Distributed vs. centralized energy storage for power system applications

0
Distributed vs. Centralized Energy
Storage for Power System Applications
Electrical Engineering Project
Supervisor: Dr Damian Flynn
Student: Numa Gueissaz
University College Dublin
2013-2014

1
Acknowledgment
I am grateful to my supervisor Dr Damian Flynn who guided me in this project and who gave me truly
helpful advices.

0
Contents
1 Introduction.......................................................................................................................................... 1
2 Literature Review ................................................................................................................................. 1
2.1 Distributed and Centralized Energy Storage Systems................................................................... 1
2.2 Applications of Energy Storage Systems ....................................................................................... 3
2.2.1 Grid Stabilization .................................................................................................................... 4
2.2.2 Load Levelling /Peak Shaving ................................................................................................. 4
2.2.3 Energy Arbitrage..................................................................................................................... 5
2.2.4 End User Peak Shaving ........................................................................................................... 5
2.2.5 Reserves.................................................................................................................................. 5
2.2.6 Black Start............................................................................................................................... 6
2.2.7 Integration of Renewable Energy........................................................................................... 6
2.2.8 Uninterruptible Power Supply (UPS)...................................................................................... 8
2.3 Types of Energy Storage Systems.................................................................................................. 9
2.3.1 Electrochemical Energy Storage........................................................................................... 12
2.3.2 Mechanical Energy Storage.................................................................................................. 13
2.3.3 Electrical Energy Storage...................................................................................................... 14
2.3.4 Thermal Energy Storage ....................................................................................................... 15
2.4 Different Energy Storage Systems for Different Applications..................................................... 16
3 Description of the system modelled .................................................................................................. 18
3.1 Description of the Software Used ............................................................................................... 18
3.2 Description of the system modelled ........................................................................................... 18
3.2.1 Generation............................................................................................................................ 18
3.2.2 Load...................................................................................................................................... 18
3.2.3 Transmission Network.......................................................................................................... 19
3.2.4 Diagram of the Test System ................................................................................................. 20
3.3 Description of the ESSs modelled................................................................................................ 22
3.4 Verification Process..................................................................................................................... 23
3.4.1 Is the load looking as expected? .......................................................................................... 23
3.4.2 Which generators are scheduled?........................................................................................ 23
3.5 Different scenarios studied ......................................................................................................... 25
3.5.1 Wind ..................................................................................................................................... 25
3.5.2 Different Locations for ESSs ................................................................................................. 27

1
4 Results ................................................................................................................................................ 27
4.1 How is storage used?................................................................................................................... 27
4.2 Savings Realized by the Load....................................................................................................... 31
4.2.1 No Wind................................................................................................................................ 31
4.2.2 Wind with Ideal Forecast...................................................................................................... 32
4.2.3 Uncertain Wind .................................................................................................................... 33
4.2.4 Conclusion Cost to Load ....................................................................................................... 34
4.3 Generators Net Revenues ........................................................................................................... 35
4.3.1 No Wind................................................................................................................................ 35
4.3.3 Uncertain Wind .................................................................................................................... 36
4.3.4 Conclusion Generator Net Revenues ................................................................................... 36
4.4 Storage Net Revenues................................................................................................................. 37
4.4.1 No Wind................................................................................................................................ 37
4.4.3 Uncertain Wind .................................................................................................................... 38
4.4.4 Conclusion ESSs Revenues.................................................................................................... 39
4.5 Variation of the Characteristics of ESSs....................................................................................... 39
4.6 How to be sure that storage is responsible for the cost reduction............................................. 40
4.7 Downsides of storage.................................................................................................................. 41
4.7.1 Peak Units............................................................................................................................. 42
4.7.2 CO2 Emissions ...................................................................................................................... 43
5 Conclusion .......................................................................................................................................... 44
6 Future work ........................................................................................................................................ 44
7 References.......................................................................................................................................... 45
8 Appendixes......................................................................................................................................... 47
A. Important Results.......................................................................................................................... 47
B. Generator Properties .................................................................................................................... 49
C. Load Distribution........................................................................................................................... 51
D. Transmission Network .................................................................................................................. 52

1
1 Introduction
In modern power systems, the role of an energy storage system (ESS) is more and more discussed.
Recently, California passed an energy storage mandate calling for 1325 MW of energy storage by
2020 [1].The goal of California is to reach 33 percent of its power supply to come from renewables by
2020, and because of the intermittency of their production, solutions must be found to integrate
those sources to the grid. Storage is one of them.
The goal of this project is to investigate the roles of ESSs in a power grid, and more specifically the
differences between centralized and distributed energy storage. Their location on the grid and how
this affects the type of ESS will also be examined. The question of the intermittency of the energy
production and how this affects the storage systems will also be analysed, thanks to the addition of
wind farms to the system.
To achieve this goal, an Australian power market modelling and simulation software, Plexos, was
used. With this software, a modified version of the IEEE reliability Test System was modelled. The
changes were made to introduce distributed ESSs and wind farms.
Once the system was modelled, different scenarios were studied. Three cases were analysed: the
first one was simply the system with no storage; the second one was a centralized ESS (a pumped-
hydro plant); and the last one was three distributed ESSs (batteries).
Wind farms were introduced to see how the uncertainty and the variability of wind were modifying
the previous cases. Wind was first model as perfectly predictable, and then with a difference in the
wind forecast and the actual generation. The purpose of this double model was to see if errors in the
forecast would change the use of storage.
The cost paid by the load will be the main criterion in the comparison of those scenarios (i.e. if ESSs
allow to do savings), but the generators revenues and some other non-economical criterions
(differences in unit commitment, emission of CO2) will also be taken into consideration.
In this report, the literature reviewed will be summed up. Different applications and types of energy
storage systems (ESS) will be reviewed, the difference between centralized ESSs and distributed ones
will be revised, and some real examples of existing ESSs will be given.
2 Literature Review
2.1 Distributed and Centralized Energy Storage Systems
Distributed ESSs can be described as numerous small on-site storage systems, connected to the edge
of the network, whereas centralized storage systems tend to be larger units, in a smaller number,
and connected to the transmission network. Another big difference between distributed ESSs and
centralized ESSs is that the distributed ones are not subject to location restrictions. The major

2
(centralized) ESS is pumped-hydro storage (PHS). However, it is only available in mountainous areas
(underground pumped-hydro has not been realized yet, even if projects exist). This is a strong
drawback in countries without big mountain chains. And even in countries with appropriate
geological formations, the sites where PHS is possible and economically feasible will soon or later all
be used. (In Europe, the potential for pumped hydropower storage is about 10 times the actual one
[2]; however some countries [3] are already using almost all their potential, like Switzerland where
over 85% [3] of the potential is used).
Distributed ESS can be located everywhere on the grid. Either near the generation or near the
consumers, connected to the transmission network or to the distribution one. They have no location
restrictions. This is a great advantage, because the applications they provide change their optimum
location on the grid. For example, a study showed that a superconductive magnetic energy storage
(SMES) combined with a StatCom would have a better performance if it was connected near a load
area instead of near generation (with however an increased cost) [4].
A scheme of a power system with a centralized ESS is shown in Figure 2.1.1.
Figure 2.1 Small network with a centralized ESS (a dam)
The power system in Figure 2.1 is basic. It is composed of generators (that can be wind farms, fuel
plants, nuclear plants, etc.) that are connected to the transmission network. A centralized ESS (a
dam) is also connected to this transmission network. As discussed above, dams are usually far from
Generation
Transmission
Network
Distribution
Network

3
the consumptions nodes and therefore are connected to the transmission network. A distribution
network is also attached to the transmission network, and different loads are connected to it. In
Figure 2.2, the same system is shown but this time with distributed ESSs (here just a battery) at every
node to emphasis the fact that they don’t have any location restrictions.
Figure 1.2 Small network with distributed ESS
2.2 Applications of Energy Storage Systems
There are a lot of different applications in the grid for ESS. The most important ones will be listed and
explained. Given that the topic of my project is related to the grid, transport applications won’t be
considered.
Those applications differ a lot in term of power and energy required, and they have different time
scales.
Transmission
Network
Distribution
Network
Generation

4
2.2.1 Grid Stabilization
Grid stability is the capacity the grid has to regain a state of equilibrium after a disturbance
(transients, interruptions, sags or swells) in order that the system remains intact [5]. Three examples
that can cause grid instability are:
1) Rotor Angle instability happens when a short circuit occurs on a transmission component
that causes the generators to accelerate and lose synchronization with the grid. The
effect is oscillation of the voltage on the line (transients). If those oscillations are not
damped, the generator will trip offline. This may lead to an overload of transmission
equipment, and then lead to the instability of the total system.
2) Frequency excursion occurs when there is a strong difference between generation and
load (e.g. when a generator trips or when generation is not sufficient).
3) Voltage instability occurs when a load require a large amount of reactive power,
exceeding the capacity of the reactive power sources. This increase in load is
accompanied by a drastic voltage drop.
Those cases will now be discussed and we’ll see how storage can help.
2.2.1.1 Rotor Angle Stability
The angular position of the rotors of synchronous generators remains constant to provide voltage in
phase with the grid voltage. However, if synchronism is lost, angular oscillations can occur. An energy
storage system can be used to regulate the oscillations, by switching between charge and discharge
modes at the frequency of the oscillations (e.g. 0.5 to 1 Hz [5]) to cancel their effect. Those storage
systems need a really fast response and the ability to switch between charge and discharge modes
really quickly. Therefore suitable technologies are SMES, super capacitor, flywheels and batteries.
2.2.1.2 Frequency Excursion Suppression
Frequency excursion suppression allows the grid to maintain a stable frequency (in a small interval,
i.e. around 50 Hz in Europe), even in case of severe system disturbance. The stability requires that
the system can quickly recover the balance between load and generation in case of the loss of a part
of the generation or a part of the load [5]. During the time when a generator trips and the time when
the replacement reserves are brought online (between 15 and 30 minutes, depending on the system
and the available units), there are spinning reserves that must be used. Usually, it’s generators that
are increasing their output. But instead of those, or in addition, energy storage systems could be
used. They need to be really fast, have a quite big power and energy output, and therefore the
suitable technologies are batteries.
2.2.1.3 Voltage Stability
To improve the voltage stability, injections of reactive and real power for small durations (order of
seconds [5]) must be produced. SMES, batteries and super capacitors could be used, because of their
ability to provide short bursts of power.
2.2.2 Load Levelling /Peak Shaving
Load levelling is the rescheduling of loads (usually industrial loads) to cut electrical demand; or the
production of energy during off peak period for storage, as shown in Figure 2.3. Peak shaving is the
reduction of the demand during peak demand or the rescheduling of certain loads from peak

5
demand to off peak demand [6]. This may avoid large investments on new transmission or
generation.[7] The suitable ESSs need to have a big capacity (if we want the load levelling or the
peak shaving to be really effective, a large amount of energy must be stored with a high efficiency),
and therefore PHS and CAES seem to be adequate. Batteries are also used. For example, in the USA, a
1.2 MW NaS battery is used in the Charleston Substation to provide 7.2 MWh of on-demand power,
and by peak shaving the load each days, the ESS enables a deferment of transmission upgrade for six
or seven years. Then the battery will be located in another substation where deferment is needed.
Figure 2.3 Concept of Load Levelling
2.2.3 Energy Arbitrage
The ability of storing energy at one time of the day and then discharging it at another time,
effectively shifting the energy consumption, is what is called energy arbitrage. The purpose is to earn
money by doing this, storing low cost energy during time of off-peak demand and selling it at high
cost during time of peak demand. The difference in price between peak and off peak demand must
be big enough to compensate the losses encountered in the storage process.
The ESSs suitable for this are high capacity and long discharge time ESSs, i.e. PHS and CAES or
batteries.
2.2.4 End User Peak Shaving
Undesired grid voltage effects at end/user level can be black out, voltage sags (short period
interruption), voltage peaks and variable fluctuation (flicker) [7]. Almost all consumers need a stable
power supply, especially industrial loads. ESSs can be used to diminish the fluctuations of those end
user negative effects. Suitable technologies are batteries, SMES, super capacitor or flywheels.
Centralized ESSs are not usable because they are usually far from the load.
2.2.5 Reserves
There are different types of reserves that must come on line at different times when there’s
suddenly not enough generation to meet the demand (a generator that trips offline, or the loss of a
transmission line). The first reserve is called the frequency response reserve (because it must be

6
available immediately after the loss of the generator or the transmission line in order to mitigate the
drop in frequency). Then there’s the spinning reserve (composed by online generator increasing their
output, it’s called “spinning” because it refers to the fact that generators are already online and
spinning) and finally the replacement reserve (it is composed by new units that are brought online in
order to compensate the loss in generation).The replacement reserve must be available within the 30
minutes that follow the fault.
Spinning reserve is defined as “the amount of generation capacity that can be used to produce active
power over a given period of time which has not yet been committed to the production of energy
during this period.” [6] It is either generators that can increase their output, or energy storage
systems that can act as generators over a period of time (until the replacement reserve comes in line,
i.e. new generators have been started to meet the demand).
Suitable technologies are really fast devices with high burst of power, like SMES flywheels or super
capacitor in the first seconds (frequency response reserve), then batteries in the first minutes and
finally CAES or PHS as “spinning” reserve.
Figure 2.4 Idealized reserve mechanism after the loss of a part of the generation
2.2.6 Black Start
It is the ability of a power source to go from a shutdown condition to an operation condition without
assistance from the electrical grid. Then the power source can re-energize the grid and help other
generators to go online. Power sources can be reenergized via ESSs, combined with diesel
generators. ESSs need to have significant power capabilities (typically 10 MVA [5]), and sufficient
energy to deploy generator from a cold state, which takes 15 minutes to one hour for a gas turbine.
Suitable technologies are batteries, PHS and CAES.
2.2.7 Integration of Renewable Energy
Wind and solar are two resources that are, by nature, variable and uncertain. Therefore, they won’t
necessarily be available when they are needed. For example, the energy produced by a solar panel is
maximal at noon as shown in Figure 2.5, but the peak consumption is around 6 pm.

7
Figure 2.5 Irradiance in January on a 35 degrees south facing tilted array located at the engineering building of UCD
This Figure doesn’t even account of the variations due to clouds or atmosphere changes. Wind is
even more volatile, and is also difficult to predict. Figures 2.6 and 2.7 show respectively wind
production and electricity consumption in Ireland on a random day, the 16th
of September 2013. We
can see that the wind falls when the consumption is reaching its peak value.
Figure 2.6. Wind generation
Figure 2.7 System demand
Source: EirGrid

8
Storage systems could help integrating the renewables by storing energy when there’s too much
production and not enough consumption and releasing it when the opposite situation occurs.
Suitable technologies are PHS, CAES and batteries. Thermal energy storage (molten salts) is also used
in combination with solar power plants. An example of this combination is the solar thermal power
plant of Andasol in Spain that uses 28 500 tons of salt to provide 50 MW of storage for 7.5 hours, at
an efficiency of 93%.
Another example is a study that has been made in the UK where it was proven that vanadium redox
batteries could be added next to a wind farm to allow it to provide frequency regulation, and
therefore improve its integration to the grid.
2.2.8 Uninterruptible Power Supply (UPS)
Some loads are so critical, that even in the case of the complete loss of the production, they still
need to be supplied. One can think about hospitals, data servers, government vital services, prisons,
etc. This concept is illustrated in Figure2.8.
Figure 2.8 Concept of Uninterruptible Power Supply
In Fairbanks, Alaska, a battery of 40 MW that has 7 minutes of storage can provide emergency power
supply to 12000 people until the diesel generators are started up.
Transmission
Network
Power station Critical load
ESS

9
2.3 Types of Energy Storage Systems
Different applications in which ESSs can be used have been reviewed. However, they all require
different characteristics. Fortunately, there are lots of different energy storage systems and they
differ in many characteristics: efficiency, cycle life, energy and power density and self-discharge.
Those ESSs can be categorized in four groups: electrochemical energy storage, mechanical energy
storage, electrical energy storage, thermal energy storage. The criterion of classification is the form
in which energy is stored.
The ESSs reviewed in this report are shown in Figure 2.9.
Figure 2.9 Different Types of Energy Storage Systems
Those ESSs are also associated with their key parameters in Table 1. For each parameter, the red
cells indicate the best EESs in the category and the blue cells indicate the worst.
Types of Energy Storage
Electrical Energy Storage
1) Super Capacitor
2) Superconductive
Magnetic Energy
Storage (SMES)
Mechanical Energy
Storage
1) Pumped Hydro
storage (PHS)
2) Flywheels
3) Compressed Air
Energy Storage (CAES)
Electrochemical Energy
Storage
Batteries
1)Lead Acid
2) Li-Ion
3) NiMH
4)NaS
5)Flow Batteries
6) Vehicle to Grid (V2G)
Thermal Energy Storage
1) Hot Water Storage
2) Phase Change
Materials

10
Name
Efficiency
[%] Cycle Life
Energy
Density
[Wh/kg]
Power
Density
[W/kg]
Self-
Discharge
Lead Acid 70-80
200-2000
cycles 20-35 25 Low
Li-Ion [7] 70-85
500-2000
cycles 100-200 360 Low
NiMh [7] 50-80
< 3000
cycles 60-80 220 High
NaS [7] 70
2000
cycles 120 120 Medium
Vanadium Redox
[7] 80
1600
cycles 25 80-150 Negligible
Pumped Hydro
Storage 65-85 >20 years 0.3
Depends on
the height Negligible
Flywheels (steel)
[7] 95
20000
cycles 5-30 1000 High
Flywheels
(composite) 95
20000
cycles 50 5000 High
CAES [7] 60-80 >20 years 10-30
Depends on
the plant Low
SMES [8] [9] [10]
>95 20 years 1-11 >10000 High
SuperCapacitor[11] 95 10 cycles Low 4000-10000 High
PCM Molten Salts 70-95 5-12 years 80-200 80-300 Medium
Table 1 Different features of ESSs
We can learn a lot of things from this table. First, the ESSs with a high power density have a rather
small energy density (e.g. SMES, Super Capacitors and flywheels). This indicates that the applications
that provide the ESSs to the grid (either short burst of power or longer period of smaller generation)
will strongly determine the type of ESSs.
We also see that PHS doesn’t have huge advantages on its competitors, except in cycle life. An
interesting fact is that it has a low energy density even though it’s used in energy applications. This
low energy density is counterbalanced (and in fact caused) by the size of the PHS systems which are
enormous. A long cycle life allows big infrastructure to be built, and therefore a long time to recover
the capital costs and therefore a low cost. This low cost is why pumped hydro systems are so widely
used.
The cost has two aspects, a cost per kW and a cost per kWh, reflecting the fact that ESSs cannot be
used in every application. The cost is shown in per kW in Figure 2.01 and in per kWh in Figure 2.11.

11
The source of the data is a paper presented at the 2012 IEEE Electrical Power and Energy Conference
[9]
Figure 2.10 Cost per kWh of different ESSs
Super capacitor and SMES are not shown in this graph because their value of 20000 and 25000$ per
kWh are out of scale.
Figure 2.11 Cost per kW of different ESSs
On those graphs we see that PHS, CAES and TES have a really low per kWh cost, whereas SMES and
batteries have a rather low cost per kW. Lead acid batteries have a low cost in both, but as shown in
Table 1, they have a low cycle life and power density, and a medium energy density and efficiency.
The different types of energy storage systems will now be reviewed.
0
500
1000
1500
2000
2500
Price[$] Cost per kWh
0
200
400
600
800
1000
1200
Price[$]
Cost per kW

12
2.3.1 Electrochemical Energy Storage
2.3.1.1 Lead Acid Battery
It is composed by a “spongy” lead as negative active material and a lead dioxide as positive active
material. The whole system is immersed in a diluted sulfuric acid electrolyte. Its main strength is the
low cost of its materials and production. Unfortunately it has a low power and energy density. It is
mainly used in cost sensitive application where the downsides (low power and energy density) are
not an issue.
2.3.1.2 Li-Ion Battery
It works on the principle of lithium ions moving between the anode and cathode to produce current.
Its strengths are the high energy density and the fact that it has no memory effect (which improves
the life cycle) and a low self-discharge. Unfortunately it is quite expensive to build and lithium is a
limited resource. The main applications are cell phones, laptop and other portable equipment.
2.3.1.3 NiMH Battery
This battery uses nickel oxyhydroxide for the positive electrode and metallic cadmium for the
negative electrode. It has a high power density, a proven safety (it has been used for many years), it
is also resistant and has a long life. However, the self-discharge is high. It was used in the first
electric/hybrid vehicle in the 90’s and 2000’s.
2.3.1.4 NaS Battery
It consists of molten sulphur at the positive electrode and molten sodium at the negative one,
separated by a solid beta alumina ceramic electrolyte. Positive sodium ions go through the
electrolyte and combine with sulphur to form sodium polysulfide. It has a high power and energy
density, a long life cycle, a rather low cost (because it is built in low cost and abundant materials).
However, it has a strong downside: the operating temperature is 300-360 degrees Celsius. Therefore
it needs to be heated externally. It is used in grid application (load levelling, UPS) and the worldwide
installed capacity is about 200 MW [11].
2.3.1.5 Flow batteries (vanadium redox)
This battery has two electrolytes that are stored in tanks, and they are pumped in the reactor where
a reversible chemical reaction happens (production or consumption of current), as shown in Figure
2.12. Therefore the total stored energy is decoupled from the rated power. The rated power depends
on the reactor size and the stored capacity depends on the auxiliary tanks size. This implies that the
capacity is limited only by the size of the tanks. Unfortunately, the energy density is rather low, and
therefore if a big capacity is needed huge tanks must be built. However, if size is not a problem, then
the decoupling is a really strong benefit. The applications provided are power quality control
applications (UPS, frequency and voltage control), emergency power, back-up power, and integration
of renewable energy. The worldwide installed capacity is 38 MW.

13
Figure 2.12 Flow Battery diagram [7]
2.3.1.6 Vehicle to Grid, V2G
Due to different factors (necessity of reducing carbon dioxide to reach the goals set by the Kyoto
Protocol [12], increase in oil prices, environmental awareness) electric or hybrid vehicles are a
growing part of the vehicle sales. Therefore, more and more electric vehicles will be connected to the
grid when they are charging. The batteries of those vehicle (lead-acid, NiMh, lithium-ion) can work as
distributed resources to smooth the load curve [13].
2.3.2 Mechanical Energy Storage
2.3.2.1 Pumped Hydro Storage (PHS)
Water is pumped from a low reservoir to a high reservoir (an artificial lake, retained by a dam) and
therefore energy is stored as potential energy. The water is then released through turbines to
generate power. It has a high capacity, a good efficiency and therefore a low cost. However, it has
geographical constraints (mountainous areas are needed). The applications are: peaking capability,
reserve supply and balancing capability to support renewables sources, load levelling and peak
shaving. It can also work as load during low demand periods, if generation is too high. The size range
depends on the dam. In Ireland, it is 292 MW for the Turlough Hill one[14], but in country with
mountains, it can be way higher, in Switzerland, La Grande Dixence has an installed capacity of 2000
MW[15] and in China the Three Gorges Dam has an installed capacity of 22500 MW [16].
2.3.2.2 Flywheels
It is an ESS that stores energy in a rotating mass. The energy stored is kinetic energy and is given by:
=
1
2
∗ ∗
Where I is the inertia of the rotating mass and is the angular velocity.
Therefore, the higher the speed, the higher the energy stored. As the intertia is given by :

14
= = ∗
in the case of a cylindrical flywheel of mass m and radius r, it is more efficient to increase the speed
of rotation than the mass in order to store more energy.
Therefore, high speed devices are working around 10000 RPM (and even up to 60000 RPM for the
most modern flywheels [17]). Energy is stored and released via an electric motor. The strength of the
flywheel is that it can provide a very high peak power. Therefore it is used in power quality
applications. It is also used in a few companies to avoid that voltage sags or voltage losses interrupt
the industrial production. Nonetheless, it seems not economically viable, as the main constructor,
Beacon Power bankrupted in 2011 [19].The size range of the devices is the kW scale [11].
2.3.2.3 Compressed Air Energy Storage (CAES)
In a CAES plant, electricity is used to compress air during off-peak periods. This compressed air is
stored in a natural cavern (salt mines) or in a reservoir. Then when the demand has grown again, the
air is heated in a gas turbine, it’s expending and gas fuels are added and the turbine is rotating.
There are only 2 existing plants, one in Huntorf, Germany, and the other one in McIntosh, Alabama,
United States. The strengths of this storage system are its high capacity and its low cost. The
downside is the special site requirement. Huntorf plant is mainly used as spinning reserve and peak
shaving and McIntosh plant is used for arbitrage, load management and peak shaving [5]. Lots of
other projects were studied but none were built. A recent project, the Norton plant in Norton, Ohio,
US (a huge 2700 MW plant) was planned first for 2005-2006, but was not realised so far. This is
apparently due to lack of investments due to the economic situation [20].
2.3.3 Electrical Energy Storage
2.3.3.1 Superconductive Magnetic Energy Storage (SMES)
The energy is stored in the magnetic field created by a current flowing through a superconducting
coil. The material, to be superconducting, must be cooled to a temperature of 4-6 K. Therefore the
operating temperature is really low and a cooling circuit, which will consume a lot of energy and
money, must be used. However, SMES have a high power density and have a really fast dynamic
response (range of 1 millisecond). Therefore they are suitable for grid voltage and angular stability
and short duration power quality [5 & 11].
2.3.3.2 Super capacitor
Super capacitors (or ultra-capacitors) work in the same way than usual capacitor: energy is stored in
the capacitor by charge separation. The capacitance is given by:
=
Then the energy stored is given by
=
1
2

15
Where is the vacuum permittivity, is the relative permittivity, A the area of the plates, d the
distance between the two plates and V the voltage applied.
To achieve a great amount of energy stored, the area of the plates (If plate capacitor) and the
relative permittivity of the dielectric must be high, while the distance between the two plates must
be as small as possible. The applications in the grid are short term high power applications
(frequency control, voltage control). It is also used in transport applications, as short duration energy
storage for hybrid-electric vehicle [5]. Another interesting application is the Bridge Power Systems:
the super capacitors can carry a critical load away from a failing source to a stable alternate one, via
a rapid isolation from the failing source. 80-90% of the short circuits are just a momentary missing
voltage replacement, which implies that no alternate source needs to be added (250 ms of average
duration [5]), and therefore super capacitors are perfectly fitting. In case of a severe fault, and during
the transition between bulk mode and island mode (the micro grid is disconnected from the bulk
grid), the super capacitors are helpful to smooth the transition between the two modes. They are
also really efficient to deal with peak load, because of its high power and energy density. They are
also useful if the load needs a high instantaneous current (motors).
2.3.4 Thermal Energy Storage
In thermal energy storage, energy is stored in form of heat, and is released either in this form
(heating systems) or transformed in electricity via a heat turbine.
2.3.4.1 Hot Water Thermal Energy Storage
Energy can be stored by heating water. This is often used when a solar thermal collector is used in a
heating system, but it can also be done at bigger scales in hot water tanks. For example, “ Am
Ackermannbogen” in Munich, Germany is a district where the heating is powered by 2761 of
solar collector. This energy, if not used directly, is stored in a tank of 5700 , which is shown in
construction in Figure 2.13 [21].
Figure 2.2 Thermal Tank Energy Storage

16
2.3.4.2 Phase Change Materials Thermal Energy Storage (PCM)
PCM are materials that are changing their state in a narrow interval of temperature (solid-solid or
solid-liquid). It’s increasing the achievable storage capacity at a given temperature, as shown in
Figure 2.14.
Figure 2.3 [22]
An example of PCM materials is molten salts, heated at 390 degrees in turns into a liquid which is
transferred via a heat exchanger and pumped in a second tank where the heat generated is used to
produce steam that is then used in a steam turbine. The process is shown in Figure 2.15 [23].
Figure 2.4 Use of TES in a solar plant [24]
2.4 Different Energy Storage Systems for Different Applications
We’ve seen that ESSs could be used in a lot of different applications in the grid. Moreover, those
applications require different characteristics in term of power, energy, time of response and duration
and therefore different ESSs are used. Figure 2.16 shows different applications in function of the
power and the duration required, and Figure 2. 17 shows different ESSs that could be used.

17
The cost of those ESSs varies a lot, but it seems that big infrastructure and big investment costs is the
most profitable way of providing kWh (see PHS TES and CAES in Figure 2.10) whereas kW cost tends
to be low in smaller systems like batteries (see Figure2. 11).
Figure 2.5 Power Requirement and Storage Time required for different applications [25]
Figure 2.6 Power and storage time of different ESSs [25]
We’ve seen that the energy, power, time were key parameters for ESSs. The question upon which my
project is based is to determine how another parameter, location, affects those ESSs.

18
3 Description of the system modelled
3.1 Description of the Software Used
PLEXOS® Integrated Energy Model is a power market modelling and simulation software. It can
optimize power systems over different time scales (long-term: 1-40 years, medium-term: 1-5 years,
short-term: < 1 year). For this project, only short term schedule was used. The optimization is
realized using mixed-integer programing (MIP). It deals with pricing, unit commitment, constraints
modelling, Monte Carlo simulation and stochastic optimization.
The simulations were run for one year, with a resolution of one hour, and a day-ahead forecast of 24
hours.
3.2 Description of the system modelled
3.2.1 Generation
The generation system is composed of 32 units that are dispatched among different nodes. There are
2 nuclear steam units (400 MW), 9 coal steam units (1x350 MW, 4x155 MW and 4x76 MW), 11 oil
steam units (3x197 MW, 3x100 MW and 5x12 MW), 4 combustion units (4x20MW) and a hydro plant
(6x50MW). All those plants have different incremental cost curves but they are all piecewise linear.
Their characteristics are shown in appendix B, in Table 3 (capacity, minimum stable factor, heat rates
and start-up costs) and Table 4 (type of fuel used, location, min up and down times, max ramp up
and ramp down times, operation and maintenance charges).
The data is coming from the IEEE reliability system, except for the min up and down times and the
ramp up and down rates that were not provided. They were taken from a European Commission JRC
Institute for Energy paper [26].
The fuel costs are not up to date and were taken as given in the IEEE reliability test system. However
even if this might introduce a small bias (oil is taken as three times more expensive than coal, not
between 7 and 10 as it is today [27]), the system was built with those prices and it didn’t evolve since
(oil plants did not shut down and were not replaced by coal plants).
Wind was also added to the system. The data was the 2013 forecast and actual generation of Ireland
taken from the EirGrid website. It was scaled to the size of the IEEE reliability system.
3.2.2 Load
Using the data from the IEEE reliability test system, a load of one year was modelled, with hourly,
daily, weekly and seasonal variations, as illustrated in Figures 3.1 and 3.2. We see in the weekly
differences that the peak load is smaller in the week-end, and that the overall consumption is also
smaller. This depicts the fact that some people are going out, some industries are closed, etc. We can
also see that late in the evening the consumption is higher than during a week day, and it could show
that people are staying awake later during the weekend.

19
Figure 7 Load during a week day and a weekend day in winter
The seasonal differences are also pretty interesting. We can see that in all cases the peak load is
around meal time (when the oven, microwave, etc. are on). But in summer, it is around noon, which
is probably a consequence of air-conditioning systems that are on during the warmest time of the
day, and in winter, the peak load is at 6pm, and this is because there’s no need for air conditioning
during winter but heating is necessary.
Figure 8 Load a Monday in summer, winter and spring/fall
The peak load is 2850 MW, and the load is dispatched among different buses as shown in Table 5 in
appendix C. The total consumption over the year is 15.3 TWh. Ireland had an electricity consumption
of about 26 TWh in 2011 [28].
3.2.3 Transmission Network
The transmission network is composed of two voltage levels: 138kV and 230 kV. There are 5
transformers between the two parts of the system. Their characteristics and the characteristics of
the lines are shown in Table 6 in appendix D.
0
500
1000
1500
2000
2500
0 4 8 12 16 20 24
LoadMW
Time [hr]
Differences between a week day and a
weekend day
Week day
Weekend day
900
1100
1300
1500
1700
1900
2100
2300
2500
0 2 4 6 8 10 12 14 16 18 20 22 24
Load[MW]
Hours
Seasonal Variations
Winter Monday
Spring/Fall Monday
Summer Monday

20
3.2.4 Diagram of the Test System
The diagram of the test system is shown in Figure 3.3. The addition of the nodes in red was done in
order to have dead-end nodes where it would be easier to add constraints that would allow the use
of storage. If storage was added anywhere in the network, it would not necessarily be used.
Difficulties had to be created to guarantee the use of storage.

21
Figure 3.3 Modified IEEE Reliability Test System

22
3.3 Description of the ESSs modelled
Two types of ESSs were modelled. Both had the same model, shown in Figure3.4. The idea is that
there are two reservoirs and that energy is pumped from the lower reservoir to the upper one,
consuming energy from the grid, and released the other way around to produce energy. The system
with the two reservoirs is a closed-loop system and the efficiency appears only in the generator, i.e.
more energy is inputted to the generator to pump energy than energy is outputted from the
generator when energy stored in the upper reservoir is released. A distributed model, which was
representing a set of flow batteries and a centralized one which was representing a pumped-hydro
system were modelled. The differences between those two systems resided in the characteristics of
the model: efficiency, energy stored, power output, (and of course location in the grid).
Figure 3.4 Model of an ESS
The distributed ESSs were modelled with the following characteristics: an efficiency of 80%, a power
output of 20 MW and a capacity of 0.02 GWh (1 hour of storage). A difference between the pump
mode and the release mode was also introduce to illustrate the fact that it takes more time to store
energy than to release it. Therefore the max release of the tail reservoir was set to 10 MW whereas
the max release of the head reservoir was set at 20 MW.
The centralized ESS was modelled with the following characteristics: an efficiency of 75%, a power
output of 60 MW and a capacity of 0.3 GWh (5 hours of storage).
The comparison between the case with centralized storage and with distributed storage was a bit
ambiguous. Indeed, how can we compare two different scenarios with different characteristics? And
it was not possible to give the same power output and the same capacity to the distributed ESSs and
to the centralized one. A compromise was found by modelling three distributed ESSs that together
would gave the same power output (60 MW) as the centralized one, but they still had a smaller
capacity.
Max
Release
Efficiency
Energy stored in
and released from
the ESS
Max
Volume
Flow of energy in
the ESS

23
3.4 Verification Process
In order to verify if the model of the IEEE Reliability system was working properly (i.e. not doing
switch on/off of units every hours, dispatching the cheapest units first), simulations of the first
month were done and the results analysed.
3.4.1 Is the load looking as expected?
The load for the first month is shown in Figure 3.5
Figure 3.5 Load for the first month
As expected, there are variations during day and night time, and also during week days and weekend
days. There are always 5 days with a peak load at about 2300 and then two weekend days with a
peak load at about 2000. We also see that generation is changing slightly every week. It was also
confirmed with excel that the load was exactly the same than the data inputted.
3.4.2 Which generators are scheduled?
Hydro is always online, because its fuel cost is 0. Therefore its output should be 6x50x24 kWh =
7.2GWh and that’s what we can see, as shown in Figure 3.6

24
Figure 3.6 Hydro generation always scheduled at 7.2 GWh
However, at first, it was not always the cheapest units that were dispatched. For example, as shown
in Figure 3.7, the nuclear units, which are the same, didn’t have the same generation output. This
appeared to be due to transmission constraints, as when they were all removed (ideal transmission
lines, i.e. no resistance or reactance and no max flow), the same units had the same production (e.g.
the two nuclear generators were both set at their maximum output).
Figure 3.7 Same units with different generation
Without transmission constraints, the units were scheduled in this order:
-Hydro (max output)
-Nuclear (max output)
-Big coal plant
-Other coal plants

25
-Big oil plants
-Small oil plants (that were almost never scheduled)
This scheme is what we expect. If expensive units (oil plants) need to be started, bigger units will be
started first because even if they have bigger start-up costs they are meant to be able to generate
energy at a lower cost than small units.
3.5 Different scenarios studied
Different scenarios were studied in this project. They are shown in Figure 3.8, and commented in
sections 3.5.1 and 3.5.2
Figure 3.8 Simulation Scenarios
3.5.1 Wind
The first big difference between those scenarios was the addition or not of wind, which could be
perfectly predictable or have a certain degree of uncertainty. Then simulations were run in each of
those sub-models without storage, with distributed storage near load (at a dead-end node and at the
node just next to it) and near generation and with centralized storage. The uncertainty of the wind
was modelled as shown in Figure 3.9, i.e. with two simulations. The first simulation was done with
the wind forecast, and the unit commitment of slow units (nuclear plants and big coal units) was then
inputted to a new simulation with a realized wind. Therefore fast units had to adapt to the big unit
that were already committed and to the changes in the wind.
Simulation
Ideal Wind Uncertain Wind
No Storage Centralized ESS
Distributed ESS
Near Generation
Distirbuted ESS
Near Dead-End
Load Node
Distirbuted ESS
at Dead-End
Load Node
No Wind

26
Figure 3.9 Wind uncertainty
The differences between wind forecast and realized wind are shown in Figure 3.10. The wind farms
generate 1270 GWh in a year for the forecast case and 1240 GWh for the realized one.
Figure 3.10 Differences in generation between wind forecast and realized wind
The wind penetration was about 20 %, which is quite big since the system has two nuclear plants
(about 20% of the total energy produced) and is a rather small system (that doesn’t have
interconnections to any other grid).
First Run:
Forecast
• Wind
forecast
• Comitment
of slow units
Second Run:
Realized
• Realized
wind
• Comitment
of fast untits
-1000
-500
0
500
1000
1
331
661
991
1321
1651
1981
2311
2641
2971
3301
3631
3961
4291
4621
4951
5281
5611
5941
6271
6601
6931
7261
7591
7921
8251
8581
8911
Generation[MW]
Time [hr]
Differences in generation between wind
forecast and realized wind

27
3.5.2 Different Locations for ESSs
3.5.2.1 Dead End Node
Distributed ESSs were added at the dead-end nodes in red on Figure 3.3 (page 21). The choice of
dead-end nodes was motivated by the fact that difficulties had to be created in order to force the
software to use storage.
3.5.2.2 Node Next to Dead-End Node
The same distributed ESSs have also been placed at buses 1, 2 and 7. Those buses are just next to the
dead-end nodes of the previous case. The purpose of those two cases was to determine whether it
was better to have storage at the extreme edge of the grid or close to the load but still
interconnected with more than one line to the rest of the grid.
3.5.2.3 Distributed Storage near Generation
The same distributed ESSs than in the two previous cases are used. But they are located at different
buses close to generation: bus 13 (3x197 MW Oil plant), bus 15 (155 MW Coal plant and 5 x12 MW
Oil plant) and bus 24 (2x155 MW Coal plant and 350 MW Coal plant).
3.5.2.4 Centralized Storage
The centralized storage was located at bus 17. This bus was chosen because it has connections to the
low voltage grid, to the nuclear plants and to a coal plant. It also doesn’t have any generation or load
on it, which is quite coherent with the fact that a pumped-hydro system is in the mountains, far from
the load and from the generation units.
4 Results
4.1 How is storage used?
Storage is used to provide energy arbitrage (see 2.2.3), i.e. storing energy during low price periods
and restoring it during high price periods. As shown in Figure 4.1, the highest load period, which is
also the highest price period, depends on the season. It is in the morning for both summer and
spring, and in the evening in winter. We also expect energy to be stored at around 4 hours in the
morning in all seasons.

28
Figure 4.1 Load in function of the hour at different seasons
The flows of the upper reservoir of an ESS for one year are shown in Figure 4. 2.
Figure 4.2 Flows of the upper reservoir of an ESS per hour per year
We see that energy is mostly stored between 23 pm and 3 am and released between 8 am and 12
am. This is not exactly what we expected but the reason is simple, when there is no wind the price is
really stable, low during night and high during the day, as shown in Figure 4.3. Therefore it doesn’t
reduce dramatically the cost if energy is stored at 3 am or 4 am.
900
1100
1300
1500
1700
1900
2100
2300
2500
0 2 4 6 8 10 12 14 16 18 20 22 24
Load[MW]
Hours
Seasonal Variations
Winter Monday
Spring/Fall Monday
Summer Monday
-2000
-1500
-1000
-500
0
500
1000
1500
2000
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Inflow[MW/hr/year]
Hours
Flows of the upper reservoir of an ESS
per hour per year

29
Figure 4.3 Price evolution in January
We can see an illustration of those flows in Figure 4. 4. It’s the lower reservoir of a distributed ESS in
the scenario without wind.
Figure 4.4 Inflow And Release of the lower reservoir of a distributed ESS
There are two interesting thing to note. Energy is stored during time of off peak demand (in orange,
the release of the down reservoir) and released during time of peak price period (in blue, inflow of
the down reservoir). The other interesting thing is that the release occurs with a doubled output and
over a shorter time period, reflecting the doubled power output given to the upper reservoir.
However, in the case with the wind, the price is evolving completely differently. We show on Figure
4.5 the inflow and release of the upper reservoir of a distributed ESS and the cost in the case of the
uncertain wind scenario.

30
Figure 4.5 Cost to load and inflow and release of the upper reservoir of a distributed ESS in the scenario with uncertain
wind generation
We can see that the release of the generator doesn’t always occur at the highest price. However,
storage is now not only used for energy arbitrage, but also to integrate the big amount of relatively
uncertain wind generation. We can see this in Figure 4.6 where the wind generation was plotted with
the inflow and release of the upper reservoir of an ESS.
Figure 4.6 Wing generation and inflow and release of the upper reservoir of an ESS in the scenario with uncertain wind
generation
We clearly see that the release of the reservoir corresponds sometimes to drops in the wind
generation.
Another interesting point is the net revenues of the ESSs. They also reflect this process of storing and
releasing energy. As shown in Figure 4.7, there is negative revenue for the generator when it’s
storing energy (from the grid point of view, it’s like the generator is consuming energy, acting like a

31
load), and a positive revenue when it’s releasing energy, acting like a “normal” generator. We also
see that the positive revenues are bigger than the negative, and the ESS is therefore earning money.
We plotted it here with the inflow and release of the upper reservoir to show how this was linked.
Figure 9 Generator net revenues, inflow and release of its upper reservoir. Non-ideal wind scenario with distributed ESS
at dead-end node
4.2 Savings Realized by the Load
The savings realized by the load in the three scenarios are shown in part 4.2.1 (no wind), 4.2.2 (ideal
forecast) and 4.2.3 (non-ideal forecast). Those savings were calculated as follows:
= ! ! !ℎ# $ % !ℎ &! ! # − ! ! !ℎ# $ % !ℎ ! #
4.2.1 No Wind
Figure 4.8 Savings Realized by the Load in the Scenario with no Wind
-8000000
-6000000
-4000000
-2000000
0
2000000
4000000
6000000
8000000
Distributed at
Dead-End
Nodes
Distributed
next to Dead-
End Nodes
Distributed
Near
Generation
Centralized
Savings Realized Without Wind

32
The first thing that strikes us when looking at Figure 4.8 is that even though all the distributed
storage are allowing savings to the load, the centralized one leads to an increase in the cost of about
500 000$. This seems to be due to the fact that the load has to pay for an additional unit on the
system that is not able to survive by itself (the generator has a negative net revenue).
We also see that the distributed case near generation doesn’t allow great savings (apparently
because it’s far from the load, therefore it can’t really act on it).
The best case is the distributed storage next to load. Not at the dead-end node (extreme edge of the
grid) but at the bus just next to it. This may have two reasons: the first one is that the ESSs maybe
need more than just one connection to the rest of the grid in order to be fully operational, and the
second reason could be that because there are also small loads at the buses next to the dead-end
nodes, the ESSs are able to “work” for both loads (the one at the dead-end and the one on the bus
next to it), allowing a better optimization.
4.2.2 Wind with Ideal Forecast
Figure 4.9 Savings Realized by the Load in the Scenario with predictable wind
The introduction of wind has allowed the centralized case to also perform savings for the load.
However, they are still below the one realized in the three distributed cases, which are all allowing
greater savings than in the scenario without wind. We also note that the distributed ESSs near
generation are now almost as performant as the two other ones. This is mainly due to the fact that
they are placed close to wind farms, and therefore have a big optimization potential due to the
variability and the uncertainty of the wind.
0
1000000
2000000
3000000
4000000
5000000
6000000
7000000
8000000
9000000
Distributed at
Dead-End Nodes
Distributed next
to Dead-End
Nodes
Distributed Near
Generation
Centralized
Savings Realized with Ideal Forecast

33
4.2.3 Uncertain Wind
Figure 4.10 Savings Realized by the Load in the Scenario with uncertain wind
In this case, the centralized ESS is generating much more than the three other storage systems,
because of its bigger capacity. Therefore, to reduce the bias discussed in section 3.3, we can divide all
results by the generation of each unit in order to obtain Figure 4.11. The results show that the
distributed ESSs near the load are still the most profitable, and therefore an addition of a lot of those
distributed ESS would me more profitable than the centralized case.
Figure 4.11 Savings per GWh of storage realized by the load. Mind the scale
However, the fact that the big capacity is now allowing the biggest savings informs us of the
importance of a big volume to integrate renewables to the grid. This is especially the case in our
system with a big nuclear generation (slow units) and no connection to other systems (no
0
2000000
4000000
6000000
8000000
10000000
Distributed at
Dead-End Nodes
Distributed next
to Dead-End
Nodes
Distributed Near
Generation
Centralized
Savings Realized With Uncertain
Wind
100000
120000
140000
160000
180000
200000
220000
Distributed at Dead-
End Nodes
Distributed next to
Dead-End Nodes
Distributed Near
Generation
Centralized
Savings Realized With Uncertain Wind
Divided By the Generation of the ESSs

34
opportunity to import or export energy). The storage units have a big role to play, as a load when the
wind generation is big and as a generator when it’s low. Figure 4.12 show how the centralized PHS
helps in the integration of wind. We can see that when the wind farms are generating a lot, the PHS
stores energy, and when the wind drops the PHS releases this energy.
Figure 4.12 Inflow and Release of the centralized ESS and wind generation of one of the 4 wind farms
4.2.4 Conclusion Cost to Load
In the scenario without wind, the savings were the biggest in the case of the distributed ESSs next to
the dead-end nodes, and were representing 1.7% of the total cost without storage (about 7 M$). The
centralized ESS by contrast increased the cost paid by the load by 1.3% (about 5.5M$). The
introduction of a perfectly predictable wind allowed the ESSs to increase their performance, and the
most profitable ones were again the distributed ESSs near the dead-en nodes with savings of 2.2 % of
the total cost (8M$). Finally, the introduction of an uncertain wind has showed the importance of a
big storage capacity in order to reduce the cost paid by the load. The biggest savings were realized by
the ESS with the biggest capacity, i.e. the centralized one with savings of 2.5 % of the total cost
(about 9.25 M$). However, the savings per GWh generated were higher in the case of the distributed
ESSs near the dead-end nodes.

35
4.3 Generators Net Revenues
4.3.1 No Wind
Figure 4.13 Generator net revenues in the scenario without wind Mind the scale
We see in Figure 4.13 that the generator net revenues are all smaller with storage than without. One
possible explanation is that storage reduces the price (the MW provided by the storage units are
“free” MW and therefore the overall price is reduced), and so all generators see their pool revenue
decrease. The peak unit has also its generation reduced.
Figure 4.14 Generator net revenues in the scenario with predictable wind Mind the scale
100000000
105000000
110000000
115000000
120000000
125000000
130000000
No Storage Distributed at
Dead-End
Nodes
Distributed
next to Dead-
End Nodes
Distributed
Near
Generation
Centralized
Generators Net Revenues without Wind
140000000
145000000
150000000
155000000
160000000
165000000
Dead-End
Nodes
Distributed
next to Dead-
End Nodes
Distributed
Near
Generation
Centralized
Generators Net Revenues with Ideal
Forecast

36
The generators have increased a lot their revenue. This is due to the fact that the wind generation
has a zero cost. Therefore its net revenue is just its income and it’s lifting the generators total
revenues up. The conventional generators have however seen their net revenues reduced.
We also see that, as in the case without wind, the introduction of storage is reducing the generator
net revenues, probably for the same reasons.
Figure 4.15 Generator net revenues in the scenario with uncertain wind Mind the scale
With the uncertainty of the wind, the distributed ESSs near the load are not reducing the revenues of
the generators. However, even if the reduction is smaller, the distributed ESSs near generation and
the centralized one are still reducing the generators net revenues.
4.3.4 Conclusion Generator Net Revenues
We’ve seen that storage tends to reduce the generator revenues. When there is no wind, the loss is
included between 2.2% (distributed near load) and 12.6 % (distributed near generation) of the total
revenues when there’s no storage. When wind with an ideal forecast is added, the generators
revenues are also dropping with storage, but this time the interval is much narrower: between 2.5%
(centralized storage) and 6.5% (distributed near generation) of the total revenues. However, when
the wind forecast is not perfect, the loss is reduced to zero in the distributed cases near the load and
is greatly reduced in the case of the ESS near the generation. The reduction of revenues is still big in
the case of the centralized ESS, with a loss of about 10% of the generator revenues.
100000000
105000000
110000000
115000000
120000000
Dead-End
Nodes
Distributed
next to Dead-
End Nodes
Distributed
Near
Generation
Centralized
Generators Net Revenues With
Uncertain Wind

37
4.4 Storage Net Revenues
4.4.1 No Wind
Figure 4.16 ESSs revenues in the scenario without Wind
The revenues of the ESSs are all positive, except in the case of the centralized ESS. The fact that all
the distributed ESSs have positive revenues can be explained by the fact that they earn money by
arbitrage. So how does it come that the centralized doesn’t earn anything? As shown in Figure 4.17,
Plexos doesn’t use the centralized ESS only for arbitrage. Because of its big size, Plexos uses it as
generator when the demand is low and as a load when it is high. It may seem strange, but maybe it
allows avoiding turning on an expensive unit, and therefore saves money. In the end this is what we
expect from a centralized unit: performing system improvements rather than local ones. The fact
that the centralized storage is not used as the distributed ones shows this difference. This is coherent
with what we’ve seen in part 4.2.3, where the centralized ESS was helping to the integration of the
wind, which is a system improvement.
Figure 4.17 Inflow and release of the upper reservoir of the centralized ESS in the case without wind
-20000
30000
80000
130000
180000
230000
280000
330000
380000
430000
Distributed at
Dead-End Nodes
Distributed next
to Dead-End
Nodes
Distributed Near
Generation
Centralized
ESSs Revenues Without Wind

38
Figure 4.18 ESSs Revenues in the Scenario with a perfectly predictable wind
When ideal wind is added, the revenues of the distributed ESSs increase a little bit. This is due to the
fact that the price differential is bigger (when there’s a lot of wind during low demand period the
price is really low and when there’s no wind during high demand period the price is really high ), and
therefore energy arbitrage allows more earnings. The centralized ESS has now a positive income,
which is however still low compared to the other one. It is also helping to the integration of wind.
Figure 4.19 ESSs Revenues in the scenario with an uncertain wind generation.
Revenues of the ESSs have increased again, and once more it’s because of an increase in the price
differential. The centralized ESS has however decreased its income. This inconstancy in the results
also shows us that it is performing larger-scale operation (i.e. Plexos is not trying to maximise its
profits but is using the centralized ESS to improve the total system).
0
50000
100000
150000
200000
250000
300000
350000
400000
450000
Distributed at
Dead-End Nodes
Distributed next
to Dead-End
Nodes
Distributed Near
Generation
Centralized
ESSs Revenues With Ideal Forecast
0
100000
200000
300000
400000
500000
Distributed at
Dead-End Nodes
Distributed next
to Dead-End
Nodes
Distributed Near
Generation
Centralized
ESSs Revenues With Uncertain Wind

39
4.4.4 Conclusion ESSs Revenues
We’ve seen that all the distributed ESSs were earning money through arbitrage. Those revenues
were increasing with the price differential, which was increasing when ideal and non-ideal
predictable wind was added to the system. The centralized ESS however didn’t have any big
revenues. This might be due to the fact that it is not performing energy arbitrage but it is used to
help the whole grid.
4.5 Variation of the Characteristics of ESSs
The distributed ESSs at the dead end nodes were modified in order to examine the effect of their
parameters. The different modifications were the following: doubled power output, doubled capacity
and increased efficiency (from 80% to 90%).
Figure 4.20 Savings realized by the load in the distributed ESS at dead-end node case
We see in Figure 4.20 that the savings realized by the load are greatly increased by the amelioration
of the efficiency and by a bigger power output. The doubled capacity also improves the savings but in
a lesser significant way. However, this must be the easiest thing to improve in flow batteries, as the
storage tanks simply need to be extended (see 2.3.1.5). This also doubles the revenues of the ESS.
Indeed, the energy stored is doubled and therefore twice as much money as in the normal case can
be earned from arbitrage.
0
1000000
2000000
3000000
4000000
5000000
6000000
7000000
8000000
9000000
10000000
Normal Doubled Capacity Increased
efficiency of 10%
Doubled Power
Output
Savings Realized by the Load

40
Figure 4.21 ESS revenues in the distributed ESS at dead-end node case
4.6 How to be sure that storage is responsible for the cost reduction
Apart from the fact that storage is the only thing which is changing between the different cases, we
can also see that those reductions in price are made at hours when the storage generator is working,
strongly relating those two variables. The correlation was calculated as follow:
=
∑( * − ̅) ∗ (-* − -.)
/∑( * − ̅) ∗ /∑(-* − -.)
Where x and y are the two variables, i.e. the flows in the reservoir (shown in Figure 4.22) and the
savings realized (shown in Figure 4.23)
The correlation coefficient found was 0.83, which implies a strong correlation between the two
variables.
Of course correlation doesn’t imply causality, but as the only difference between the scenarios is the
addition of storage, we can assume that storage is the source of the savings.
0
100000
200000
300000
400000
500000
600000
700000
800000
900000
Normal Doubled Capacity Increased
efficiency of 10%
Doubled Power
Output
ESSs Revenues

41
Figure 10 Power flows in the upper reservoir of one of the storage in the distributed scenario with an ideal wind forecast
The negative flows correspond to inflows and the positive flows correspond to release. We can see
that energy is stored at night and released during the day. Note that the sum of the inflow and
release is zero, which is reflecting the nature of the closed loop system formed by the two reservoirs
(see Figure 3.4).
Figure 4.23 Savings made in the case of the distributed scenario with an ideal wind forecast
4.7 Downsides of storage
-2000
-1500
-1000
-500
0
500
1000
1500
2000
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Inflow[MW/hr/year]
Hours
Flows of the upper reservoir of an ESS
per hour per year
-100000
-50000
0
50000
100000
150000
200000
250000
300000
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Savings[$/hr/year]
Hours
Difference in the cost to load between the case with
distributed storage and the case witout storage

42
4.7.1 Peak Units
Some peak units are not used anymore due to the introduction of storage, but they still could be
needed in case of failure of another plant in the system.
We can see in Figure 4.24 that in the case without storage, this unit is generating only at the peak
load, once in a year. But the addition of storage has the result that the unit is not used anymore, as
shown in Figure 4.25.
Figure 4.24 Peak plant generation in the case without storage
Figure 4.25 Peak plant generation in the case with storage. Mind the scale, the unit is actually not generating and the
spikes are approximation of the software
The fact that a plant is not used anymore would lead to its shut down. We could argue that this is the
law of the market and that if a unit is too expensive than it has to be replaced by a cheaper
competitor. Nonetheless, storage cannot produce energy, and in case of a failure somewhere in the
system, this peaking unit might be needed. Therefore, to keep it in the market, capacity payments
should be provided. The question arising now is, is it really profitable to add storage if it leads to an
increase in capacity payments. And the answer is: it depends if those capacity payments are lower
than the savings realized by storage.

43
4.7.2 CO2 Emissions
Another effect of storage is the increase in CO2 emissions. In the test system with no wind, as hydro
is always constant and at max output, the only remaining sources are coal, oil and nuclear plants. So
a decrease in nuclear generation implies an increase in coal and oil generation and therefore an
increase in CO2 emissions.
Figure 4.26 shows the decrease in nuclear generation in the scenario without wind.
Figure 4.26 Generation of the nuclear plants over one year. Mind the scale
We can see that the decrease in nuclear (i.e. the increase in coal and oil generation) is about 6.05
GWh. This is only a little fraction of the total generation (15.3TWh), but this is still the equivalent to
the annual greenhouse gas emissions from almost 900 passenger vehicles [29]. Although it might
seems very small, it’s not negligible, especially for country that are struggling to reach the Kyoto
Protocol requirements.
4292
4294
4296
4298
4300
4302
4304
4306
4308
No Storage Centralized Distributed Near
Gen
Distributed Near
Load
NuclearGeneration[GWh]
Total nuclear generation

44
5 Conclusion
This project has shown that the addition of storage to the test system was profitable for the load
which was able to save money. Distributed energy storage systems near dead-end nodes but still
interconnected to the rest of the network were able to provide the best savings (1.7% of the total
cost paid by the load). However when a big amount of uncertain wind generation was added, the
centralized ESS was providing the best savings because of its high capacity and because it was not
only doing energy arbitrage. The parameters of the distributed ESSs were also studied and it
appeared that the best way to improve the load savings was to improve the efficiency of the storage
devices. However, it is in practice nearly impossible to do and an increased capacity, even if it didn’t
greatly improved the savings in the cost paid by the load was significantly increasing the revenues of
the ESSs.
However, some downsides have also arisen. The generation revenues were reduced (up to 10%) by
the addition of storage. ESSs also introduced an increase in the use of coal and oil plants over
nuclear, which led to an increase in C02 emissions. Storage also forced some peaking units to shut
down.
We’ve seen that the location of the storage systems strongly influence the application they are
providing. In the case of the distributed units, energy arbitrage and a bit of peak shaving were
providing but the centralized ESS was also helping to the integration of renewable energy.
6 Future work
The modelling of a real system with real characteristics (reserves constraints, electricity market, a
smaller resolution of the simulation, a longer day-ahead forecast, and generator outage) should be
undertaken to see if the same phenomenon would be observed in a tangible case. The precision of
the storage model should also be increased (addition of constraints like ramp up and down rates,
self-discharge) to allow different applications (frequency control, etc.). Different ESS should also be
modelled.
A further analysis including the capital costs of the storage systems will also have to be undertaken to
determine if the ESSs are economically viable for the people who invest in it.
When it comes to the integration of renewables, storage should also be rigorously compared to some
alternatives like curtailment or interconnections with bigger networks.
Another point that will have to be taken into account is the energetic cost of production and
operation of the ESSs. A recent study of Stanford University has shown that the overall energetic cost
of batteries was not compensated by the energy saved from curtailment in the case of wind farms
[30 & 31].

45
7 References
[1] Newmann L.H., “What California’s Energy Storage Requirement Really Means”, IEEE Spectrum, 8
Nov 2013
[2] European Commission website
http://ec.europa.eu/dgs/jrc/index.cfm?id=1410&obj_id=16750&dt_code=NWS&lang=en
[3] World Energy Council, “World Energy resources: Hydro”, 2013
http://www.worldenergy.org/wp-content/uploads/2013/10/WER_2013_5_Hydro.pdf
[4] Arsoy A., “Power Converter and SMES in Controlling Power System Dynamics”, Industry
Application Conference 2000
[5] EPRI-DOE Handbook of Energy Storage for Transmission & Distribution Applications, EPRI, Palo
Alto, CA and the U.S. Department of Energy, Washington DC , 2003
[6] Mohd A., “Challenges in Integrating Distributed Energy Storage Systems into Future Smart Grid”,
IEEE International Symposium on Industrial Electronics, 2008
[7] Vazquez S., “Energy Storage Systems for Transport and Grid Applications“, IEEE Transactions on
Industrial Electronics volume 57, Dec.2010
[8] Boyes J.D., “Technologies for Energy Storage. Flywheels and Super Conducting Magnetic Energy
Storage”, Power Engineering Society Summer Meeting, 16-20 Jul 2000
[9] Bhuiyan F.A., « Energy Storage Technologies for Grid-Connected and Off-Grid Power System
Applications », 2012 IEEE Electrical Power and Energy Conference
[10] Mohd H.A., “An Overview of SMES Applications in Power and Energy Systems”, IEEE Transaction
on Sustainable Energy, volume 1, April 2010
[11] Styczynski Z.A., “Electric Energy Storage and its Tasks in the Integration of Wide-Scale Renewable
Resources”, CIGRE/IEEE PES Joint Symposium, 29-31 July 2009
[12] United Nations, Kyoto Protocol
http://unfccc.int/essential_background/kyoto_protocol/items/1678.php
[13] Tu Yiyun, “Research on Vehicle-to-Grid Technology”, 2011 International Conference on
Computer Distributed Control and Intelligent Environmental Monitoring, 19-20 Feb. 2011
[14] ESB Website
http://www.esb.ie/main/about-esb/ardnacrusha.jsp
[15] Grande-Dixence Dam Website
http://www.grande-dixence.ch/
[16] China Three Gorges Corporation Website
http://www.ctgpc.com/

46
[17] "Breakthrough in Ricardo Kinergy ‘second generation’ high-speed flywheel technology"; Press
release date: 22 August 2011. Retrieved 2012-07-03
[18] Lawrence R.G., “Flywheel UPS”, IEEE Industry Applications Magazine volume 9, May-June 2003
[19] Postelwait J., « Beacon Power emerges from bankruptcy with new energy storage project »,
PowerGrid International, 24 June 2013.
[20] Morehead B., “Mine Use Project still in Limbo “, The Norton Post, 2 Jan. 2013
[21] Reuss, M., « Solar District Heating in Germany », Bavarian Centre for Applied Energy Research,
ZAE BAYERN
[22] Hauer, A., “Thermal Energy Storage, Technology Brief”, IEA-ETSAP and IRENA 2013
[23] Siemens AG Energy Sector, “Steam turbines for CSP plants”, 2011
[24] Copyright Solar Millenium AG
[25] Oudalov, A., Buehler, T., Chartouni, D., “Utility Scale Applications of Energy Storage”, ABB
Switzerland Ltd., IEEE Energy 2030, Atlanta, Georgia, SA, 17-18 November 2008
[26] Bruynooghe, C., Eriksson, A., Fulli, G., “Load-following operating mode at Nuclear Power Plants
(NPPs) and incidence on Operation and Maintenance (O&M) costs. Compatibility with wind power
variability”. JRC Scientific and Technical Reports, European Commission, Institute for Energy, 2010
[27] Sustainable Energy Authority of Ireland (SEAI), “Commercial Fuel Cost Comparison January
2014”, 01 January 2014
[28] Central Intelligence Agency, the World Factbook about Ireland,
https://www.cia.gov/library/publications/the-world-factbook/geos/ei.html
[29]U.S. Environmental Protection Agency, Greenhouse Gas Equivalencies Calculator
http://www.epa.gov/cleanenergy/energy-resources/calculator.html#results
[30] Barnhart, C., Dale, M., Brandt, A., Benson, S., “The energetic implications of curtailing versus
storing solar- and wind-generated electricity”, Energy & Environmental Science, 28 August 2013
[31] Edgette, C., Barnhart, C., “ Clarifying the Confusion-Storage and Cost Effectiveness”, Alt Energy
Stocks, 29 September 2013
http://www.altenergystocks.com/archives/2013/09/clarifying_the_confusion_storage_and_cost_eff
ectiveness.html

47
8 Appendixes
A. Important Results
Table 2 Important Results
Type of
storage
Scenario Cost To
Load [$]
Generation
cost [$]
Generators
NET Revenues
[$]
Storage
Generation
[GWh]
Storage
Revenues [$]
No Storage No Wind 419358043 169361977 124760722 0 0
No Storage Perfect
Wind
373567342 114051575 159940176 0 0
No Storage Non
Perfect
Wind
373672727 121551519 119050237 0 0
Distributed at
dead-end
node
No Wind 412415063 169098787 121988208 20.88 375817
Distributed at
dead-end
node
Perfect
Wind
365739445 113686594 154226181 21.0864 408384
Distributed at
dead-end
node
Non
Perfect
Wind
369574298 121276908 119043912 20.7264 451125
Distributed
2mwh
No Wind 411495517 168789696 121845792 41.76 799129
Distributed
2mwh
Perfect
Wind
368617860 113360003 156816002 40.548 934068
Distributed
2mwh
Non
Perfect
Wind
377539176 126259897 117306600 40.404 959569
Distributed
90%
No Wind 410219570 169099453 119614907 20.8836 346787
Distributed
90%
Perfect
Wind
365727437 113672593 155073516 21.6024 429960
Distributed
90%
Non
Perfect
Wind
370070275 121237514 118730627 20.8404 457444
Distributed
2MW
No Wind 410632980 169103296 120587545 20.88 349104
Distributed
2MW
Perfect
Wind
365176283 113700278 154530601 20.988 373667
Distributed
2MW
Non
Perfect
Wind
369960277 121280342 119350633 21.99 409298
Distributed
next to dead-
end node
No Wind 412306489 169101455 121503948 20.88 377940

48
Distributed
next to dead-
end node
Perfect
Wind
365569844 113692656 155033054 20.9328 409644
Distributed
next to dead-
end node
Non
Perfect
Wind
369471944 121270707 119066689 20.8332 447132
Distributed
Near
Generation
No Wind 416865997 169237163 109022903 19.914 161557
Distributed
Near
Generation
Perfect
Wind
365733843 113835491 149556467 20.1408 280463
Distributed
Near
Generation
Non
Perfect
Wind
370369924 121382257 116364853 20.7948 282731
Centralized No Wind 424892115 169204534 118814713 20.88 -8400
Centralized Perfect
Wind
372840039 113863562 155935031 55.56 62822
Centralized Non
Perfect
Wind
364395485 120247921 107110344 53.04 11971

49
B. Generator Properties
Table 3, Characteristics of the generation system.
Name
of the
plant
Max
Capacity
[MW]
Min
Stable
factor
[%]
Load
Point
1
[MW]
Load
Point
2
[MW]
Load
Point
3
[MW]
Load
Point
4
[MW]
Heat Rate
[Gj/MWh]
Heat
Rate
2
Heat
Rate
3
Heat
Rate 4
Startup
Costs
[$]
Nuclear
Plant
400 25 100 200 320 400 13.2 11.4 10.7 10.6 40000
155
MW
Coal
Plant
155 35 54.25 93 124 155 11.8 10.7 10.3 10.2 9300
2x76
MW
Coal
plant
76 20 15.2 38 60.8 76 16.5 13.6 12.6 12.7 5320
350
MW
Coal
plant
350 40 140 227.5 280 350 10.7 10.1 10.
02
10.0
2
1400
0
3x100
MW Oil
Plant
100 25 25 55 80 100 13.7 11.2 10.7 10.6 3000
3x197
MW Oil
plant
197 40 68.95 118.2 157.6 197 11.3 10.4 10.4 10.1 4925
5x12
MW Oil
Plant
12 20 2.4 6 9.6 12 16.5 13.6 12.6 12.7 120
2x20
MW Oil
plant1
20 80 16 20 15.8 15.3 200

50
Table 4 Characteristics of the generation system
Name of
the
plant
Fuel
used
Number
of Units
Location Min Up
Time
[hr]
Min Down
Time [hr]
Max
Ramp Up
[MW/min]
Max
Ramp
Down
[MW/min]
FO&M Charge
[$/kW/year]
Nuclear
Plant1
LWR 1 Bus18 136 24 8 5
Nuclear
Plant2
LWR 1 Bus21 136 24 8 8 5
155 MW
Coal
Plant1
Coal 1 Bus15 8 2 8 15.5 0.8
155 MW
Coal
Plant2
Coal 1 Bus16 8 2 10 15.5 0.8
155 MW
Coal
Plant3
Coal 1 Bus24 8 2 10 15.5 0.8
155 MW
Coal
Plant4
Coal 1 Bus24 8 2 10 15.5 0.8
2x76
MW
Coal
plant1
Coal 2 Bus01 6 2 10 15.2 10
2x76
MW
Coal
plant2
Coal 2 Bus02 6 2 10 15.2 10
350 MW
Coal
plant
Coal 1 Bus24 10 2 10 35 4.5
3x100
MW Oil
Plant
Oil
$6
3 Bus07 10 2 20 10 0.8
3x197
MW Oil
plant
Oil
$6
3 Bus13 10 2 10 20 0.7
5x12
MW Oil
Plant
Oil
$6
5 Bus15 5 2 20 1 10
2x20
MW Oil
plant1
Oil
$2
2 Bus01 2 2 1 2 0.3
2x20
MW Oil
plant2
Oil
$2
2 Bus02 2 2 2 2 0.3

51
C. Load Distribution
Table 5 Distribution of the load among the buses
Node Voltage
[kV]
Load participation factor
[p.u.]
Bus01 138 0.038
Bus02 138 0.034
Bus03 138 0.063
Bus04 138 0.026
Bus05 138 0.025
Bus06 138 0.048
Bus07 138 0.044
Bus08 138 0.06
Bus09 138 0.061
Bus10 138 0.068
Bus11 230 0
Bus12 230 0
Bus13 230 0.093
Bus14 230 0.068
Bus15 230 0.0555
Bus16 230 0.035
Bus17 230 0
Bus18 230 0.0585
Bus19 230 0.032
Bus20 230 0.045
Bus21 230 0
Bus22 230 0
Bus23 230 0
Bus24 230 0
busstorage1 138 0.048
WindNode 230 0

52
D. Transmission Network
Table 6 Transmission Network Properties
Name Node from Node to Voltage [kV] Resistance [p.u.] Reactance [p.u.]
01-2 Bus01 Bus02 175 0.0026 0.0129
01-3 Bus01 Bus03 175 0.00546 0.2112
01-5 Bus01 Bus05 175 0.0218 0.0845
02-4 Bus02 Bus04 175 0.0328 0.1267
02-6 Bus02 Bus06 175 0.0497 0.192
03-9 Bus03 Bus09 175 0.0023 0.0839
04-9 Bus04 Bus09 175 0.0268 0.1037
05-10 Bus05 Bus10 175 0.0139 0.0883
06-10 Bus06 Bus10 175 0.0159 0.0605
07-8 Bus07 Bus08 175 0.0159 0.0614
08-10 Bus08 Bus10 175 0.0427 0.1651
08-9 Bus08 Bus09 175 0.0427 0.1651
linebusstorage1 busstorage1 Bus01 150 0.0026 0.0129
11-13 Bus11 Bus13 500 0.0061 0.0476
11-14 Bus11 Bus14 500 0.0054 0.0418
12-13 Bus12 Bus13 500 0.0061 0.0476
12-23 Bus12 Bus23 500 0.0124 0.0966
13-23 Bus13 Bus23 500 0.0111 0.0865
14-16 Bus14 Bus16 500 0.005 0.0389
15-16 Bus15 Bus16 500 0.0063 0.049
15-21 Bus15 Bus21 500 0.0063 0.049
15-21 (1) Bus15 Bus24 500 0.0067 0.0519
15-24 Bus15 Bus24 500 0.0022 0.0173
16-17 Bus16 Bus17 500 0.0033 0.0259
16-19 Bus16 Bus19 500 0.003 0.0231
17-18 Bus17 Bus18 500 0.0018 0.0144
17-22 Bus17 Bus22 500 0.0135 0.1053
18-21 Bus18 Bus21 500 0.0033 0.0259
18-21 (1) Bus18 Bus21 500 0.0033 0.0259
19-20 Bus19 Bus20 500 0.0051 0.0396
19-20 (1) Bus19 Bus20 500 0.0051 0.0396
20-23 Bus20 Bus23 500 0.0028 0.0216
20-23 (1) Bus20 Bus23 500 0.0028 0.0216
21-22 Bus21 Bus22 500 0.0087 0.678
Wind1to20 WindNode Bus20 500 0.0087 0.678

Distributed vs. centralized energy storage for power system applications

More Related Content

What's hot (20)

Viewers also liked (8)

Similar to Distributed vs. centralized energy storage for power system applications (20)

More from Andrew Gelston (20)

Distributed vs. centralized energy storage for power system applications