 9 reveal that the pro posed market-based modelling of the interruptible DSM processes in the planning phase of the conceptualised MG reduces the estimated whole- life cost of the system by at least 21% and up to a maximum of 32% (with an incentive resolution of $0.02/kWh), as compared to the BAU inter ruptible DR-integrated and non-DR-integrated MG planning cases, respectively. Table 9 Breakdown of the total discounted system cost under different DR provision strategies. Cost component Cost subcomponent Simulation case MBI-DR BAU- DR NO-DR Total discounted equipment-related costs (( ∑ c∈CNPCc 20− yr ) + NPCI 20− yr ) [$] 18.25 m 21.88 m 25.62 m Total discounted MG operational costs (NPV( 20− yr ∑8760 t=1 OCMG(t))) Total discounted incentive payment to the aggregators (NPV 20− yr ( ∑8760 t=1 IMGO(t) ∑ j∈JD j LA(t))) [$] 3.99 m 3.48 m − Total discounted cost of electricity imports (NPV 20− yr ( ∑8760 t=1 costim(t))) [$] 0.46 m 2.76 m 7.46 m Total discounted FCEV2G electricity provision costs (NPV( 20− yr ∑8760 t=1 πFCEV2GPFCEV2G(t))) [$] 0.42 m 0.49 m 0.50 m Total discounted operating costs of the biopower plant Total discounted emission credits (NPV 20− yr ( ∑8760 t=1 costem(t))) [$] 0.52 m 0.57 m 0.62 m Total discounted biomass feedstock costs* (NPV 20− yr (72 × ∑8760 t=1 MBP(t))) [$] 0.49 m 0.54 m 0.58 m Total discounted income derived from electricity exports (− NPV 20− yr ( ∑8760 t=1 incomeex(t))) [$] − 2.41 m − 2.42 m − 2.97 m Whole-life cost of the system (WLC) [$] 21.72 m 27.3 m 31.81 m * The total cost of the biomass feedstock is not systematically affected by changes in the endogenous variables of the model in this study. That is, the total cost imposed by the biomass feedstock was calculated outside the optimisation model and the results were then corrected accordingly. Table 10 Size of the MG equipment in the cost-minimal solution under different DR provision strategies. Component Simulation case MBI-DR BAU-DR NO-DR PV plant NPV [no.] 3,594 3,690 4,742 STDEC* [%] 3.54 3.04 3.33 Wind plant NWT [no.] 4 5 6 STDEC* [%] 24.11 26.35 26.73 Micro-hydro power plant NMH [no.] 6 6 6 STDEC* [%] 1.91 1.59 1.36 Biopower plant NBP [no.] 4 4 7 STDEC* [%] 0.77 0.64 0.96 Transformer NT [kVA] 310 320 329 STDEC* [%] 0.11 0.10 0.08 Hydrogen tank NHT [kg] 6,079 7,904 9,168 STDEC* [%] 16.93 18.11 18.16 Electrolyser NE [no.] 122 144 157 STDEC* [%] 4.14 4.08 3.80 Fuel cell NFC [no.] 238 378 440 STDEC* [%] 6.75 8.58 8.66 Battery bank N1600 [no.] 2 2 2 N400 [no.] 0 1 2 N100 [no.] 2 0 3 STDEC* [%] 17.49 15.06 16.00 Super-capacitor bank NSC [no.] 1,982 2,090 2,136 STDEC* [%] 14.53 12.61 11.01 FCEV2G setup NFCEV2G [kW] 504 573 608 STDEC* [%] 0.57 0.53 0.49 Hydrogen station NS [kg-H2/h] 6.14 7.94 9.15 STDEC* [%] 0.42 0.45 0.45 Inverter N900 [no.] 5 6 7 N115 [no.] 2 3 5 N33 [no.] 1 3 1 STDEC* [%] 8.73 8.86 8.97 * STDEC stands for the share of the total discounted equipment-related costs, which can be expressed explicitly in mathematical terms as (( ∑ c∈CNPCc 20− yr ) + NPCI 20− yr ). S. Mohseni et al.
- 25. Applied Energy 287 (2021) 116563 25 Furthermore, the results summarised in Tables 9 and 10 are indica tive of the high efficiency of the proposed model for the aggregator- activated, responsive load-aware MG capacity design in the following ways: 1. While the total discounted equipment-related costs in the BAU case are higher by ~20% than the MBI case, the total discounted income derived from electricity exports has remained at nearly the same level. This is because the majority of this extra cost is spent on the backup power equipment, the energy output of which, according to the MG’s hourly operational strategy in Fig. 5, cannot be sold to the main grid–for energy efficiency considerations. To examine the robustness of this assumption, a further unreported model was run in both the MBI and BAU simulation cases, where the backup power was allowed to be sold into the utility grid, while maintaining the rest of the model unchanged. A comparative analysis of the results of the two models for the investigated test-case is presented in Sup plementary Material (Additional File 4: Table S1). The results show that the relative difference of the total discounted equipment-related costs in the MBI and BAU cases reduces to ~15%, from ~20% for the base-case model, when the sale of backup power into the grid is not prohibited. The key factor underpinning this change is that the un reported optimisation model that supports the sale of backup power to the main grid finds the opportunity to arbitrage intertemporal differences in wholesale prices and buy-back rates. The unreported model, therefore, increases the proportion of total nominal storage to generation capacity in the optimal equipment capacity configuration as compared to the base-case model. More specifically, the propor tion of the share of the total back-up components’ capacity to the share of the total primary generation components’ capacity in the system’s whole-life cost increases from 1.97 and 1.85 to 3.51 and 3.22 in the MBI and BAU model realisations, respectively, at rela tively modest extra total equipment-related costs–that is, ~9% and ~5%, respectively. This, however, increases the MG’s total net in come from the exchange of energy with the utility grid by ~76% and ~429%, respectively, in the aforementioned two cases. As a conse quence, the MG’s whole-life cost reduces by ~3% and ~5%, respectively, in the two cases mentioned above–but at the cost of higher total energy dissipated as a result of increased energy con version rocesses. 2. The total discounted income derived from electricity exports in the non-DR-integrated case is higher by ~23% in comparison with the base case, which is mainly due to the increased excess of renewable energy generation in low-demand periods. Note that the export of energy is seen merely as a means to avoid spillage of non- dispatchable renewable energy, and the low export tariff makes it irrational for the solution algorithm to optimise the capacity of the MG equipment for energy export purposes. It should not be over looked, however, that energy export made a fair contribution to the cost-efficiency of the proposed MG system in all of the cases studied. It is also important to note that the solution algorithm, in the MBI case, has almost always avoided buying and storing electricity from the upstream grid at times of low demand, but the surplus renewable energy is sold to the grid at these times due to: (1) the higher level of feed-in-tariff than the system’s levelised cost of energy (LCOE) at most of the off-peak times of the year, and (2) the fact that the battery and SC banks soon reach their maximum capacity limits when the MG system is lightly loaded. This is while the total discounted cost of electricity imports occupies ~10% and ~24% of the total discounted system costs in the BAU DR-integrated and non-DR-integrated cases, respectively. 3. In all of the investigated cases, the optimised size of the electrolyser unit is considerably lower compared to those in established size combinations–of electrolyser to hydrogen reservoir to fuel cell–in the literature (see, for example, [124,125,126]). This is due to the spe cific conditions of the case study site, where load demand is subject to a high degree of seasonality. Accordingly, an electrolyser of lower capacity is sufficient for the purpose–since the hydrogen tank can be filled gradually during the low season, from October through June. That is also why the optimum capacity of the electrolyser experi enced the least changes among the backup power equipment in the three scenarios investigated. 4. As planned, the fuel cell generation using the stored hydrogen has accounted for seasonal load levelling. The optimal capacity of the fuel cell is more highly impacted by the proposed interruptible DR implementation as compared to the battery and SC banks. This observation implies that peak load shaving–fulfilled by exploiting the responsive loads–has had a substantial role in smoothing out the seasonal variation in load demand, and, in turn, improving the load factor of the annual load power demand profile. In other words, much of the suggested DR scheduling strategy’s positive impact on the cost-efficiency of the conceptualised MG is derived from its implementation in the winter high season. This also explains the marked increase in the size of the WT, hydrogen tank, fuel cell, and the electrical loads’ inverter–as the main drivers of increasing the equipment-related costs–when the DR is implemented in a BAU manner, or, more significantly, when no DR scheduling process is implemented. To provide a clearer picture of the impact of the pro posed DSM model on the load power demand data fed into the optimal capacity planning algorithm, the monthly mean 24-h Fig. 14. Convergence process of the MFOA in its best and worst runs throughout 30 simulation cases. S. Mohseni et al.
- 26. Applied Energy 287 (2021) 116563 26 electricity consumption profile is presented in Fig. 15 for the simu lation cases under study. According to the figure, realising the pro posed DSM model under the BAU and MBI cases shaves ~24% and ~38% off the maximum peak power demand compared to the NO- DR case, respectively. This, consequently, increases the load factor from 0.25 in the NO-DR case to 0.31 and 0.35 in the BAU and MBI simulation cases, respectively. 5. The relatively low share of biomass in the optimised energy resource mix, in spite of its vast potential in the site under study (see Fig. 11), is revealing in two ways: (1) the solution algorithm has succeeded in restricting the bioenergy use to a sustainable level by imposing an emission penalty and, more importantly, (2) it gives credence to the idea that biomass resources need to be deployed in a way that con tributes primarily to energy security–in favour of a deep green approach to renewable energy system planning [127]. More specif ically, the biopower plant in the conceptualised MG plays a critical role in improving the system’s self-sufficiency, as it is the only dis patchable power generation unit in the system. 5.3.1. Financial appraisal To demonstrate the financial sustainability of the long-term invest ment proposal, this sub-section compares the LCOE of the MG with the existing retail electricity prices at the site, as well as the LCOE reported in the literature for the most similar projects. More specifically, the project was benchmarked against the studies in the literature that met the following three criteria: (1) a self-sufficiency ratio of at least 85% if the system is grid-connected, (2) powered by 100% renewable energy, and (3) tailored to the electrification of small- to medium-scale communities. The LCOE, in the MG context, represents the average revenue per unit of energy generated that would be required to recoup the lifetime costs of the system. Accordingly, the LCOE [$/kWh] of the MG under study can be calculated as follows [128]: LCOE = WLC ∑PL n=1 ( ∑8760 t=1 PL(t)+ ∑8760 t=1 PS(t))Δt (1+ir)n , (49) where PL represents the project lifetime [years], ir denotes the real in terest rate per annum [%], and the terms ∑8760 t=1 PL(t) and ∑8760 t=1 PS(t) respectively denote the total annual electric and hydrogen power de mand on the MG, which are discounted to reflect the NPV of future energy flows. By solving Eq. (49), the LCOE of the proposed MG is found to be $0.08/kWh, while the most recent yearly average retail price of elec tricity is as high as $0.22/kWh at the studied site. That is, implementing the proposed MG system is expected to realise savings of at least 64% in the community’s energy costs if financed as a community-owned renewable energy project. Note that the MG’s LCOE is calculated for the case where the aggregator-mediated demand-side flexibility re sources are scheduled using the proposed non-cooperative game-theo retic DSM framework–and the whole-life cost of the MG is $21.72 m. Table 11 benchmarks the conceptualised system in terms of the LCOE with the most similar projects in the literature. As can be seen from Table 11, the LCOE of the simulated MG is highly competitive with that of the best value reported in the recent literature for a community-scale, 100%-renewable electrification project. Add to this the fact that a carbon-free, hydrogen-based, light-duty trans portation fleet is integrated into the proposed MG, making it one of the first of its kind. This provides additional support for the economic sus tainability of the conceptualised community energy system. Based on the above premises, the modelled MG provides an evidence base to inform the energy sector and climate change policy, infrastruc ture providers, and the wider modelling community of both the tech nical feasibility and economic viability of leveraging the potential synergies in the integration of energy networks for electricity, heating, and transport to realise economy-wide deep decarbonisation. 6. Conclusions The projections on the uptake of demand response programmes used in the long-term capital infrastructure planning of sustainable energy systems are substantially influenced by the biases and preferences of end-consumers, which can be modelled in terms of the elasticity of the customer supply of demand response capacity. This study, one of the first to provide an understanding of end-consumer behavioural traits in long-term demand-side management schemes, developed a comfort- aware, demand response-integrated optimisation model for equipment capacity-planning of renewable energy systems. To this end, the study developed a two-stage, aggregator-mediated, non-cooperative game- theoretic demand-side management market design to improve the ac curacy of the long-term forecasts of end-users’ participation in incentive-directed demand response programmes. The proposed model provides an effective framework for improving the accuracy of Fig. 15. Comparison of the monthly mean daily profile for load power demand in different simulation cases. S. Mohseni et al.
- 27. Applied Energy 287 (2021) 116563 27 investment assessments made for demand response-aided energy sys tems by adopting the endogenous Stackelberg leader–follower re lationships in two stages, namely: first, for interactions between the micro-grid operator and responsive load aggregators, and second, for aggregator-customer exchanges. Moreover, the devised model success fully generalised the long-term, community-level renewable energy system design problem in the following four areas: 1. It guaranteed a level playing field for a variety of clean energy technologies–in the interest of energy diversification–where the use of biomass resources is limited to a sustainable level by imposing a new constraint term. 2. It implemented the potential of cross-vector integration (in partic ular, power-to-gas technology) in conjunction with the value of fuel cell electric vehicles in vehicle-to-grid operation to improve the flexibility of energy systems with deep penetration of renewables. 3. It allowed for a meta-heuristic solution algorithm based on the moth- flame optimisation algorithm to find the cost-optimal mix of micro- grid assets, whilst facilitating long-term decision-making on the de livery of aggregator-mediated incentive-responsive loads using a realistic example. The use of a case study illustrated the application of the model in the town of Ohakune, New Zealand, demonstrating that many of the challenges for integrating a 100%-renewable energy system can be surmounted. 4. The suggested solution algorithm was also shown to be efficient in nearing the formulated problem’s globally optimum solution. In addition, a comparative analysis of the proposed market-driven and business-as-usual realisations of the interruptible load programme verified the validity of the proposed modelling framework as a de cision support tool for utilities to make reliable forecasts about the engagement of different classes of end-consumers in demand response programmes. This is particularly important when designing greenfield renewable energy systems, or as micro-grids are used to increase the penetration of responsive loads. The numeric results obtained from the model’s application to the test-case system of Ohakune have revealed two novel insights: 1. The use of the proposed two-stage demand-side management market design for the projection of flexible demand resources brings higher- order information about micro-grid operator-aggregators-customers interactions into the analysis, which can be leveraged towards improving the economic viability of renewable energy systems. Notably, as compared to the case where demand-side resources are managed using a business-as-usual interruptible load approach, the model results have indicated that a cost saving of at least 21% (equating to approximately $5.5 m) can be generated for the simu lated micro-grid in Ohakune, while imposing the same discomfort cost on end-users. 2. The large-scale supply of demand-side flexibility resources, enabled by demand response aggregators, has great potential in reducing the estimated life-cycle cost of sustainable energy systems. Specifically, the evidence from this study demonstrates that assisting the con ceptualised micro-grid with incentive-driven, market-directed demand-side management processes reduces the total discounted system costs by circa 32% (equating to around $10 m in this case study). In this light, a thorough analysis of the value of lost load to the target customers–in the interest of improving the accuracy of the forecasted willingness of the end-users to deliver their demand response resources–is of paramount importance in the design phase of all-renewable micro-grids. This is especially true for the devel opment of first-access energy systems in remote areas where the values of unserved energy are expected to be lower than those esti mated for urban and industrial customers. In conclusion, this paper has shown that capturing the flexible de mand potential of small- to medium-scale customers during the planning phases of a hydrogen-based grid-connected micro-grid system can pave the way toward achieving greater energy independence, energy de mocracy, and energy security in rural and semi-urban areas in a cost- effective and environmentally efficient way. CRediT authorship contribution statement Soheil Mohseni: Conceptualization, Methodology, Data curation, Formal analysis, Investigation, Resources, Software, Validation, Visu alization, Writing - original draft. Alan C. Brent: Supervision, Project administration, Writing - review & editing. Scott Kelly: Supervision, Writing - review & editing. Will N. Browne: Supervision, Writing - re view & editing. Daniel Burmester: Supervision, Writing - review & editing. Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Table 11 Comparative evaluation of the proposed MG’s LCOE against those of the comparable schemes. Reference Renewable energy system architecture Case study site(s) Climatic conditions Unsubsidised LCOE [$/kWh]*, † Hosseinalizadeh et al., 2016 [129] An on-grid PV/WT/BESS/FC MG Four villages in Iran, namely Moaleman, Ghadamgah, Marvdasht, and Nikouyeh Diverse climatic conditions 0.54–1.60 Shang et al., 2016 [130] An insular PV/WT/BESS/DG MG An unnamed island near Singapore Tropical/equatorial 0.14 Chauhan and Saini, 2017 [28] A stand-alone PV/WT/BESS/DG/BP/ MHPP MG Chamoli district, Uttarakhand state, India Warm temperate 0.07–0.10 Fu et al., 2018 [131] Stand-alone solar PV systems U.S.-wide Diverse climatic conditions 0.13–0.16 Li, 2019 [132] A grid-independent PV/BESS/FC MG A community centre in Kunming, China Humid subtropical 1.55 Rezk et al., 2019 [133] A grid-independent PV/FC hybrid renewable energy system The city of Minya, Egypt Mediterranean 0.06 This study A grid-tied PV/WT/MHPP/BP/FC/ BESS/SC MG The town of Ohakune, New Zealand Temperate 0.08 Key: BESS = Battery Energy Storage System, BP = Biopower Plant, DG = Diesel Generator, FC = Fuel Cell, LCOE = Levelised Cost of Energy, MG = Micro-Grid, MHPP = Micro-Hydro Power Plant, PV = Photovoltaic, SC = Super-Capacitor, WT = Wind turbine. † Where appropriate, the LCOE values were adjusted to 2019 U.S. dollars. * For cases where different configurations of the proposed system are investigated, or the conceptualised system is optimised under different climatic conditions, or the optimisation process is carried out in a multi-objective search space or in a stochastic way, the value of LCOE is reported as a range, rather than a certain value. S. Mohseni et al.
- 28. Applied Energy 287 (2021) 116563 28 Acknowledgements The authors would like to acknowledge excellent comments and suggestions on the categorisation of load demand from the audience of the 2020 17th International Conference on the European Energy Market (EEM) in Stockholm, Sweden, as well as constructive comments from the anonymous reviewers. Appendix A. Supplementary material Supplementary data to this article can be found online at https://doi. org/10.1016/j.apenergy.2021.116563. References [1] Hui H, Ding Y, Shi Q, Li F, Song Y, Yan J. 5G network-based Internet of Things for demand response in smart grid: a survey on application potential. Appl Energy 2020;257:113972. https://doi.org/10.1016/j.apenergy.2019.113972. [2] Mohseni S, Moghaddas-Tafreshi SM. A multi-agent system for optimal sizing of a cooperative self-sustainable multi-carrier microgrid. Sustain Cities Soc 2018;38: 452–65. https://doi.org/10.1016/j.scs.2018.01.016. [3] Niu J, Tian Z, Zhu J, Yue L. Implementation of a price-driven demand response in a distributed energy system with multi-energy flexibility measures. Energy Convers Manag 2020;208:112575. https://doi.org/10.1016/j. enconman.2020.112575. [4] Ihsan A, Jeppesen M, Brear MJ. Impact of demand response on the optimal, techno-economic performance of a hybrid, renewable energy power plant. Appl Energy 2019;238:972–84. https://doi.org/10.1016/j.apenergy.2019.01.090. [5] Modarresi Ghazvini A, Olamaei J. Optimal sizing of autonomous hybrid PV system with considerations for V2G parking lot as controllable load based on a heuristic optimization algorithm. Sol Energy 2019;184:30–9. https://doi.org/ 10.1016/j.solener.2019.03.087. [6] Jing Z, Zhu J, Hu R. Sizing optimization for island microgrid with pumped storage system considering demand response. J Mod Power Syst Clean Energy 2018;6(4):791–801. https://doi.org/10.1007/s40565-017-0349-1. [7] Kim H, Bae J, Baek S, Nam D, Cho H, Chang HJ. Comparative analysis between the Government Micro-Grid Plan and computer simulation results Based on real data: the practical case for a South Korean Island. Sustainability 2017;9(2):197. https://doi.org/10.3390/su9020197. [8] Amir V, Azimian M. Dynamic multi-carrier microgrid deployment under uncertainty. Appl Energy 2020;260:114293. https://doi.org/10.1016/j. apenergy.2019.114293. [9] Gelazanskas L, Gamage KAA. Demand side management in smart grid: a review and proposals for future direction. Sustain Cities Soc 2014;11:22–30. https://doi. org/10.1016/j.scs.2013.11.001. [10] Haider HT, See OH, Elmenreich W. A review of residential demand response of smart grid. Renew Sustain Energy Rev 2016;59:166–78. https://doi.org/ 10.1016/j.rser.2016.01.016. [11] Esther BP, Kumar KS. A survey on residential Demand Side Management architecture, approaches, optimization models and methods. Renew Sustain Energy Rev 2016;59:342–51. https://doi.org/10.1016/j.rser.2015.12.282. [12] Wang J, Zhong H, Ma Z, Xia Q, Kang C. Review and prospect of integrated demand response in the multi-energy system. Appl Energy 2017;202:772–82. https://doi.org/10.1016/j.apenergy.2017.05.150. [13] Robert FC, Sisodia GS, Gopalan S. A critical review on the utilization of storage and demand response for the implementation of renewable energy microgrids. Sustain Cities Soc 2018;40:735–45. https://doi.org/10.1016/j.scs.2018.04.008. [14] Jordehi AR. Optimisation of demand response in electric power systems, a review. Renew Sustain Energy Rev 2019;103:308–19. https://doi.org/10.1016/j. rser.2018.12.054. [15] Behrangrad M. A review of demand side management business models in the electricity market. Renew Sustain Energy Rev 2015;47:270–83. https://doi.org/ 10.1016/j.rser.2015.03.033. [16] Jin X, Wu Q, Jia H. Local flexibility markets: literature review on concepts, models and clearing methods. Appl Energy 2020;261:114384. https://doi.org/ 10.1016/j.apenergy.2019.114387. [17] Kahrobaee S, Asgarpoor S, Qiao W. Optimum sizing of distributed generation and storage capacity in smart households. IEEE Trans Smart Grid 2013;4(4): 1791–801. https://doi.org/10.1109/TSG.2013.2278783. [18] Yu L, Li YP, Huang GH, An CJ. A robust flexible-probabilistic programming method for planning municipal energy system with considering peak-electricity price and electric vehicle. Energy Convers Manag 2017;137:97–112. https://doi. org/10.1016/j.enconman.2017.01.028. [19] Varasteh F, Nazar MS, Heidari A, Shafie-khah M, Catalão JPS. Distributed energy resource and network expansion planning of a CCHP based active microgrid considering demand response programs. Energy 2019;172:79–105. https://doi. org/10.1016/j.energy.2019.01.015. [20] Cardoso G, Stadler M, Bozchalui MC, Sharma R, Marnay C, Barbosa-Póvoa A, et al. Optimal investment and scheduling of distributed energy resources with uncertainty in electric vehicle driving schedules. Energy 2014;64:17–30. https:// doi.org/10.1016/j.energy.2013.10.092. [21] Hosseinnia H, Nazarpour D, Talavat V. Multi-objective optimization framework for optimal planning of the microgrid (MG) under employing demand response program (DRP). J Ambient Intell Humaniz Comput 2019;10(7):2709–30. https:// doi.org/10.1007/s12652-018-0977-y. [22] Moghaddas-Tafreshi SM, Jafari M, Mohseni S, Kelly S. Optimal operation of an energy hub considering the uncertainty associated with the power consumption of plug-in hybrid electric vehicles using information gap decision theory. Int J Electr Power Energy Syst 2019;112:92–108. https://doi.org/10.1016/j. ijepes.2019.04.040. [23] Martins VF, Borges CLT. Active distribution network integrated planning incorporating distributed generation and load response uncertainties. IEEE Trans Power Syst 2011;26(4):2164–72. https://doi.org/10.1109/ TPWRS.2011.2122347. [24] Zhu L, Yan Z, Lee WJ, Yang X, Fu Y, Cao W. Direct load control in microgrids to enhance the performance of integrated resources planning. IEEE Trans Ind Appl 2015;51(5):3553–60. https://doi.org/10.1109/TIA.2015.2413960. [25] Atia R, Yamada N. Sizing and analysis of renewable energy and battery systems in residential microgrids. IEEE Trans Smart Grid 2016;7(3):1204–13. https://doi. org/10.1109/TSG.2016.2519541. [26] Pazouki S, Haghifam MR. Optimal planning and scheduling of energy hub in presence of wind, storage and demand response under uncertainty. Int J Electr Power Energy Syst 2016;80:219–39. https://doi.org/10.1016/j. ijepes.2016.01.044. [27] Schachter JA, Mancarella P, Moriarty J, Shaw R. Flexible investment under uncertainty in smart distribution networks with demand side response: assessment framework and practical implementation. Energy Policy 2016;97: 439–49. https://doi.org/10.1016/j.enpol.2016.07.038. [28] Chauhan A, Saini RP. Size optimization and demand response of a stand-alone integrated renewable energy system. Energy 2017;124:59–73. https://doi.org/ 10.1016/j.energy.2017.02.049. [29] Nojavan S, Majidi M, Esfetanaj NN. An efficient cost-reliability optimization model for optimal siting and sizing of energy storage system in a microgrid in the presence of responsible load management. Energy 2017;139:89–97. https://doi. org/10.1016/j.energy.2017.07.148. [30] Amrollahi MH, Bathaee SMT. Techno-economic optimization of hybrid photovoltaic/wind generation together with energy storage system in a stand- alone micro-grid subjected to demand response. Appl Energy 2017;202:66–77. https://doi.org/10.1016/j.apenergy.2017.05.116. [31] Chen J, Zhang W, Li J, Zhang W, Liu Y, Zhao B, et al. Optimal sizing for grid-tied microgrids with consideration of joint optimization of planning and operation. IEEE Trans Sustain Energy 2018;9(1):237–48. https://doi.org/10.1109/ TSTE.2017.2724583. [32] Zheng Y, Jenkins BM, Kornbluth K, Kendall A, Træholt C. Optimization of a biomass-integrated renewable energy microgrid with demand side management under uncertainty. Appl Energy 2018;230:836–44. https://doi.org/10.1016/j. apenergy.2018.09.015. [33] Xiao H, Pei W, Dong Z, Kong L. Bi-level planning for integrated energy systems incorporating demand response and energy storage under uncertain environments using novel metamodel. CSEE J Power Energy Syst 2018;4(2): 155–67. https://doi.org/10.17775/CSEEJPES.2017.01260. [34] Husein M, Chung I-Y. Optimal design and financial feasibility of a university campus microgrid considering renewable energy incentives. Appl Energy 2018; 225:273–89. https://doi.org/10.1016/j.apenergy.2018.05.036. [35] Gazijahani FS, Salehi J. Reliability constrained two-stage optimization of multiple renewable-based microgrids incorporating critical energy peak pricing demand response program using robust optimization approach. Energy 2018;161: 999–1015. https://doi.org/10.1016/j.energy.2018.07.191. [36] Amir V, Jadid S, Ehsan M. Optimal planning of a multi-carrier microgrid (MCMG) considering demand-side management. Int J Renew Energy Res 2018;8(1): 238–49. [37] Mohseni S, Brent A, Burmester D, Chatterjee A. Optimal sizing of an islanded micro-grid using meta-heuristic optimization algorithms considering demand- side management. In: Proceedings of the 2018 Australas. Univ. Power Eng. Conf., Auckland, New Zealand, 27–30 Nov. 2018, p. 1–6. https://doi.org/10.1109 /AUPEC.2018.8757882. [38] Nazari A, Keypour R. A two-stage stochastic model for energy storage planning in a microgrid incorporating bilateral contracts and demand response program. J Energy Storage 2019;21:281–94. https://doi.org/10.1016/j.est.2018.12.002. [39] Prathapaneni DR, Detroja KP. An integrated framework for optimal planning and operation schedule of microgrid under uncertainty. Sustain Energy, Grids Networks 2019;19:100232. https://doi.org/10.1016/j.segan.2019.100232. [40] Bhamidi L, Sivasubramani S. Optimal planning and operational strategy of a residential microgrid with demand side management. IEEE Syst J, Early Access 2019:1–9. https://doi.org/10.1109/JSYST.2019.2918410. [41] Mohseni S, Brent AC, Burmester D. A demand response-centred approach to the long-term equipment capacity planning of grid-independent micro-grids optimized by the moth-flame optimization algorithm. Energy Convers Manag 2019;200:112105. https://doi.org/10.1016/j.enconman.2019.112105. [42] Salyani P, Abapour M, Zare K, Babri T. Optimal stochastic planning of DERs in a game theory framework considering demand response and pollution issues. In: Demand Response Application in Smart Grids: Concepts and Planning Issues, Volume 1. Cham: Springer International Publishing; 2020. p. 193–214. https:// doi.org/10.1007/978-3-030-31399-9_8. [43] Burger S, Chaves-Avila JP, Batlle C, Perez-Arriaga IJ. The Value of Aggregators in Electricity Systems. CEERP WP 2016-001. MIT Center for Energy and Environmental Policy Research; 2016. Available: https://energy.mit.edu/ S. Mohseni et al.
- 29. Applied Energy 287 (2021) 116563 29 wp-content/uploads/2016/01/CEEPR_WP_2016-001.pdf [Accessed: 18-Oct.- 2020]. [44] Stevenson T, Batstone S, Reeve D, Poynton M, Comendant C. Transitioning to zero net emissions by 2050: moving to a very low-emissions electricity system in New Zealand. SAPERE Research Group – Prepared for the New Zealand Productivity Commission; 2018. Available: https://www.productivity.govt.nz/assets/Docum ents/3374eca8c4/Transitioning-to-zero-net-emissions-by-2050.pdf [Accessed: 18-Oct.2020]. [45] Okur Ö, Voulis N, Heijnen P, Lukszo Z. Aggregator-mediated demand response: minimizing imbalances caused by uncertainty of solar generation. Appl Energy 2019;247:426–37. https://doi.org/10.1016/j.apenergy.2019.04.035. [46] Nash J. Non-cooperative games. Ann Math 1951;54(2):286–95. https://doi.org/ 10.2307/1969529. [47] Basar T, Olsder GJ. Dynamic noncooperative game theory, vol. 23. Siam; 1999. [48] Canadian Solar Inc. CS6K-270|275|280P; Dec. 2017, PV Module Product Datasheet V5.552_EN. Available: https://www.collectiu-solar.cat/pdf/2-Panel-Ca nadian_Solar-Datasheet-CS6K.pdf [Accessed: 4-Mar.-2020]. [49] Patel MR. Wind and solar power systems: design, analysis, and operation. CRC Press; 2005. [50] Murugaperumal K, Raj PADV. Feasibility design and techno-economic analysis of hybrid renewable energy system for rural electrification. Sol Energy 2019;188: 1068–83. https://doi.org/10.1016/j.solener.2019.07.008. [51] Meteonorm. Bern, Switzerland: Genossenschaft Meteotest; 2018. Version 7.3.0. [52] Alstom. ecotècnia ECO 48/750; Aug. 2014, ecotècnia ECO 48/750 Datasheet. [Online]. Available: https://en.wind-turbine-models. com/turbines/791-ecot-cnia-eco-48-750#datasheet/ [Accessed: 4-Mar.-2020]. [53] Wagner R, Antoniou I, Pedersen SM, Courtney MS, Jørgensen HE. The influence of the wind speed profile on wind turbine performance measurements. Wind Energy Int J Prog Appl Wind Power Convers Technol 2009;12(4):348–62. [54] Gipe P. Wind power. Wind Eng 2004;28(5):629–31. https://doi.org/10.1260/ 0309524043028145. [55] Sunceo Hydro. XJ50-100SCTF6-Z; 2015. [Online]. Available: https://www.micro -hydro-power.com/100kw-hydro-turbine-generator/ [Accessed: 4-Mar.-2020]. [56] Khan MRB, Jidin R, Pasupuleti J. Multi-agent based distributed control architecture for microgrid energy management and optimization. Energy Conserv Manag 2016;112:288–307. https://doi.org/10.1016/j.enconman.2016.01.011. [57] Bosona T, Gebresenbet G. Modeling hydropower plant system to improve its reservoir operation. Int J Water Resour Environ Eng 2 (4):87–94. [Online]. Available: https://academicjournals.org/journal/IJWREE/article-full-text-pdf/34 24C7E1713 [Accessed: 19-Dec.-2020]. [58] All Power Labs. Old PP20 vs New PP30 Cogen-CS: Technical Specifications & Comparisons; Jan. 2019, Rev 03. Available: http://www.allpowerlabs.com/wp- content/uploads/2019/07/PP30-vs-PP20-Spec-Sheet-Public-2019-Rev-03-cu rrent-July-2019.pdf [Accessed: 4-Mar.-2020]. [59] Singh S, Singh M, Kaushik SC. Feasibility study of an islanded microgrid in rural area consisting of PV, wind, biomass and battery energy storage system. Energy Convers Manag 2016;128:178–90. https://doi.org/10.1016/j. enconman.2016.09.046. [60] Ahmed OY, Ries MJ, Northrop WF. Emissions factors from distributed, small-scale biomass gasification power generation: Comparison to open burning and large- scale biomass power generation. Atmos Environ 2019;200:221–7. https://doi. org/10.1016/j.atmosenv.2018.12.024. [61] Moghaddas-Tafreshi SM, Mohseni S, Karami ME, Kelly S. Optimal energy management of a grid-connected multiple energy carrier micro-grid. Appl Therm Eng 2019;152:796–806. https://doi.org/10.1016/j.applthermaleng.2019.02.113. [62] Qin H, Kimball JW. Solid-state transformer architecture using AC-AC dual-active- bridge converter. IEEE Trans Ind Electron 2013;60(9):3720–30. https://doi.org/ 10.1109/TIE.2012.2204710. [63] Leonics Co. Apollo GTP-500; 2018, P.LEN.BRO.INV.156 Rev. 12.00. Available: http://www.leonics.com/product/renewable/inverter/dl/GTP-500-156.pdf [Accessed: 4-Mar.-2020]. [64] Akram U, Khalid M, Shafiq S. An innovative hybrid wind-solar and battery- supercapacitor microgrid system—development and optimization. IEEE Access 2017;5:25897–912. https://doi.org/10.1109/ACCESS.2017.2767618. [65] Eaton Corporation. XLR-48 Supercapacitor; Feb. 2019, Technical Data 10510. Available: https://datasheet.octopart.com/XLR-48R6167-R-Eaton-datasheet- 130052459.pdf [Accessed: 4-Mar.-2020]. [66] CellCube Energy Storage Systems Inc. Use your own power grid: Intelligent storage systems based on vanadium redox flow technology; 2018, D6144/ 0714ND3. Available: https://static1.squarespace.com/static/5b1198ada2772c6 585959926/t/5b57363f88251b71261fc4a1/1532442177499/CellCube+-+Use+ Your+Own+Power+Grid.pdf [Accessed: 4-Mar.-2020]. [67] H-TEC Systems. The core of electrolysis: PEM electrolyser stacks – Designed for ideal integration in systems; 2019, H-TEC Series-S: S 30/30. Available: https ://www.h-tec-systems.com/fileadmin/Content/PDFs/19022019/H-TEC_SYS TEMS_Datenblatt_Stacks_SE30_30_EN.PDF [Accessed: 4-Mar.-2020]. [68] Ballard Power Systems Inc. FCgen–1020ACS; May 2015, SPC5101559-0H. Available: https://www.ballard.com/docs/default-source/spec-sheets/fcgen-102 0-acs-v2.pdf?sfvrsn=c3ebc380_4 [Accessed: 19-Dec.-2020]. [69] Graham JDT, Mulvenna AJ, Mufford WE, Borck JG, Ko J, Harper MAM. Hydrogen fueling station. U.S. Patent 6,810,925, issued 2 Nov. 2004. [70] Pure Energy Centre Limited. Hydrogen fueling station; 2019. [Online]. Available: https://pureenergycentre.com/hydrogen-fueling-station/ [Accessed: 4-Mar.- 2020]. [71] Takagi H. Queueing analysis: a foundation of performance evaluation. Amsterdam: North-Holland; 1991. [72] Robledo CB, Oldenbroek V, Abbruzzese F, van Wijk AJM. Integrating a hydrogen fuel cell electric vehicle with vehicle-to-grid technology, photovoltaic power and a residential building. Appl Energy 2018;215:615–29. https://doi.org/10.1016/j. apenergy.2018.02.038. [73] Ahmad MS, Mehmood MA, Luo H, Shen B, Latif M, Ghani WAWAK, et al. Pyrolysis and thermogravimetric study to elucidate the bioenergy potential of novel feedstock produced on poor soils while keeping the environmental sustainability intact. Sustain 2019;11(13). https://doi.org/10.3390/su11133592. [74] Winter C-J, Sizmann RL, Vant-Hull LL, editors. Solar power plants: fundamentals, technology, systems, economics. Springer Science & Business Media; 2012. [75] My Solar Quotes. Solar Power Buy-Back Rates; 2020. [Online]. Available: https ://www.mysolarquotes.co.nz/about-solar-power/residential/solar-power-buy-ba ck-rates-nz/ [Accessed: 4-Mar.-2020]. [76] [dataset] CliFlo: New Zealand’s National Climate Database. [Online]. Available: http://cliflo.niwa.co.nz/ [Retrieved: 10-Oct.-2019]. [77] Elert G. Energy density of hydrogen. Energy Density of Hydrogen− The Physics Factbook 2017;22. [78] XiaoW, Dunford WG, Capel A. A novel modeling method for photovoltaic cells. In: Proceedings of the 2004 IEEE 35th Annual Power Electronics Specialists Conference (IEEE Cat. No.04CH37551), Aachen, Germany, 20–25 Jun. 2004, pp. 1950–56. https://doi.org/10.1109/PESC.2004.1355416. [79] Engineering ToolBox. Wind Shear; 2008. [Online]. Available: https://www. engineeringtoolbox.com/wind-shear-d_1215.html/ [Accessed: 4-Mar.-2020]. [80] Interagency Working Group on Social Cost of Greenhouse Gases, United States Government. Technical support document: technical update of the social cost of carbon for regulatory impact analysis under executive order 12866. The August 2016 revision of the 2013 Technical Support Document on the Social Cost of Carbon. Available: https://www.epa.gov/sites/production/files/2016-12/do cuments/sc_co2_tsd_august_2016.pdf [Accessed: 4-Mar.-2020]. [81] Nerini FF, Broad O, Mentis D, Welsch M, Bazilian M, Howells M. A cost comparison of technology approaches for improving access to electricity services. Energy 2016;95:255–65. https://doi.org/10.1016/j.energy.2015.11.068. [82] Thomas D, Deblecker O, Ioakimidis CS. Optimal design and techno-economic analysis of an autonomous small isolated microgrid aiming at high RES penetration. Energy 2016;116:364–79. https://doi.org/10.1016/j. energy.2016.09.119. [83] Duman AC, Güler Ö. Techno-economic analysis of off-grid PV/wind/fuel cell hybrid system combinations with a comparison of regularly and seasonally occupied households. Sustain Cities Soc 2018;42:107–26. https://doi.org/ 10.1016/j.scs.2018.06.029. [84] Mohseni S, Brent AC. Economic viability assessment of sustainable hydrogen production, storage, and utilisation technologies integrated into on- and off-grid micro-grids: a performance comparison of different meta-heuristics. Int J Hydrogen Energy 2019. https://doi.org/10.1016/j.ijhydene.2019.11.079. [85] Oldenbroek V, Smink G, Salet T, van Wijk AJM. Fuel cell electric vehicle as a power plant: Techno-economic scenario analysis of a renewable integrated transportation and energy system for smart cities in two climates. Appl Sci 2020; 10(1):143. https://doi.org/10.3390/app10010143. [86] Oldenbroek V, Hamoen V, Alva S, Robledo CB, Verhoef LA, van Wijk AJM. Fuel cell electric vehicle-to-grid: experimental feasibility and operational performance as balancing power plant. Fuel Cells 2018;18(5):649–62. https://doi.org/ 10.1002/fuce.201700192. [87] Farahani SS, van der Veen R, Oldenbroek V, Alavi F, Lee EHP, van de Wouw N, et al. A hydrogen-based integrated energy and transport system: the design and analysis of the car as power plant concept. IEEE Trans Syst Man Cybern Mag 2019;5(1):37–50. https://doi.org/10.1109/MSMC.2018.2873408. [88] Steward DM. Critical Elements of Vehicle-to-Grid (V2G) Economics. No. NREL/ TP-5400-69017, Golden, CO, USA: National Renewable Energy Lab (NREL); 2017. [89] Soshinskaya M, Crijns-Graus WHJ, van der Meer J, Guerrero JM. Application of a microgrid with renewables for a water treatment plant. Appl Energy 2014;134: 20–34. https://doi.org/10.1016/j.apenergy.2014.07.097. [90] Khan MRB, Jidin R, Pasupuleti J, Shaaya SA. Optimal combinations of PV, wind, micro-hydro and diesel systems for a seasonal load demand. In: Proceedings of the 2014 IEEE International Conference on Power and Energy (PECon), Kuching, Malaysia, 1–3 Dec. 2014, pp. 171–76. https://doi.org/10.1109/PECON.201 4.7062435. [91] Chauhan A, Saini RP. Techno-economic optimization based approach for energy management of a stand-alone integrated renewable energy system for remote areas of India. Energy 2016;94:138–56. https://doi.org/10.1016/j. energy.2015.10.136. [92] Naderi M, Bahramara S, Khayat Y, Bevrani H. Optimal planning in a developing industrial microgrid with sensitive loads. Energy Rep 2017;3:124–34. https://doi. org/10.1016/j.egyr.2017.08.004. [93] Mohseni S, Moghaddas-Tafreshi SM. Development of a multi-agent system for optimal sizing of a commercial complex microgrid. arXiv preprint, arXiv: 1811.12553; 2018. [94] Mohseni S, Moghaddas-Tafreshi SM. A multi-agent approach to optimal sizing of a combined heating and power microgrid. arXiv preprint, arXiv:1812.11076; 2018. [95] Mohseni S, Brent AC, Burmester D. Community resilience-oriented optimal micro- grid capacity expansion planning: The case of Totarabank eco-village, New Zealand. Energies 13 (15):3970. https://doi.org/10.3390/en13153970. [96] Mohseni S, Brent AC, Kelly S. A hierarchical, market-based, non-cooperative game-theoretic approach to projecting flexible demand-side resources: towards more realistic demand response-integrated, long-term energy planning models. S. Mohseni et al.
- 30. Applied Energy 287 (2021) 116563 30 In: Proceedings of the 2020 17th International Conference on the European Energy Market (EEM), Stockholm, Sweden, 16–18 Sep. 2020, pp. 1–6. https:// doi.org/10.1109/EEM49802.2020.9221977. [97] Yu M, Hong SH. Incentive-based demand response considering hierarchical electricity market: a Stackelberg game approach. Appl Energy 2017;203:267–79. https://doi.org/10.1016/j.apenergy.2017.06.010. [98] Fahrioglu M, Alvarado L. Designing incentive compatible contracts for effective demand management. IEEE Trans Power Syst 2000;15(4):1255–60. https://doi. org/10.1109/59.898098. [99] Fahrioglu M, Alvarado L. Using utility information to calibrate customer demand management behavior models. IEEE Trans Power Syst 2001;16(2):317–22. https://doi.org/10.1109/59.918305. [100] Mirjalili S. Moth-flame optimization algorithm: a novel nature-inspired heuristic paradigm. Knowledge-Based Syst 2015;89:228–49. https://doi.org/10.1016/j. knosys.2015.07.006. [101] Goldberg DE, Holland JH. Genetic algorithms and machine learning. Mach Learn 1988;3:95–9. https://doi.org/10.1023/A:1022602019183. [102] Kennedy J, Eberhart R. Particle swarm optimization. In: Proceedings of the 1995 International Conference on Neural Networks, Perth, WA, Australia, 27 Nov.–1 Dec. 1995, p. 1942–48. https://doi.org/10.1109/ICNN.1995.488968. [103] Mohseni S, Brent AC, Burmester D. A comparison of metaheuristics for the optimal capacity planning of an isolated, battery-less, hydrogen-based micro-grid. Appl Energy 2020;259:114224. https://doi.org/10.1016/j. apenergy.2019.114224. [104] Mohseni S, Brent AC, Burmester D. A sustainable energy investment planning model based on the micro-grid concept using recent metaheuristic optimization algorithms. In: Proceedings of the 2019 IEEE Congress on Evolutionary Computation (CEC), Wellington, New Zealand, 10–13 Jun. 2019, pp. 219–26. https://doi.org/10.1109/CEC.2019.8790007. [105] Mohseni S, Brent AC, Burmester D, Chatterjee A. Stochastic optimal sizing of micro-grids using the moth-flame optimization algorithm. In: Proceedings of the 2019 IEEE Power & Energy Society General Meeting (PESGM), Atlanta, GA, USA, 4–8 Aug. 2019, pp. 1–5. https://doi.org/10.1109/PESGM40551.2019.8973570. [106] Trading Economics. New Zealand − Real Interest Rate. [Online]. Available: htt ps://tradingeconomics.com/new-zealand/real-interest-rate-percent-wb-data.ht ml/ [Retrieved: 5-Feb.-2020]. [107] Hübert T, Boon-Brett L, Buttner W. Sensors for safety and process control in hydrogen technologies. CRC Press; 2016. [108] Chen Q, Xia M, Zhou Y, Cai H, Wu J, Zhang H. Optimal planning for partially self- sufficient microgrid with limited annual electricity exchange with distribution grid. IEEE Access 2019;7:123505–20. https://doi.org/10.1109/ ACCESS.2019.2936762. [109] Chowdhury S, Zhang J, Messac A. Avoiding premature convergence in a mixed- discrete particle swarm optimization (MDPSO) algorithm. In: Proceedings of the 53rd AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and Materials Conference, Honolulu, Hawaii, USA, 23–26 Apr. 2012, p. 1678. https://doi.org/ 10.2514/6.2012-1678. [110] Khan B, Singh P. Selecting a meta-heuristic technique for smart micro-grid optimization problem: a comprehensive analysis. IEEE Access 2017;5:13951–77. https://doi.org/10.1109/ACCESS.2017.2728683. [111] MATLAB. Natick, MA: The MathWorks Inc; 2018. R2018b, Version 9.5. [112] Anonymous. Ohakune Useful Information; 2011. [Online]. Available: http:// www.visitohakune.co.nz/page/ohakune-useful-information/12/ [Accessed: 4- Mar.-2020]. [113] Environmental Protection Authority. Summary and Analysis – Carrots and Parsnips. No. APP201045, 2012. Available: https://www.epa.govt.nz/assets/F ileAPI/hsno-ar/APP201045/fb604d3064/APP201045-Summary-and-Ana lysis-Carrot-and-Parsnip.pdf [Accessed: 4-Mar.-2020]. [114] Hall P, Jack M. Bioenergy options for New Zealand − Analysis of large-scale bioenergy from forestry. No. 1124-2019-3124, Scion − Next Generation Biomaterials, 2009. Available: https://niwa.co.nz/sites/niwa.co.nz/files/impor ted/__data/assets/pdf_file/0007/95668/Large-scale-forestry-for-bioenergy.pdf [Accessed: 4-Mar.-2020]. [115] [dataset] Anderson B, Eyers D, Ford R, Ocampo DG, Peniamina R, Stephenson J, et al. New Zealand GREEN Grid Household Electricity Demand Study 2014–2018. Colchester, Essex: UK Data Service. https://doi.org/10.5255/UKDA-SN-853334. [116] Anderson J. Pulling the plug on network congestion. [B.A. dissertation]. Dunedin, New Zealand: Department of Economics, University of Otago; 2009. [117] Tayan O, Alginahi YM, Kabir MN, Al Binali AM. Analysis of a transportation system with correlated network intersections: a case study for a central urban city with high seasonal fluctuation trends. IEEE Access 2017;5:7619–35. https://doi. org/10.1109/ACCESS.2017.2695159. [118] The Electricity Market Information: The New Zealand Electricity Authority’s wholesale database. [Online]. Available: https://www.emi.ea.govt.nz/Wholesale /Reports/ [Retrieved: 12-Feb.-2020]. [119] Concept Economics Pty Ltd. Investigation of the value of unserved energy − Stage 1. No. ABN 73 129 990 530, Prepared for Electricity Commission; 2008. Available: https://www.researchgate.net/profile/Deb_Chattopadhyay/publicati on/322255908_Investigation_of_the_Value_of_Unserved_Energy/links/5a4e8 ff5458515e71b085a3f/Investigation-of-the-Value-of-Unserved-Energy.pdf [Accessed: 19-Dec.-2020]. [120] Xi X, Sioshansi R. Using price-based signals to control plug-in electric vehicle fleet charging. IEEE Trans Smart Grid 2014;5(3):1451–64. https://doi.org/10.1109/ TSG.2014.2301931. [121] Nwulu NI, Xia X. Multi-objective dynamic economic emission dispatch of electric power generation integrated with game theory based demand response programs. Energy Convers Manag 2015;89:963–74. https://doi.org/10.1016/j. enconman.2014.11.001. [122] Nwulu NI, Xia X. Implementing a model predictive control strategy on the dynamic economic emission dispatch problem with game theory based demand response programs. Energy 2015;91:404–19. https://doi.org/10.1016/j. energy.2015.08.042. [123] Mani S, Sokhansanj S, Bi X, Turhollow A. Economics of producing fuel pellets from biomass. Appl Eng Agric 2006;22(3):421–6. https://doi.org/10.13031/2 013.20447. [124] Baghaee HR, Mirsalim M, Gharehpetian GB. Multi-objective optimal power management and sizing of a reliable wind/PV microgrid with hydrogen energy storage using MOPSO. J Intell Fuzzy Syst 2017;32(3):1753–73. https://doi.org/ 10.3233/JIFS-152372. [125] Li B, Roche R, Paire D, Miraoui A. Sizing of a stand-alone microgrid considering electric power, cooling/heating, hydrogen loads and hydrogen storage degradation. Appl Energy 2017;205:1244–59. https://doi.org/10.1016/j. apenergy.2017.08.142. [126] Mukherjee U, Maroufmashat A, Ranisau J, Barbouti M, Trainor A, Juthani N, et al. Techno-economic, environmental, and safety assessment of hydrogen powered community microgrids; case study in Canada. Int J Hydrogen Energy 2017;42 (20):14333–49. https://doi.org/10.1016/j.ijhydene.2017.03.083. [127] Kampman B, Bergsma G, Schepers B, Croezen H, Fritsche UR, Henneberg K, et al. BUBE: better use of biomass for energy - background report to the position paper of IEA RETD and IEA bioenergy. IEA RETD and IEA Bioenergy. No. 10.3844.56; 2010. Available: https://www.ieabioenergy.com/wp-content/uploads/2013/1 0/Better-Use-of-Biomass-for-Energy-Background-Report.pdf [Accessed: 4-Mar.- 2020]. [128] Lai CS, Locatelli G, Pimm A, Wu X, Lai LL. A review on long-term electrical power system modeling with energy storage. J Cleaner Prod 2021;280:124298. https:// doi.org/10.1016/j.jclepro.2020.124298. [129] Hosseinalizadeh R, Shakouri GH, Amalnick MS, Taghipour P. Economic sizing of a hybrid (PV–WT–FC) renewable energy system (HRES) for stand-alone usages by an optimization-simulation model: Case study of Iran. Renew Sustain Energy Rev 2016;54:139–50. https://doi.org/10.1016/j.rser.2015.09.046. [130] Shang C, Srinivasan D, Reindl T. An improved particle swarm optimisation algorithm applied to battery sizing for stand-alone hybrid power systems. Int J Electr Power Energy Syst 2016;74:104–17. https://doi.org/10.1016/j. ijepes.2015.07.009. [131] Fu R, Feldman DJ, Margolis RM. US solar photovoltaic system cost benchmark: Q1 2018. No. NREL/TP-6A20-72399. Golden, CO, USA: National Renewable Energy Lab (NREL); 2018. [132] Li C. Techno-economic study of off-grid hybrid photovoltaic/battery and photovoltaic/battery/fuel cell power systems in Kunming, China. Energy Sources Part A Recover Util Environ Eff 2019;41(13):1588–604. https://doi.org/10.1080/ 15567036.2018.1549134. [133] Rezk H, Sayed ET, Al-Dhaifallah M, Obaid M, El-Sayed AHM, Abdelkareem MA, et al. Fuel cell as an effective energy storage in reverse osmosis desalination plant powered by photovoltaic system. Energy 2019;175:423–33. https://doi.org/ 10.1016/j.energy.2019.02.167. S. Mohseni et al.