Optimal state estimation techniques for accurate measurements in internet of things enabled microgrids using deep neural networks

International Journal of Electrical and Computer Engineering (IJECE)
Vol. 12, No. 4, August 2022, pp. 4288~4301
ISSN: 2088-8708, DOI: 10.11591/ijece.v12i4.pp4288-4301  4288
Journal homepage: http://ijece.iaescore.com
Optimal state estimation techniques for accurate measurements
in internet of things enabled microgrids using deep neural
networks
Sudhakar Rao Padupanambur1
, Mohammed Riyaz Ahmed2
, Bangalore Prabhakar Divakar1
1
School of Electrical and Electronics Engineering, REVA University, Bengaluru, India
2
School of Multidisciplinary Studies, REVA University, Bengaluru, India
Article Info ABSTRACT
Article history:
Received Dec 20, 2021
Revised Jan 9, 2022
Accepted Feb 5, 2022
The employment of microgrids in smart cities is not only changing the
landscape of power generation, transmission, and distribution but it helps in
green alleviation by converting passive consumers into active produces
(using renewable energy sources). Real-time monitoring is a crucial factor in
the successful adoption of microgrids. Real-time state estimation of a
microgrid is possible through internet-of-things (IoT). State estimation can
provide the necessary monitoring of grid for many system optimization
applications. We will use raw and missing data before we learn from data,
the processing must be done. This paper describes various Kalman variants
use for preprocessing. In this paper a formulated approach along with
algorithms are described for optimal state estimation and forecasting, with
weights update using deep neural networks (DNN) is presented to enable
accurate measurements at component and system level model analysis in an
IoT enabled microgrid. The real load data experiments are carried out on the
IEEE 118-bus benchmark system for the power system state estimation and
forecasting. This research paves a way for developing a novel DNN based
algorithms for a power system under dynamically varying conditions and
corresponding time dependencies.
Keywords:
Deep neural network
Internet of things
Kalman filter
Microgrid
Recurrent neural network
Smart grid
State estimation
This is an open access article under the CC BY-SA license.
Corresponding Author:
Sudhakar Rao Padupanambur
School of Electrical and Electronics Engineering, REVA University
560064, Rukmini Knowledge Park, Kattigenahalli, Yelahanka, Bengaluru, India
Email: sudhakar.rao@reva.edu.in
1. INTRODUCTION
The centralized power distribution has dominated the energy sector in the last century; but the
exponential growth of load, need of power quality, environmental issues and shortage of fossil fuels have
challenged the conventional methods [1]. As a consequence, many researchers started investigation on
adoption of small-scale distributed generation by distributed energy resources (DERs). Recently, microgrid
has received huge momentum due to its economic merit and environment friendly nature. With small,
distributed grids consisting of sophisticated devices for power distribution, energy conversion and storage,
microgrid has become a game changer in the energy sector [2]. This new trend has shifted the paradigm from
centralized energy generation and distribution to local energy conversion, energy storage, power quality and
optimization to enable the citizen of smart city to actively participate in production and distribution via smart
grids [3]. The ever-increasing demand and usage of power have triggered investigations in the way that
electricity is generated, distributed, managed, and consumed. In the field of power distribution, hitherto
unprecedented requirements are expected to be satisfied [4]. Though microgrid has many advantages such as

Int J Elec & Comp Eng ISSN: 2088-8708 
Optimal state estimation techniques for accurate measurements in … (Sudhakar Rao Padupanambur)
4289
lower cost, high reliability, reduced carbon footprint, improved power quality, and asset security, its ability to
convert passive consumers into active suppliers by enabling them to harvest renewable sources of energy is
the key reason for the massive adoption of it [5].
Despite the fact that microgrid has increased our optimism about renewable sources, it comes with
certain limitations such as frequent maintenance of equipment and need for load selection, power factor
correction and new schemes of revenue collection. Microgrids are designed to enable two-way power flow
and two-way data flow that require pervasive (and ubiquitous) instrumentation and connectivity on just about
to every device within it [6]. Highly instrumented microgrids can strengthen grid resilience and help
minimize outages in the larger utility grid. The ability of microgrid to operate autonomously from larger
grids could prove to be a major driver of internet of things (IoT) adoption in smart cities of the future. IoT
comes into picture when massive connectivity of sensors is required for analytics through machine learning
for optimization [7]. As presented in Figure 1, the IoT ensures smooth functioning of a microgrid by enabling
remote monitoring and control via web applications thereby conveying real-time info on hand-held
multimedia devices. The data stored in smart meters will enable the electricity board to fix the tariff, create
policies, and analyze the trends in electricity usage [8]. Fog computing can be used for analysis of voltage,
current, and power locally while the data is stored at the cloud, centrally. The research gaps are identified on
the research challenges, and these are presented in Table 1 of literature review conducted in brief.
Figure 1. IoT communication network and microgrid dynamic state estimation process
Table 1. Summary of literature review conducted
References Variable parameter Methodology adopted Research gap
[9] Efficiency of microgrid
system
Control of energy storage under price
fluctuation.
Problem of optimality in the battery
[10] Optimality of micro-grid
system
Manage the microgrid under load
shifting approach
Optimal charging and discharging of
battery cycles.
[11] Minimized energy cost Battery charge from utility of grid only Optimal exploitation of renewable energy.
[12] Power flow from battery
to grid
A model free reinforcement learning. Cost minimizing storage policy.
[13] Electrical storage shifting Load shifting approach Usage of heterogeneous approaches.
The convergence of technological, economic, and environmental forces, are driving the IoT and
microgrids simultaneously, then each drives the other forward. Be it the IoT driving microgrids or vice-versa,
the continued expansion of each is good for the other. The current trends we see in the microgrids and IoT, as
well as energy generation and storage technologies, evoke visions of a vast network of microgrids balancing
and optimizing energy use in a deeply connected world. The proliferation of IoT enabled microgrids will
open up new frontiers of services and applications for next-generation power and energy systems. Although
many communication infrastructures are proposed, a systematic study of a suitable two-way communication
infrastructure for sensing, estimating, and controlling the microgrid incorporating multiple distributed energy

 ISSN: 2088-8708
Int J Elec & Comp Eng, Vol. 12, No. 4, August 2022: 4288-4301
4290
sources is yet to be carried out. The problem statement is narrowed down as to develop a novel deep learning
(DL) algorithm for power systems under dynamically varying conditions and corresponding time
dependencies. Based on the problem statement the key contributions of this research are: i) a formulated
approach to enable accurate measurements at component and system level model analysis using a deep neural
network (DNN) architecture in an IoT enabled microgrid, ii) to describe several Kalman variants used for
pre-processor of raw, redundant data and produce a reliable state estimate and forecasting using a DNN, and
iii) to investigate the various ways of distributed system state estimation via Kalman filters and its variants
with a DNN.
2. RESEARCH METHOD
In the mathematical foundations of Kalman filter, a recursive algorithm is that works on dynamic
systems and uses measurement, estimation, prediction, and updating is presented in (1).
𝑥𝑘 = 𝐴𝑘𝑥𝑘 + 𝐵𝑘𝑢𝑘 + 𝑤𝑘 (1)
With 𝑥𝑘 Є 𝑅𝑛, 𝐴𝑘, 𝑤𝑘, 𝐵𝑘 representing transition matrix, process noise, and control input model respectively,
and measurement (or observation) of 𝑧𝑘 is depicted in (2).
𝑧𝑘 = 𝐻𝑘𝑥𝑘 + 𝑣𝑘 (2)
With 𝑧𝑘 Є 𝑅𝑚, and 𝐻𝑘 being observation model if process noise wk and the measurement noise 𝑣𝑘 are
assumed to be an independent random variable with Gaussian probability density functions and zero mean
value. Let us go about to understand the model with equations, first of which is the state update (3).
𝑥𝑘 = 𝑥𝑘−1 + 𝑧𝑘 − 𝐻𝑘−1𝑥𝑘−1 (3)
Here xk−1 represents the previous estimate for the kth
instant. The difference between the latest measurement
and the previous estimate gives the measurement residual called innovation represented by yk. The weight of
innovation decreases with a greater number of iterations as the Kalman gain decreases.
This state update equation gives the present estimate at the kth
instant. State extrapolation equation is
the immediate shift from the state update equation after an instant. In (4) extrapolates the current state to the
next state.
𝑥𝑘−1 = 𝑥|𝑘−1|+ 𝐾𝑘𝑦𝑘 (4)
Hence giving the prediction of the next state. From (4), the factor that is responsible for the addition of the
difference between the present measurement and the previous estimate to the previous estimate is called
Kalman gain [14]. This is defined by the third Kalman Filter equation, Kalman gain is presented in (5),
Kk = Pk−1𝐻𝐾
𝑇
(HkPk − 1𝐻𝐾
𝑇
+ Rk) (5)
This is the ratio of the in estimate to the sum of the uncertainty in estimate with the uncertainty in the
measurement.
[0 ≤ 𝐾𝑛 ≤ 1]
From (3)
𝑥𝑘 = 𝑥𝑘−1(1 − 𝐻𝑘 . 𝐾𝑘 + 𝑍𝑘 𝐾𝑘 ) (6)
Hence, the weight is given to the measurement and 1-Hk is given to the estimate, as the number of iterations
increases the significance of the measurements decreases and the new value comes closer to the previous
estimate value [15]. The uncertainty in the estimate needs to be updated with the updating of the state of the
dynamic system, for finding the present uncertainty in the estimate using the previous uncertainty, estimate
uncertainty update equation or covariance update (7):
𝑃𝑘 = (1 − 𝐻𝑘 . 𝐾𝑘 ) 𝑃𝑘−1 (7)

4291
where (1 − 𝐾𝑘 ≤ 1) as the number of iterations increases, uncertainty in the present estimate decreases.
Estimate uncertainty extrapolation equation as presented in (8) is applied to realize the present uncertainty
using the previous uncertainty.
𝑃𝑘−1 = (𝐴𝑘 . 𝑃𝑘−1 )𝑇𝑘 + 𝑄 (8)
So, (4) and (8) is categorized in priori, and represents the prediction of co-variance and state of the system
[16], [17]. The equation (3) and (6) is categorized in posteriori and represents the correction of covariance
and state of the system with the block diagram as depicted in Figure 2 which describes the workflow of
categorized priori along with prediction of co-variance and state of system.
The Table 2 describes the summary of all Kalman filter variants based on state estimators. The
different state estimators are extended, unscented, ensembled, and particle Kalman filters [18], [19]. They
present a system model (linear, locally linear, non-linear) description for each variant of thee Kalman filter
state estimator. Further the arranged distributions (Gaussian, non-Gaussian) along with a computational cost
(low, medium, and high) is presented for each of the Kalman filter state estimator variants [20], [21].
Figure 2. Workflow of categorized priori along with prediction of co-variance and state of system
Table 2. Summary of all Kalman based state estimators
State estimator System model Assumed distribution Computational cost
Kalman Filter (KF) Linear Gaussian Low
Extended KF (EKF) Locally linear Gaussian Low
Unscented KF (UKF) Non-linear Gaussian Medium
Ensembled KF (EnKF) Non-linear Non-gaussian High
Particle KF (PKF) Non-linear Non-gaussian High
The Figure 3 illustrates the IoT based smart grid network here in several instances of IoT based
system, it has a tiny-computing device for monitoring, controlling, which often very limited in terms of
resources and capacities, energy usage and also are constrained to communication bandwidth and network
efficiency. Message queuing telemetry transport (MQTT) is a publishing/subscribing messaging protocol, it
is designed to be open, simple, lightweight. In MQTT as a client, subscriber and publisher exist,
communications are managed by a broker. Subscribers are linked to a broker and subscribe to interesting
topics. The publisher also communicates and publishes issues with the broker. The publishing message will
soon be sent to the subscriber who has subscribed to the message via the broker. In this case, its message is
published with a topic, a smart meter in the distribution board measures how much power is consumed.

 ISSN: 2088-8708
4292
IoT-gateway: the IoT-gateway is a linking bridge between the sensor network and the MQTT
network. This system is constructed of two components; it is responsible for the communication of devices
and the collection of data. The contact of the device is, any request arrives via MQTT at the IoT-gateway,
then an IoT gateway operates a device via the network of the target device and responds via MQTT to the
requestor.
As depicted in Figure 4(a) a real time power system state estimation and forecasting of the historical
voltage data is presented. The Figure 4(b) represents deep neural network model unfolded without the
outputs, and Figure 4(c) presents deep neural network based real time monitoring of the power system. For
tracking, automatic control and two-way communication functions, intelligent sensors and actuators are
deployed. In other words, the voltage fluctuation is observed by several smart sensors, which will relay the
information to the local intelligent control unit. The control unit calculates the deviation of the device state
from a pre-defined reference value based on the received signal. In order to monitor the state variations, the
predicted states are then used by voltage regulators.
Figure 3. IoT based smart grid network
Figure 4. Real time power system state estimation and forecasting (a) deep recurrent neural network (RNN)
based real time power system state estimation, (b) a deep RNN architecture model, and (c) DNN based real
time monitoring of power system
3. ALGORITHM DESIGN
In the algorithm the weights are taken by using importance sampling. The samplings (support
points) are taken sequentially as measurements and in this algorithm 1. The weights are updated to approach
the true posterior density. The algorithm 1 discretizes the system state into a collection of state vectors
analyzed. By integrating the ensemble of the state model, the mean and error covariance can be found [22],
[23]. As the number of ensembles increases the probability density function can be approximated more
accurately [24].

4293
The algorithm 2 reduces the computational work, as it is sufficient to use a limited number of model
states for reasonable statistical convergence. X2[x1, ..., xn]=[xi], is n×N matrix, prior ensemble matrix,
D =[d1,...., dn], is m×N matrix, where d is the data column and has random vector from n-dimensional normal
distribution N (0, R). The research further steps into algorithm 3 using the steps illustrated in the algorithm 2.
In algorithm 3 the deep recurrent neural networks (Deep RNN’s) are introduced for state estimation and
forecasting using the historical data of the estimated voltages measured from sigma points in state space and
measurements voltage data obtained from the sensors as illustrated in algorithm 2. The algorithm 3 is framed
taking the Figure 4(a) to Figure 4(c) into the consideration [25].
Algorithm 1. Algorithm for optimal state estimation and accuracy measurements
Start
Input: sigma points: system dimensions (2N+1)
N=Dimensionality of System;
while (N)
Calculate:
X: sigma point matrix;
µ: mean of Gaussian;
λ: scaling factor;
ψ: distance of sigma point from mean point;
if (ψ>=λ)
P:find covariance matrix weights;
else if (sum of all weights=1)
µ`: predicted mean;
ξ’: predicted co-variance;
w: weights of sigma point;
g: non-linear function;
end if()
end if()
Obtain the measurements from sensors;
Compute the update step;
end while()
Perform optimal state estimation;
Obtain accuracy;
Stop
Algorithm 2. Algorithm for weight update
Start
Input: Sigma Points: System Dimensions (2N +1)
N=Dimensionality of System;
while (N)
Calculate:
Z: transformed sigma points in measurements space;
z: mean in measurement space;
S: co-variance in measurement space;
Q: noise;
H: function that maps sigma points to measurement;
end while ()
if (Sigma points in state space and measurement)
Obtain the measurements from sensors;
Compute the weight update step;
end if ()
Stop
The RNN power system state estimation solver with minimized complexity is presented in
algorithm 3. Using the appropriate step sizes {μi} and η, the sequence {vi} generated and converges to a
sigma point illustrated in algorithm 2. A large number of inner loop iterations (K) are required to
approximate the solution. The quasi-reverse Bi has to be computed per outer loop iteration [26]. The next
outer loop iterates vi+1 can be obtained, once the inner optimum variable ui
*
is found with (9).
𝑣𝑖+1 = [𝐵𝑖(𝑢𝑖
∗
+ 𝑧) + 𝑣𝑖]/2 (9)
The equation (8) follows the definition of ui
*
and with respect to initialization, the u0
0
can be initialized as 0
without generality loss and ui
0
=ui-1
*
for i>1.
Algorithm 3. Minimized complexity recurrent NN
Start
Obtain the matrices (Hm)of the physical microgrid (zm)
zm=vT > Hmv+ ꞓm,∀m=1,...,M
Input: Data {(zm , Hm) M }m=1 , step size {µi }, ηi and initialization v0 = 1and u0
0 =0;

 ISSN: 2088-8708
4294
for i=0,1……..,I do
Initialize ui
0 ;
for k = 0,1,……., K do
update ui
k+1 ;
end for
update vi+1
end for
Stop
4. RESULTS OBTAINED
Microgrids, stand-alone distributed systems are required to shift the present power system towards
renewable sources of energy. IoT is used to communicate the information retrieved from sensors and is used
in smart grids for smart connectivity. Its use in smart grids for energy management systems extends the
benefits of a smart grid beyond automation, distribution, and monitoring [27].
The Figures 5 and 6 illustrate plot results of EKF, UKF and modified UKF for smart grids that
require state estimation for real-time supervisory control and monitoring. There are various techniques to
perform state estimation, some of which are discussed in this paper. The recursive systematic convolution
code is adopted from the communication point of view to introduce redundancy throughout the framework.
In the end, the Viterbi soft output decoder is used to retrieve device information from noisy measurements
and fluctuations of transmission. The Kalman filter is then used to estimate machine states, which serves as a
reference for the control algorithm to be implemented [28], [29].
Figure 5. Plot results of EKF and UKF
Figure 6. Plot results of UKF and modified UKF
Time (Second) Time (Second)

4295
We generate real states and measurements for this model as depicted in Figure 7 which presents of
estimation error and 3σ interval for UKF, EKF and BS error. The first 100 values of the filtering results with
UKF, EKF and bootstrap filter (BS) are calculated and obtained filtering results for UKF are as depicted in
Figure 8. It is observed that BS gives clearly better performance than UKF and EKF. The comparison of
estimation error and 3σ interval for UKF, extended UKF, modified EKF and BS error are as illustrated in
Figures 9 and 10 respectively. Since no source is ideal, with the increase in the number of generators i.e., by
decentralization the maintenance of constant voltage and frequency is a challenge.
Figure 7. Comparison of estimation error and 3σ interval for UKF, EKF, and BS error
Figure 8. Comparison of real signal and UKF model filtered estimate
Power quality is mainly affected by the existence of transient voltages, irregular changes in load,
harmonics, voltage sag, and swells. Maintaining the sinusoidal waveform could also be a challenge due to the
same reasons. In simple words, power quality is the ability of the system (grid) to supply constant voltage
and frequency with sinusoidal waveform, although it is impossible to eliminate the variations in power grids,
they can, however, be maintained under tolerance limits.
A discrete-time state space equation models the microgrid. Then, in order to obtain the appropriate
system details, cost-effective smart sensors are deployed. The recursive systematic convolution code is

 ISSN: 2088-8708
4296
adopted from the communication point of view to introduce redundancy throughout the framework. The
Table 3 illustrates mean square error obtained for several variants of Kalman filter. It is noteworthy that
extended Kalman filter has got a high mean square error and bootstrap Kalman Filter has got a least mean
square error. Derived from Table 2, Figures 11(a) and 11(b) depicts the estimation error estimation and
angles in the microgrid sensor voltage magnitudes from test instances 1000-1050 for IEEE-118 bus 50 and
bus 80 respectively.
Figure 9. Comparison of estimation error and 3σ interval for UKF 1 and extended UKF 2 error
Figure 10. Comparison of estimation error and 3σ interval for UKF, modified EKF and BS error
The Table 4 as depicted compares the existing results in terms of computational efficiency, energy
consumption, switching probability, misclassification error, and the throughput in the previous works carried
out with our proposed work and also the % improvement or the % enhancement in the obtained results are as
described in tabular column. Constant monitoring and prediction of states in grids will be a boon to
maintaining good power quality. Power quality can also be improved using PWM active filters, power
conditioning equipment like passive shunt L-C Filters, and by correcting power factor.

4297
Table 3. Tabular column of mean square error (MSE) obtained
Kalman filter variant Mean square error
UKF1-MSE 36.8932
UKF2-MSE 28.6441
Extended KF-MSE 75.1919
PKF-MSE 67.9096
BS-MSE 8.1377
CKF-MSE 60.6998
Ensembled KF-MSE 63.760
(a)
(b)
Figure 11. Microgrid sensor voltage magnitudes in angles and estimation error of IEEE-118 bus system from
test instances 1000 to 1050 for (a) bus 50 and (b) bus 80
Further using Table 3, Figures 12(a) and 12(b) are obtained and presents the estimation error
estimation and angles in the microgrid sensor voltage magnitudes from test instances 1,000-1,050 for
IEEE-118 bus 100 and bus 105 respectively. The forecasting performance was evaluated with RNN based
power system state forecasting scheme in terms of normalized root mean square error (RMSE) as described
in Table 2 with |v’-v|2 /N of the forecast v’ related to v (ground truth) [29]. The deep RNN and the activation
functions were trained and tested on ground truth voltage time series. Here as illuminated in Figure 13 the

 ISSN: 2088-8708
4298
microgrid sensor voltage magnitudes in angles and estimation error of IEEE-118 Bus system is calculated for
the bus index at time 200. Further the Figure 14 represents microgrid sensor voltage magnitudes in angles
and estimation error of IEEE-118 Bus system is calculated for the bus number at test instance of 1,000.
Further the Table 5 illuminates the parameters of the simulation in terms of utility energy (PU), battery
parameters (maximum charging rate, depth of the discharge, stored energy (PB) and the algorithm parameters
such as convergence condition and efficiency of the algorithm is presented.
Table 4. Comparison of existing results with proposed work with % improvement
Parameter Previous Research Proposed Research % Enhancement
[30] [31] [32] [33]
Computational Efficiency 0.96 0.92 0.86 0.82 0.8 0.6%
Energy consumption 12 J 15 J 13.2 J 12.4 J 14.7 J 11.5%
Switching probability 0.20 - - 0.21 0.27 0.2%
Misclassification error (%) 2.3 2.33 2.27 2.2 2.4 0.5%
Throughput (Mp/s) 124.3 10 - 100 112.5 5.6%
(a)
(b)
Figure 12. Microgrid sensor voltage magnitudes in angles and estimation error of IEEE-118 bus system from
test instances 1,000-1,050 for (a) bus 100 and (b) bus 105

4299
Figure 13. Microgrid sensor voltage magnitudes in angles and estimation error of IEEE-118 bus system for
the bus index at time 200
Figure 14. microgrid sensor voltage magnitudes in angles and estimation error of IEEE-118 Bus system for
the bus number at test instance of 1,000
Table 5. Simulation parameters
Serial Number Parameters
I PU (Utility Energy) PU
1
8.9 J
PU
2
13.2 J
PU
3
15 J
II BATTERY Maximum charging rate/maximum discharging rate 5.28 kW h
Maximum charging rate 0.526
Depth of Discharge 36%
PB (stored energy) 12 J
III Algorithm Parameters γ 0.96
Convergence condition (µ) 0.002
5. CONCLUSION
This paper investigates the various ways of distributed system state estimation via Kalman filters
and its variants. The work focuses on detailing the working of Kalman filter with different works related to it
discussed briefly. Comparisons on the variants of Kalman filter have been carried out and optimal filters for
various settings are being advised based on their discussed mathematical modelling. With more optimization

 ISSN: 2088-8708
4300
of the performing index, and appreciable flexibility factor by variants, Kalman filter are hence recommended
to improve the state estimation of the distributed microgrids. This paper presented a formulated approach is
presented to enable accurate measurements at component and system level model analysis in an IoT enabled
microgrid. This paper described several Kalman variants used for pre-processor of raw, redundant data and
produces a reliable state estimate and further investigates the various approaches of distributed system state
estimation via Kalman filters and its variants. In this research a formulated approach along with algorithms is
presented for optimal state estimation and forecasting, with weights update using DNN. The real load data
experiments are carried out on the IEEE 118-bus benchmark system for the power system state estimation
and forecasting and to enable calculations at component and system level model analysis in an IoT enabled
microgrid. The future scope of this research is to develop a novel DNN based algorithms for a power system
under dynamically varying conditions and corresponding time dependencies.
REFERENCES
[1] J. M. C. Clark, R. B. Vinter, and M. M. Yaqoob, “Shifted Rayleigh filter: a new algorithm for bearings-only tracking,” IEEE
Transactions on Aerospace and Electronic Systems, vol. 43, no. 4, pp. 1373–1384, Oct. 2007, doi: 10.1109/TAES.2007.4441745.
[2] T. Malkamaki and S. J. Ovaska, “Optimal state estimation for improved power measurements and model verification: Theory,” in
2011 International Green Computing Conference and Workshops, Jul. 2011, pp. 1–6, doi: 10.1109/IGCC.2011.6008597.
[3] A. Sharma, S. C. Srivastava, and S. Chakrabarti, “Testing and validation of power system dynamic state estimators using real time
digital simulator (RTDS),” IEEE Transactions on Power Systems, vol. 31, no. 3, pp. 2338–2347, May 2016, doi:
10.1109/TPWRS.2015.2453482.
[4] M. M. Rana, L. Li, and S. W. Su, “An adaptive-then-combine dynamic state estimation considering renewable generations in
smart grids,” IEEE Journal on Selected Areas in Communications, vol. 34, no. 12, pp. 3954–3961, Dec. 2016, doi:
10.1109/JSAC.2016.2611963.
[5] W. Zhao et al., “Consortium blockchain-based microgrid market transaction research,” Energies, vol. 12, no. 20, Oct. 2019, doi:
10.3390/en12203812.
[6] M. M. Rana, L. Li, and S. W. Su, “Distributed dynamic state estimation over a lossy communication network with an application
to smart grids,” in IEEE 55th Conf. on Decision and Control (CDC), 2016, pp. 6657–6662, doi: 10.1109/CDC.2016.7799294.
[7] S. Bhattacharjee and C. Nandi, “Implementation of industrial internet of things in the renewable energy sector,” in The Internet of
Things in the Industrial Sector, Springer International Publishing, 2019, pp. 223–259.
[8] P. Yang, Z. Tan, A. Wiesel, and A. Nehorai, “Power system state estimation using PMUs with imperfect synchronization,” IEEE
Transactions on Power Systems, vol. 28, no. 4, pp. 4162–4172, Nov. 2013, doi: 10.1109/TPWRS.2013.2272220.
[9] M. M. Rana, L. Li, and S. W. Su, “Microgrid state estimation and control using Kalman filter and semidefinite programming
technique,” International Energy Journal, vol. 16, no. 2, pp. 47–56, 2016.
[10] P. Nayak and S. Jena, “Estimation of harmonics in microgrid using unscented kalman filter,” International Journal of Pure and
Applied Mathematics, vol. 114, no. 9, pp. 73–81, 2017.
[11] A. Salian, S. Shah, J. Shah, and K. Samdani, “Review of blockchain enabled decentralized energy trading mechanisms,” in IEEE
Int. Conf. on System, Computation, Automation and Networking (ICSCAN), 2019, pp. 1–7, doi: 10.1109/ICSCAN.2019.8878731.
[12] M. Rana, L. Li, and S. Su, “Kalman filter based microgrid state estimation and control using the IoT with 5G networks,” in IEEE
PES Asia-Pacific Power and Energy Engineering Conference (APPEEC), 2015, pp. 1–5, doi: 10.1109/APPEEC.2015.7380989.
[13] M. E. Farmer and M. Allison, “Power demand prediction in smart microgrids using interacting multiple model Kalman filtering,”
in Proceedings of the Workshop on Communications, Computation and Control for Resilient Smart Energy Systems, Jun. 2016,
pp. 1–8, doi: 10.1145/2939940.2939947.
[14] F. Auger, M. Hilairet, J. M. Guerrero, E. Monmasson, T. Orlowska-Kowalska, and S. Katsura, “Industrial applications of the
kalman filter: a review,” IEEE Transactions on Industrial Electronics, vol. 60, no. 12, pp. 5458–5471, Dec. 2013, doi:
10.1109/TIE.2012.2236994.
[15] J. Liu, A. Benigni, D. Obradovic, S. Hirche, and A. Monti, “State estimation and learning of unknown branch current flows using
decentralized kalman filter with virtual disturbance model,” in 2010 IEEE International Workshop on Applied Measurements for
Power Systems, Sep. 2010, pp. 31–36, doi: 10.1109/AMPS.2010.5609327.
[16] P. Nayak, J. M. Pattnaik, S. Samantaray, and S. R. Singh, “Comparative study of harmonics estimation in micro grid using
adaptive extended Kalman filter,” in 2016 International Conference on Electrical, Electronics, and Optimization Techniques
(ICEEOT), Mar. 2016, pp. 434–438, doi: 10.1109/ICEEOT.2016.7755527.
[17] M. Rana, L. Li, and S. Su, “Kalman filter based distributed state estimation with communication systems,” in 2015 IEEE PES
Asia-Pacific Power and Energy Engineering Conference (APPEEC), Nov. 2015, pp. 1–5, doi: 10.1109/APPEEC.2015.7380988.
[18] M. M. Rana and L. Li, “Kalman filter based microgrid state estimation using the internet of things communication network,” in
12th Int. Conf. on Information Technology-New Generations, 2015, pp. 501–505, doi: 10.1109/ITNG.2015.86.
[19] P. Nayak and B. N. Sahu, “A robust extended Kalman filter for the estimation of time varying power system harmonics in noise,”
in 2015 IEEE Power, Communication and Information Technology Conference (PCITC), Oct. 2015, pp. 635–640, doi:
10.1109/PCITC.2015.7438074.
[20] J. Huang, V. Gupta, and Y.-F. Huang, “Electric grid state estimators for distribution systems with microgrids,” in 2012 46th
Annual Conference on Information Sciences and Systems (CISS), Mar. 2012, pp. 1–6, doi: 10.1109/CISS.2012.6310739.
[21] M. Rana, L. Li, and S. Su, “Distributed microgrid state estimation using smart grid communications,” in 2015 IEEE PES Asia-
Pacific Power and Energy Engineering Conference (APPEEC), Nov. 2015, pp. 1–5, doi: 10.1109/APPEEC.2015.7380942.
[22] M. Rana and L. Li, “An overview of distributed microgrid state estimation and control for smart grids,” Sensors, vol. 15, no. 2,
pp. 4302–4325, Feb. 2015, doi: 10.3390/s150204302.
[23] F. Shabani, M. Seyedyazdi, M. Vaziri, M. Zarghami, and S. Vadhva, “State estimation of a distribution system using WLS and
EKF techniques,” in IEEE Int. Conf. on Information Reuse and Integration, 2015, pp. 609–613, doi: 10.1109/IRI.2015.101.
[24] M. M. Rana and L. Li, “Renewable microgrid state estimation using the internet of things communication network,” in 19th Int.
Conf. on Advanced Communication Technology (ICACT), 2017, pp. 823–829, doi: 10.23919/ICACT.2017.7890232.
[25] S. Saxena, H. Farag, A. Brookson, H. Turesson, and H. Kim, “Design and field implementation of blockchain based renewable

4301
energy trading in residential communities,” in 2019 2nd International Conference on Smart Grid and Renewable Energy (SGRE),
Nov. 2019, pp. 1–6, doi: 10.1109/SGRE46976.2019.9020672.
[26] F. B. Armani, F. Grimaccia, S. Leva, and M. Mussetta, “Seamless Grid: an off-chain model proposal for scalable P2P electricity
markets and grids management,” in 2019 IEEE Milan PowerTech, Jun. 2019, pp. 1–6, doi: 10.1109/PTC.2019.8810523.
[27] S. Baldi, S. Yuan, P. Endel, and O. Holub, “Dual estimation: Constructing building energy models from data sampled at low
rate,” Applied Energy, vol. 169, pp. 81–92, May 2016, doi: 10.1016/j.apenergy.2016.02.019.
[28] F. O. Igbinovia and J. Krupka, “The prospect of the internet of renewable energy (IoRE) in electricity networks,” in 2019 IEEE
International Symposium on Technology and Society (ISTAS), Nov. 2019, pp. 1–6, doi: 10.1109/ISTAS48451.2019.8937918.
[29] J. Li et al., “Decentralized on-demand energy supply for blockchain in internet of things: A microgrids approach,” IEEE
Transactions on Computational Social Systems, vol. 6, no. 6, pp. 1395–1406, Dec. 2019, doi: 10.1109/TCSS.2019.2917335.
[30] M. Onyeka Okoye et al., “A blockchain-enhanced transaction model for microgrid energy trading,” IEEE Access, vol. 8,
pp. 143777–143786, 2020, doi: 10.1109/ACCESS.2020.3012389.
[31] Y. Kim and K. Moon, “Implementation of secure system for blockchain-based smart meter aggregation,” The Journal of the
Institute of Internet, Broadcasting and Communication, vol. 20, no. 2, pp. 1–11, 2020.
[32] E. S. Kang, S. J. Pee, J. G. Song, and J. W. Jang, “A blockchain-based energy trading platform for smart homes in a microgrid,”
in 2018 3rd International Conference on Computer and Communication Systems (ICCCS), Apr. 2018, pp. 472–476, doi:
10.1109/CCOMS.2018.8463317.
[33] S. Tan, X. Wang, and C. Jiang, “Privacy-preserving energy scheduling for ESCOs based on energy blockchain network,”
Energies, vol. 12, no. 8, Apr. 2019, doi: 10.3390/en12081530.
BIOGRAPHIES OF AUTHORS
Sudhakar Rao Padupanambur received the B.E. and M. Tech. degree in
Electrical and Electronics Engineering from Mysore University, Mysore, India, 1984 and 1993
respectively. He is pursuing Ph.D. degree in Electrical and Electronics Engineering at REVA
University, Bengaluru, India. From 1987 till 2006, he was with Nitte Mahalinga Adyanthaya
Institute of Technology, Nitte, India as an Assistant Professor in department of Electrical and
Electronics Engineering. Since 2006, he has been with REVA University, Bengaluru, India,
where he is currently an Associate Professor in School of Electrical and Electronics
Engineering. His research interests include power system, microgrid, and IoT. He is having
the experience of 34 years in teaching and research. He can be contacted at email:
sudhakar.rao@reva.edu.in.
Mohammed Riyaz Ahmed received the B.E. and M. Tech. degree in electronics
and communication engineering and computer networking from Visvesvaraya Technological
University, Belgaum, India, 2007 and 2010, respectively. He received the Ph.D. degree in
electronics and communication engineering from the Jain University, Bengaluru, India, in
2016. Since 2011, he has been with REVA University, Bengaluru, India, where he is currently
an Associate Professor and Assistant Director of School of Multidisciplinary Studies. He is
advisor of IEEE EMBS student chapter and mentor to IEEE ComSoc student chapter at REVA
University. He has served as TPC member of IEEE ICAECC conference, and technical
reviewer for 7 IEEE journals and transactions along with numerous achieved journals and
conference proceedings. His research interests include the computational cognitive
neuroscience, WSNs, 5G and beyond, Genomic data sequencing, tribology, neuromorphic
engineering, spintronics, IoT and edge computing, and green technologies. He is involved in
investigation and teaching of Design thinking as an enabler for individual, institutional and
international sustainable development. He has published more than 100 research papers in
refereed journals, international conferences and as invited speaker. He is also the founder and
PI of CPS lab at REVA University. Riyaz is the advisor and the consultant to a number of
Higher Educational Institutes for their transformation into an Entrepreneurial University for
Knowledge Economy. He can be contacted at email: riyaz@reva.edu.in.
Bangalore Prabhakar Divakar completed his Ph.D. from the Hong Kong
Polytechnic University in 1998 and then continued to work in the same university as research
associate, research fellow and faculty from 1998 to 2009 working on various fields related to
fuel cells, Battery management system, HID and LED lighting and SMES. He joined REVA
as a professor in the school of EEE in 2009 and now is the Dean, Research and Innovation
Council from 2014. He has published articles in IEEE transactions and has 2 USA patents to
his credits. He has completed one research project granted from VTU research funds. He is the
recipient of the Meritorious Service Award (Team project, The Hong Kong Polytechnic
University, HK, 2009) and Award of Excellence for Teaching, REVA ITM, Sept -2014. He is
an International committee member for PESA conference organized at Power Electronics
research center, Hong Kong and International member of IEEE Transportation Electrification
Committee-IEEE TEC. He started innovation club in the school of EEE attracting
students across schools to participate in collaborative projects. He can be contacted at email:
divakar@reva.edu.in.

Optimal state estimation techniques for accurate measurements in internet of things enabled microgrids using deep neural networks

More Related Content

Similar to Optimal state estimation techniques for accurate measurements in internet of things enabled microgrids using deep neural networks (20)

More from IJECEIAES (20)

Recently uploaded (20)

Optimal state estimation techniques for accurate measurements in internet of things enabled microgrids using deep neural networks