Power Management Controller for Microgrid Integration of ...

Research ArticlePower Management Controller for Microgrid Integration ofHybrid PV/Fuel Cell System Based on Artificial DeepNeural Network

Abdulbaset Abdulhamed Mohamed Nureddin , Javad Rahebi , and Adel Ab-BelKhair

Department of Electrical and Computer Engineering, Altinbas University, Turkey

Correspondence should be addressed to Javad Rahebi; [email protected]

Received 16 April 2020; Revised 27 August 2020; Accepted 24 November 2020; Published 8 December 2020

Academic Editor: Huiqing Wen

Copyright © 2020 Abdulbaset Abdulhamed Mohamed Nureddin et al. This is an open access article distributed under the CreativeCommons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided theoriginal work is properly cited.

Nowadays, the power demand is increasing day by day due to the growth of the population and industries. The conventional powerplant alone is incompetent to meet the consumer demand due to environmental concerns. In this present situation, the essentialthing is to be find an alternate way to meet the consumer demand. In present days most of the developed countries concentrateto develop alternative resources and invest huge money for its research and development activities. Most renewable energysources are naturally friendly sources such as wind, solar, fuel cell, and hydro/water sources. The results of power generationusing renewable energy sources only depend on the availability of the resources. The availability of renewable energy sourcesthroughout the day is variable due to fluctuations in the natural resources. This research work discusses two major renewableenergy power generating sources: photovoltaic (PV) cell and fuel cell. Both of them provide foundations for power generation,so they are very popular because of their impressive performance mechanisms. The mentioned renewable energy-based powergenerating systems are static devices, so the power losses are generally ignorable as compared to line losses in the main grid. ThePV and fuel cell (FC) power systems need a controller for maximum power generation during fluctuations in the inputresources. Based on the investigation report, an algorithm is proposed for an advanced maximum power point tracking (MPPT)controller. This paper proposes a deep neural network- (DNN-) based MPPT algorithm, which has been simulated usingMATLAB both for PV and for FC. The main purpose behind this paper has been to develop the latest DNN controller forimproving the output power quality that is generated using a hybrid PV and fuel cell system. After developing and simulatingthe proposed system, we performed the analysis in different possible operating conditions. Finally, we evaluated the simulationoutcomes based on IEEE 1547 and 519 standards to prove the system’s effectiveness.

1. Introduction

Right now, power generation and transmission require sub-stantially large quantities of fossil fuels. The fossil fuel-basedpower generation systems cannot meet the consumerdemand in future due to limited availability of fuel. Further,environmental considerations will limit their usage as theyemit more greenhouse gases. This leads to warning levelsand climate change, and the availability of fossil fuels isnow limited. Nowadays, many counties are concentratingon alternate sources, replacing fossil fuels with greener andecofriendly fuels. Presently, the combination of solar and fuel

cells is a promising renewable energy (RE) source [1–5]. Thepower, which is extracted out of a PVmodule, is substantiallyvulnerable to environmental factors, natural fluctuations,and differences in operating conditions [6–8]. It is a fact thatwhen the requirement for power is critical, solar energy failsto generate sustainable power. To fulfill the power demand,many processes and devices have been introduced. Forinstance, for understanding the energy demand in the peakperiods, such as at night, a hybrid system consisting of a fuelcell and a PV was developed without power storage. It has afew drawbacks such as the issue of integrating a microgrid,and besides, such a system is difficult to operate under

HindawiInternational Journal of PhotoenergyVolume 2020, Article ID 8896412, 21 pageshttps://doi.org/10.1155/2020/8896412

https://orcid.org/0000-0003-4145-882X

https://orcid.org/0000-0001-9875-4860

https://orcid.org/0000-0002-3302-7360

https://creativecommons.org/licenses/by/4.0/

https://creativecommons.org/licenses/by/4.0/

https://doi.org/10.1155/2020/8896412

unstable conditions or has irregular, unmanageable, stochas-tic, and highly variable sources. To be precise, researchersfocused on the following area such as high penetration levelto the microgrid or when there are issues such as poor loadfollowing, load discrepancy, frequency deviation, voltage var-iability, reliability problems, and poor power quality.

Based on the literature reports, various algorithms suchas the Whale Optimization Algorithm (WOA), Water CycleAlgorithm (WCA), Moth-Flame Optimizer (MFO), andHybrid Particle Swarm-Gravitational Search Algorithm(PSOGSA) have been both developed and applied for design-ing the optimized microgrid and analyzing power quality

Duty cycle

Vin

Vin Vout

VL

L

D

S

VL

Vout

DL

S+ +

MPPT- DNN

MPPT - DNNFuelcellcurrent

�ree-phaseinverter

PWM

PWM

Vabc

Grid

Iabc

Vpv & Ipv

�ree-phaseinverter

g

A

B

C

Voltage

source

converter

DNN

fontroller

+

–

g

A

B

C

+

+ +

–

Figure 1: Block diagram of the proposed microgrid system and its controller.

Irradiance

Temperature

MOSFET

D

L

CC Gate

PV

Duty cycle - PWM

VPV , IPV

R

DNN MPPTcontroller

Figure 2: Proposed DNN-based MPPT algorithm.

2 International Journal of Photoenergy

Ir

T

m

+

–

i+ –+

gDS

+

+

v+–

MPPT_Power

D P

PV_Power

××PVP

PVP

PVMPPT

PVMPPT

NNET Input Output

Function fitting neural network

+

1

PV Power

1

– Ve

+ Ve

2

Irr

Temp

Irradiance

PV Voltage

Figure 3: Simulation model of DNN-based MPPT algorithm for PV systems.

Input

2 1

Hidden

1500 1

Output

OutputW

b+

W

b+

Figure 4: DNN model for PV MPPT system.

TrainValidation

TestBest

Best validation performance is 4.7867e-11 at epoch 61

Mea

n sq

uare

d er

ror (

mse

)

61 Epochs

102

100

10–2

10–4

10–6

10–3

10–10

0 10 20 30 40 50 60

Figure 5: DNN-PV MPPT controller showing best validation performance.

3International Journal of Photoenergy

Error histogram with 20 bins

3000

2500

2000

1500

1000

500

0

Inst

ance

s

–1.7

e-05

–1.5

e-05

–1.3

e-05

–1.1

e-05

–8.7

e-06

–6.6

e-06

–4.6

e-06

–2.6

e-06

–5.5

e-07

1.49

e-06

3.52

e-06

5.56

e-06

7.59

e-06

9.63

e-06

1.17

e-05

1.37

e-05

1.57

e-05

1.78

e-05

1.98

e-05

2.18

e-05

Errors = Targets - Outputs

TrainingValidation

TestZero error

Figure 6: DNN-PV MPPT controller error histogram.

Gradient = 9.8252e-08, at epoch 61

Mu = 1e-09, at epoch 61

Validation Checks = 0, at epoch 61

100

10–5

10–5

10–10

0.5

–0.5

1

–1

0

0 10 20 30 40 50 60

61 Epochs

Gra

dien

tM

uVa

l fai

l

Figure 7: DNN-PV MPPT controller with training data.


issues [9]. Another important factor of the hybrid renewableenergy system is optimization of net present cost (NPC), fuelcost, operation cost, and cost of energy (COE) of the hybrid.The optimization of the key performance indicators of thehybrid power system is done by using the hybrid optimiza-tion model for electric renewables (HOMER) [10]. Theenergy flow is from the energy source to the electrical systemfor a residential application in conjunction with an intelligentdemand management control strategy [11]. The real andreactive power controller is challenging role in hybrid renew-able energy systems. An adaptive Lyapunov-based rapid ter-minal sliding mode control is superior over the traditional PIcontrol because of its faster error tracking capability androbustness. A Q-V-based inverter control is employed tointerface the single phase grid and the hybrid system [12].The optimal power flow (OPF) formulation includes theforecasted active power generation of WT and PV as depen-dent variables, whereas the voltage magnitude at WT and PVbuses is considered as control (decision) variables [13].

The hybrids of Renewable Energy Sources (RESs) dealwith optimal performance in cost consideration, emission,

Validation: R = 1

Target

Out

put~

= 1⁎

targ

et +

–1

.1e–

07

0.9

0.8

0.7

0.6

0.5

0.40.90.80.70.60.50.4

All: R = 1

0.80.60.4 1

Out

put~

= 1⁎

targ

et +

–1

.2e–

08

Training: R = 1

Target

0.9

0.8

0.7

0.6

0.5

0.40.90.80.70.60.50.4

Test: R = 1

Out

put~

= 1⁎

targ

et +

–3.

7e–0

9O

utpu

t~=

1⁎ta

rget

+ 4

.7e–

08 0.9

0.8

0.7

0.6

0.5

0.4

1

0.80.60.4 1

DataFitY = T

DataFitY = T

TargetTarget

DataFitY = T

DataFitY = T

0.9

0.8

0.7

0.6

0.5

0.4

1

Figure 8: DNN-PV MPPT controller showing best regression.

Table 1

(a) DNN data of MPPT controller for PV systems

DNN-FC MPPT Values

Best validation 4.7867e-11

Mu 1e-09

Gradient 9.8252e-08

Training 1

Validation 1

Test 1

Overall regression 1

(b) Comparative analyses of MPPT controller for PV systems

Controller Fuzzy DNN

PV 17820W 18080W


Irradiance

28

26

24

22

20

W/M

2

PV Power

2

1.5

1

0.5

0W

atts

Temp

1000800600400200

0

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

C

Time

MPPT_PowerTempI

(a)

MPPT_PowerPV_Power

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

3

2.5

2

1.5

0.5

0

1

DNN - MPPT

Time

W

Trace selection

MPPT_Power

Signal statistics

Value TimeMaxMin

MeanMedianRMS

Peak to peak

1.808e+044.267e–201.808e–201.634e+041.780e+041.677e+04

0.9950.000e+00

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

3

2

1.5

1

0.5

0

2.5

Fuzzy MPPT

W

Time

PV_PowerMPPT_Power

Trace selection

Signal Statistics

MaxValue Time

MinPeak to peakMeanMedianRMS

1.782e+04 0.9980.000e+00 4.267e–20

1.464e+041.729e+041.544e+04

MPPT_Power

1.782e+04

‣‣‣‣‣

(b)

Figure 9: (a) Simulation results of DNN-PV MPPT power with respect to irradiance and temperature. (b) Comparison of fuzzy- and DNN-based MPPT controller.


PV boost converter voltage and current

Volt

and

amp

500

400

300

200

100

0

Time

0.10 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Voltage measurementCurrent measurement

Voltage

Current

Voltage

CurrentC

Figure 10: Simulation results of DNN-PV MPPT boost converter voltage and current with respect to irradiance and temperature.

MOSFET

L

D

C R

Duty cycle-PWMMPPT

controllerFuel cellcurrent

Fuel cell

Figure 11: Proposed DNN-based optimum fuel flow algorithm for the fuel cell power system.

DNN Based MPPT controller for 6 kW fuel cell stack

v+– +

m

+

–

+

–

i+ –

– FC

+

–

Fuel cell power

×

NNET InputOutput


1

2– Ve

+ Ve

1

<Utilization (%) [O2 (Yellow); H2 (Magenta)]>

<Voltage>

<Current>

<Stack efficiency (%)>

Fuel flow rate (lpm)

DC bus current

<Stack consumption (lpm) [Air(Yellow); Fuel(Magenta)]>

DC bus voltage

+ FC

100Vdc BoostDC/DC Converter

(average value)

FuelFr

y_O2

H2 Air

Scope

Figure 12: Simulation model of the DNN controller-based optimum fuel flow algorithm for a fuel cell power system.


Error histogram with 20 bins

Inst

ance

s

Error=Tragets–Outputs

1500

1000

500

0

–0.

0004

4

–0.

0004

–0.

0003

6

–0.

0003

1

–0.

0002

7

–0.

0002

3

0.000123

0.000166

0.00021

0.000253

0.000297

0.000384

0.000341

–0.

0001

8

–0.

0001

4

–9.

5e–05

–5.

2e–05

–8.

1e–06

3.54

e–05

7.9e–05

TrainingValidation

TestZero error

Figure 15: Error histogram of the DNN-based fuel flow controller.

Best validation performance is 1.4367e–80 at epoch 5M

ean

squa

red

erro

r (m

se)

5 Epochs

0 0.5 1.5 2.5 3.53 4 4.5 52110–8

10–6

10–4

10–2

100

102

Train TestValidation Best

Figure 14: DNN-based fuel flow controller showing best validation performance.

Input

1 1

Hidden layer

1000 1

Output layer

OutputW

b+

W

b+

Figure 13: DNN model for fuel cell MPPT system.


space management and adequate reliability in power genera-tion, and power quality and optimum power utilization [14].

Hence, novel and innovative technologies based onfeasible ideas are needed to deal with the mentioned issuesso as to increase the RE sources’ penetration in the powergrid.

This paper has proposed a novel and deep neural net-work algorithm to solve the power system issues with inte-grated PV and FC generators. This system is represented inFigure 1. The main contributions of this work include thefollowing:

(1) Design and model the MPPT controller for PV andfuel cells using DNN and analyze its performanceunder various weather conditions

(2) Design and develop a hybrid PV-fuel cell energy sys-tem and model a smart inverter for microgrid inte-gration of a hybrid system

(3) Design a DNN-based voltage source controller that issynchronous with the microgrid and hybrid PV andfuel cell energy system

(4) The proposed DNN-based smart inverter helps toreduce harmonics and improves the power qualityand its stability

The organization of this paper is as follows: the DNN-based MPPT algorithm for a PV system is given in Section2. This section also analyzes the proposed algorithm’s per-formance under various weather conditions. The DNN-based MPPT algorithm for fuel cell energy system and per-formance analysis under various fuel pressures and itsresults are explained in Section 3. The proposed DNN-based microgrid-integrated hybrid PV and fuel cell powersystem are explained in Section 4. In Section 5, the DNN-based smart inverter controller is modelled, and its perfor-mance is explained. The results for simulation are given inSection 6 while Section 7 includes the final conclusions.

2. Deep Neural Network-Based MPPT for PV

In the present situation, renewable energy sources arerequired to make the maximum power point tracking algo-rithm generate maximum power under various weather con-ditions [15]. The researchers have been focused to createdifferent MPPT algorithms, including the incremental con-ductance, P&O, feedback voltage and current, fuzzy, ANN,PSO ANFIS, and other controllers [16–18]. This paperthrows light on a new advanced MPPT algorithm called aDeep Neural Network (DNN) controller for photovoltaicsystems. The proposed DNN-based MPPT algorithms were

Gradient = 2.9948e–08, at epoch 5

Mu = 1e–08, at epoch 5

Validation checks = 0, at epoch 5

5 Epochs

54 4.53.532.521.510.50

1

0.5

0

–0.5

–1

Val f

ail

Val f

ail

10–2

10–4

10–6

10–8

10–10

10–5

100G

radi

ent

Gra

dien

tM

uM

u

Figure 16: Training data of the DNN-based fuel flow controller.


developed using 75,000 data entries, such as PV voltage, cur-rent, and its corresponding duty cycle, as Figure 2 shows. TheMPPT algorithm of a PV system simulation model has beendeveloped and modelled with MATLAB, illustrated inFigure 3. The proposed simulation was performed using a50 kWmodel PV array, which is connected with a boost con-verter, and it is controlled by the proposed MPPT algorithm.The developed DNN algorithm has 2 inputs, including a PVcurrent and voltage, and besides, 1500 hidden layer neuronsare used between the input and output layers, as Figure 4indicates. The proposed MPPT algorithm’s output generatesa duty cycle for a PV-connected boost converter. This DNNlayer has been used for training the MPPT network usingthe data, such as input (PV voltage, PV current) and outputin the form of a duty cycle of a converter.

The best validation performance of the proposed DNNcontroller is presented in Figure 5. The following details areobserved while training the DNN for the MPPT controllersuch as the error (targeted output) histogram data of the pro-posed DNN controller being very small, which is obvious inFigure 6. The gradient and validation checks of the proposedDNN trained network are presented in Figure 7. Finally, the

Training: R=1

Target

Out

put ~

=1⁎

targ

et +

-5.3

e-08

1

0.9

0.8

0.7

0.6

0.5

0.5 0.6 0.7 0.8 0.9 1

DataFitY = T

Validation: R=1

Target

Out

put ~

=1⁎

targ

et +

-6.5

e-07

1

0.9

0.8

0.7

0.6

0.5

0.5 0.6 0.7 0.8 0.9 1

DataFitY = T

DataFitY = T

Test: R=1

DataFitY = T

Target

Out

put ~

=1⁎

targ

et +

-9 .3

e-08

1

0.9

0.8

0.7

0.6

0.5

All: R=1

Target

Out

put ~

=1⁎

targ

et +

-6.5

e-07

1

0.9

0.8

0.7

0.6

0.5

0.5 0.6 0.7 0.8 0.9 10.5 0.6 0.7 0.8 0.9 1

Figure 17: Training and validation of the DNN-based fuel flow controller.

Table 2

(a) DNN data of the MPPT controller for the FC system

DNN-fuel cell MPPT Values


Mu 1.00e-8

Gradient 2.9948e-8

Training 1

Validation 1

Test 1


(b) Fuel cell efficiency comparison: fuzzy and DNN

Controller Fuzzy DNN

Efficiency 57.01% 57.09%


Volt

& am

p

160

140

120

100

80

60

40

20

0

0 1 2 3 5 6 7 8 9 104

Time

<Voltage>,<current>,DC bus voltage, DC bus current

<Voltage><Current>

DC bus voltageDC bus current

Figure 19: Simulation results for DNN-fuel cell converter voltage and current as well as DC bus voltage and current.

600050004000300020001000

0

Wat

t

Fuel cell output power

Time

4948.848.648.448.2

48

0 1 2 3 4 5 6 7 8 9 10

Ipm

Fuel flow rate (Ipm)

Figure 18: Simulation results for the DNN-fuel cell power with respect to optimum fuel flow.

Fuel cell voltage

Fuel cell current

140120100

80604020

0

605040302010

0

Volt

Time

Am

p

0 1 2 3 4 5 6 7 8 9 10

Figure 20: Fuel cell voltage and current waveform.


proposed DNN-based MPPT algorithm has been developedand is presented with its best regression for training data, testdata, validation data, and overall performance data inFigure 8. The overall DNN data are presented in Table 1.

The proposed 50 kW PV system has been simulated inMATLAB, and the proposed DNN-based MPPT algorithmwas applied. The proposed simulation model has been simu-lated under various weather conditions, and its performance

was analyzed in various operating conditions. The 50 kW PVsystem’s simulation results are illustrated in Figure 9(a),which also shows various solar irradiance values with respectto the corresponding PV output power and MPPTcontroller-based PV output power. The comparative analysesof the PV MPPT with fuzzy and DNN is presented inFigure 9(b) and Table 1(b); the boost converter’s currentwaveform and output voltage are illustrated in Figure 10.

Power distribution grid

�ree-phase source

�ree-phase transformer

Power distribution grid

e

�ree-phase transformer

Figure 21: The proposed simulation model of a DNN-based controller with a microgrid integrated with a hybrid PV and a FC energy system.

Current regulator PWM generatorPWM

Vd & Vq conv

Id

Vdc ref

VDC RegulatorVdc Measure

Id & Iq

Vd & Vq

Wt

PLL & measurementVabc & Iabc

Figure 22: DNN controller-based voltage source controller for smart inverters.


3. Deep Neural Network-BasedMPPT for PowerSystem Using a Fuel Cell

The DNN controller develops rules and limits to assureoptimum fuel flow for a fuel cell system, as shown inFigure 11. For the DNN learning rule, 44,000 entries of datawere used to develop the optimum fuel flow rule. The sim-ulation model of the projected algorithm-based fuel cellsystem has been simulated using MATLAB, as Figure 12illustrates. A 6 kW fuel cell power system was simulated,and a boost convertor was employed for stepping upvoltage-supported optimized fuel flows [19]. The controlof fuel flow will lead to the fuel cell’s current limitation as

well reduce hemic losses in the fuel cell, which helps toincrease fuel cell operating efficiency. In Figure 13, theDNN layer has been illustrated. This layer has a singleinput, which is the fuel cell current, and there are a thou-sand hidden layers between the input and the output. Itonly has a single output, which is the fuel flow pressure.

The best validation performance of the proposed DNNcontroller is presented in Figure 14. The following infor-mation is observed and noted such as the error(target − output) histogram data of the proposed DNNcontroller being very small, as shown in Figure 15. Thegradient and validation checks of the proposed DNN-trained network are presented in Figure 16. Finally, the

1

3

2

1

VdVq_mes

Feed forward

IdIq_mes

IdIq_ref


Input NNET Output

Rtot_pu1

0.002

0.002

0.06

0.06

Ltot_pu1

Rtot_pu5

Ltot_pu2

Vd_mes + Id⁎R - Iq⁎L + deriv(Id)⁎L = Vd_convVq_mes + Id⁎L + Iq⁎R + deriv(Iq)⁎L = Vq_conv

Sign convention: current going out of the converter = positive currentId positive --> the converter generates active power ("invertermode") = active power P positiveIq positive --> the converter absorbs reactive power ("inductive mode") = reactive power Q negative

VdVq_conv

PI_Ireg1

Current regulator(with feedforward)

Ltot = Lxfo + LchokeRtot = Rxfo + Rchoke

Harmonic filter neglectedFor Ltot >> Rtot :

(Vd_prim - Vd_conv) ~ = +Iq⁎Ltot(Vq_prim - Vq_conv) ~ = –Id⁎Ltot

Figure 23: DNN-based current regulator for smart inverter.

Input

2 2

Hidden

2500 2

Output

OutputW

b+

W

b+

Figure 24: DNN model for the current regulator of a smart inverter.


1110986 754310 210–10

10–8

10–6

10–4

10–2

100

102

11 Epochs

Mea

n sq

uare

d er

ror (

mse

)

Best validation performance is 7.7813e–10 at epoch 11

TrainValidation

TestBest

Figure 25: DNN-current regulator showing best validation performance.

TrainingValidation

TestZero error

Error histogram with 20 Bins

–0.0

0024

–0.0

0022

–0.0

0019

–0.0

0017

–0.0

0015

–0.0

0013

–0.0

0011

–9.5

e–05

–7.6

e–05

–5.8

e–05

–4e–

05

–2.2

e–05

–3.8

e–06

–1.4

4e–0

5

3.26

e–05

5.07

e–05

6.89

e–05

8.71

e–05

0.00

0105

–0.0

002

Errors = Targets – Outputs

Inst

ance

s

0.5

1.5

2.5

2

1

0

Figure 26: Error histogram of a DNN-based current regulator.


proposed DNN-based fuel flow control algorithm has beendeveloped that showed best regression for training data,test data, validation data, and the overall performancedata, as seen in Figure 17. The overall design and DNNdata are presented in Table 2.

We applied the proposed DNN-based MPPT algorithmto a 6 kW fuel cell power system and simulated it inMATLAB environment. The proposed fuel flow controllersystem has been analyzed under various fuel pressures.Figure 18 illustrates a power system that uses a fuel cell togenerate power at various fuel flow pressures. The currentwaveform and output voltage of a boost converter are shownin Figure 19, and for the fuel cell, they are presented inFigure 20. Table 2(b) shows the comparative analyses of fuelcell efficiency with fuzzy and DNN controller.

4. Integration of Microgrid with Hybrid PV andFuel Cell Power Systems

In this section, we will discuss how a microgrid is integratedwith a hybrid power system, its components like fuel cells orPVs, and how it is controlled by a DNN-based MPPT algo-rithm. The detailed simulation model is presented inFigure 21. In this simulation model, a 74 kW hybrid PV(50 kW) and a fuel cell (6 × 4 = 24 kW) energy system areintegrated into a power microgrid with the support of a smartinverter, which is controlled by a DNN-based voltage sourcecontroller [15, 20, 21] as shown in Figure 22. In this control-

ler, there are three major subcontrollers, which are designedincluding a voltage regulator, a phase lock loop, and a currentregulator. Finally, the PWM signals will be generated throughthe current regulator for the smart inverter through a syn-chronous connection between the hybrid 74 kW PV/fuel celland the microgrid system.

5. Deep Neural Network Controller-Based Grid-Integrated Smart Inverter

A Deep Neural Network (DNN) controller can regulate thecurrent in a grid-connected hybrid PV and FC energy sys-tem with the help of an exploitation sensing inverter. Duringthis DNN learning, the algorithmic rule is used with 75,000data items for providing training to develop roles to run acurrent regulator algorithm. The simulation model of theprojected current regulator-based grid-connected smartinverter has been simulated using MATLAB, as Figure 23indicates.

The projected simulation model, the 74 kW hybrid PV(50 kW), and the fuel cell (6 × 4 = 24 kW) energy system havebeen used for microgrid integration. The DNN layer thatconsists of a pair of input neurons, 2500 hidden-layer neu-rons, and two output neurons is shown in Figure 24. ThisDNN layer has been used for guiding the current regulatornetwork used in the subsequent data like input (direct axisand quadrant axis current) while the output is a PWM signalfor a grid-integrated smart inverter.

Gradient = 7.8828e–05, at epoch 11

Mu = 1e–11, at epoch 11

Validation checks = 0, at epoch 11

11 Epochs

11109876543210

1

0.5

0

–0.5

–1

Val f

ail

10–10

10–5

10–5

100G

radi

ent

Mu

Figure 27: Training data of a DNN-based current regulator.


The DNN controller has been trained using the abovesystem, and its best validation performance has been pre-sented in Figure 25. The proposed DNN controller error(target − output) histogram data is presented in Figure 26.The proposed DNN-trained network gradient and validationcheck is presented in Figure 27. Finally, the proposed DNN-based current regulator algorithm has been developed and ispresented with its best regression for training data, test data,validation data, and overall performance data in Figure 28.The regression predicts an output variable (PWM) as a func-tion of the inputs (observed direct and quadrant axis cur-rent). The overall DNN data are given in Table 3.

6. Results and Discussion

The proposed system has been developed and simulatedunder varying operational conditions using the MATLABatmosphere. The simulation model, the simulation results,and the subsequent information were determined. Figure 29shows the load of real power consumed by the consumersfrom the microgrid network, which was 110 kW. The reac-

tive power is consumed by the consumer’s load from themicrogrid, which is approximately 170 kVR, as shown inFigure 30. Figure 31 depicts that the full power (74 kilo-watts) was generated by combining a PV and a fuel cellin a hybrid power generation system. The most importantobjective of this analysis is to maintain the microgrid gridvoltage and current with no or negligible oscillations.Figure 32 shows the microgrid voltage and current

Training: R = 1

DataFitY = T

Target

1.5 1.6 1.7 1.8 1.9 2

Validation: R = 1

1.9

1.8

1.7

1.6

1.5

2

Out

put ~

=1⁎ta

rget

+ –

5.6e

–05

DataFitY = T

Target

1.9

1.8

1.7

1.6

1.5

1.5 1.6 1.7 1.8 1.9 2

2

Out

put ~

=1⁎ta

rget

+ –

5.1e

–05

Test: R = 1

DataFitY = T

Target

1.5 1.6 1.7 1.8 1.9 2

All: R = 1

1.9

1.8

1.7

1.6

1.5

2

Out

put ~

=1⁎ta

rget

+ –

6.3e

–05

DataFitY = T

Target

1.9

1.8

1.7

1.6

1.5

1.5 1.6 1.7 1.8 1.9 2

2

Out

put ~

=1⁎ta

rget

+ –

5.6e

–05

Figure 28: DNN-based current regulator—training, test, and validation.

Table 3: DNN data of a controller for the grid system.

DNN grid Values


Mu 1e-11

Gradient 7.8828e-5

Training 1

Validation 1

Test 1



0 0.1 0.2

10

15

5

00.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Wat

t

×104 Load – active power

Time

Figure 29: Real power consumed by load.

2

1.5

1

0.5

021.510.50 2.5

Time

Load - reactive power

VAR

×105

Figure 30: Reactive power consumed by the load.

54

3210

PV

Wat

t

×104

FC2.5

21.5

10.5

0

Wat

t

×104

Wat

t

6

4

2

0

0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

×104

Time

Hybrid PV + FC

Figure 31: Total power generated by hybrid PV and FC energy system.


0

400300200100

–100–200–300–400

400600

2000

–400–200

–600

Load current

Volta

ge (v

olt)

Power system distribution voltage

0 0.05 0.1 0.15

Time

Curr

ent (

amp)

Figure 32: Waveform of the microgrid voltage and current.

0 0.5 1 1.5 2 2.5

Time

–2

0

2

4

6

8

10

12

Wat

ts

Grid power supplied to load

Figure 33: Grid-supplied power to the load.

0 200 400 600 800 1000

Frequency (Hz)

0

0.5

1

1.5

2

2.5

3

3.5

Mag

(% o

f fun

dam

enta

l)

Fundamental (50 Hz) = 362, THD = 3.55%

Figure 34: THD values for PV system voltage.


0 200 400 600 800 1000

Frequency (Hz)

0

0.2

0.4

0.6

0.8

Mag

(% o

f fun

dam

enta

l)

Fundamental (50 Hz) = 8.345, THD = 0.90%

Figure 35: THD values for PV system current.

0 200 400 600 800 1000

Frequency (Hz)

0

0.5

1

1.5

2

2.5

3

3.5

Mag

(% o

f fun

dam

enta

l)


Figure 36: THD values for FC voltage.

0 200 400 600 800 10000

2

4

6

8

10

12

Frequency (Hz)

Mag

(% o

f fun

dam

enta

l)


Figure 37: THD values for FC current.


waveform in the grid-integrated nonlinear supply of powerthrough a hybrid power generation system that is equippedwith a PV and a fuel cell. Power management is one of themain tasks of a power generating station because it mustmeet the clients’ demand. In the projected model, thehybrid PV/FC generated around 74 kilowatts of power.The microgrid alone provided the power to the buyer’sload, which is given in Figure 33. During this analysis, theother major task is to boost the power quality and to cutback on the THD value at the point of common coupling.This simulation deeply analyzes the THD value of all thegenerating sources and the microgrid. The PV system’sTHD values for the voltage and the current are given inFigures 34 and 35, respectively. The next two Figures 36and 37 present the FC energy system’s THD values for cur-rent and voltage, respectively. Finally, the microgrid load,voltage, and current THD values are analyzed and givenin Figures 38 and 39, respectively. The THD values are pre-sented in tabulated form in Table 4.

7. Conclusion

This paper has been written after detailed investigation anddevelopment of an advanced DNN controller-based MPPTcontroller algorithm for PV, which has been analyzed in var-ious atmospheric conditions. The PEM cell system has beendeveloped to manage the fuel flow pressure by DNN, whichis primarily based on the algorithm to enhance the fuel cellefficiency. Furthermore, the simulation results were evalu-ated and analyzed in different weather changes and fuel flowfluctuations to make sure that the proposed controller

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

Mag

(% o

f fun

dam

enta

l)

0 200 400 600 800 1000

Frequency (Hz)


Figure 38: THD values for microgrid current load.

0 200 400 600 800 1000

Frequency (Hz)

Mag

(% o

f fun

dam

enta

l)


0

0.5

1

1.5

2

2.5

3

3.5

Figure 39: THD values for microgrid voltage load.

Table 4: THD comparative values for grid, fuel cell, and PV.

Profile Voltage-THD Current-THD

Grid 3.55% 0.4%

FC 3.55% 4.46%

PV 3.55% 0.9%


algorithm is effective. The proposed MPPT results are com-pared with a fuzzy logic controller and prove the proposedcontroller effectiveness. The projected grid integration ofthe hybrid PV and fuel cell energy system has been developedand simulated in the MATLAB atmosphere. The target ofthis analysis is implementing a simulation model of a smartinverter-based microgrid, which was integrated with a hybridPV and the fuel cell energy system using the DNN algorithmin MATLAB atmosphere. The projected system has beensimulated in various operational conditions, and the resultsare reported in this paper. The improvements in the powerquality are the second most significant purpose of thisresearch. This objective was achieved through finding theTHD values of the PV, fuel cell, distributed grid voltage,and current profile. The proposed system THD valuesare less than 5% according to the IEEE 1547 and 519 stan-dards. The projected system and DNN controller effec-tively improved the power quality and the environmentalsurroundings.

Data Availability

All time available.

Conflicts of Interest

The authors declare that they have no conflicts of interest.

References

[1] M. Nurunnabi, N. K. Roy, E. Hossain, and H. R. Pota, “Sizeoptimization and sensitivity analysis of hybrid wind/PVmicro-grids- a case study for Bangladesh,” IEEE Access,vol. 7, pp. 150120–150140, 2019.

[2] R. Singh, R. C. Bansal, A. R. Singh, and R. Naidoo, “Multi-objective optimization of hybrid renewable energy systemusing reformed electric system cascade analysis for islandingand grid connected modes of operation,” IEEE Access, vol. 6,pp. 47332–47354, 2018.

[3] M. Xu, L. Wu, H. Liu, and X. Wang, “Multi-objective optimalscheduling strategy for wind power, PV and pumped storageplant in VSC-HVDC grid,” The Journal of Engineering,vol. 2019, no. 16, pp. 3017–3021, 2019.

[4] M. I. Ghiasi, M. A. Golkar, and A. Hajizadeh, “Lyapunovbased-distributed fuzzy-sliding mode control for buildingintegrated-DC microgrid with plug-in electric vehicle,” IEEEAccess, vol. 5, pp. 7746–7752, 2017.

[5] X. Gong, F. Dong, M. A. Mohamed, O. M. Abdalla, and Z. M.Ali, “A secured energy management architecture for smarthybrid microgrids considering PEM-fuel cell and electric vehi-cles,” IEEE Access, vol. 8, pp. 47807–47823, 2020.

[6] A. Elgammal and M. El-Naggar, “Energy management insmart grids for the integration of hybrid wind–PV–FC–batteryrenewable energy resources using multi-objective particleswarm optimisation (MOPSO),” The Journal of Engineering,vol. 2018, no. 11, pp. 1806–1816, 2018.

[7] B. S. Sami, “Intelligent energy management for off-grid renew-able hybrid system using multi-agent approach,” IEEE Access,vol. 8, pp. 8681–8696, 2020.

[8] S. Rehman, H. U. R. Habib, S. Wang, M. S. Büker, L. M.Alhems, and H. Z. Al Garni, “Optimal design and model pre-

dictive control of standalone HRES: a real case study for resi-dential demand side management,” IEEE Access, vol. 8,pp. 29767–29814, 2020.

[9] A. A. Z. Diab, H. M. Sultan, I. S. Mohamed, O. N. Kuznetsov,and T. D. Do, “Application of different optimization algo-rithms for optimal sizing of PV/wind/diesel/battery storagestand-alone hybrid microgrid,” IEEE Access, vol. 7,pp. 119223–119245, 2019.

[10] T. Adefarati, R. C. Bansal, and J. John Justo, “Techno-eco-nomic analysis of a PV–wind–battery–diesel standalone powersystem in a remote area,” The Journal of Engineering, vol. 2017,no. 13, pp. 740–744, 2017.

[11] N. T. Mbungu, R. C. Bansal, and R. Naidoo, “Smart energycoordination of a hybrid wind/ PV with battery storage con-nected to grid,” The Journal of Engineering, vol. 2019, no. 18,pp. 5109–5113, 2019.

[12] S. B. Santra, K. Kumar, P. Biswal, and C. K. Panigrahi, “Lyapu-nov based fast terminal sliding mode Q-V control of grid con-nected hybrid solar PV and wind system,” IEEE Access, vol. 6,pp. 39139–39153, 2018.

[13] Z. Ullah, S. Wang, J. Radosavljević, and J. Lai, “A solution tothe optimal power flow problem considering WT and PV gen-eration,” IEEE Access, vol. 7, pp. 46763–46772, 2019.

[14] J. S. Ojo, P. A. Owolawi, and A. M. Atoye, “Designing a greenpower delivery system for base transceiver stations in south-western Nigeria,” SAIEE Africa Research Journal, vol. 110,no. 1, pp. 19–25, 2019.

[15] I. Mahendravarman, S. A. Elankurisil, M. Venkateshkumar,A. Ragavendiran, and N. Chin, “Artificial intelligentcontroller-based power quality improvement for microgridintegration of photovoltaic system using new cascade multi-level inverter,” Soft Computing, vol. 24, no. 24, pp. 18909–18926, 2020.

[16] S. Yushu, Z. Zhenxing, Y. Min, J. Dongqiang, P. Wei, andX. Bin, “Research overview of energy storage in renewableenergy power fluctuation mitigation,” CSEE Journal of Powerand Energy Systems, vol. 6, 2020.

[17] R. Indumathi, M. Venkateshkumar, and R. Raghavan, “Integra-tion of D-Statcom based photovoltaic cell power in low voltagepower distribution grid,” in IEEE-International Conference onAdvances in Engineering, Science and Management (ICAESM-2012), pp. 460–465, Nagapattinam, Tamil Nadu, 2012.

[18] C. S. Chin, J. Xiao, A. M. Y. M. Ghias, M. Venkateshkumar,and D. U. Sauer, “Customizable battery power system formarine and offshore applications: trends, configurations, andchallenges,” IEEE Electrification Magazine, vol. 7, no. 4,pp. 46–55, 2019.

[19] B. Xu, D. Chen, M. Venkateshkumar et al., “Modeling apumped storage hydropower integrated to a hybrid power sys-tem with solar-wind power and its stability analysis,” AppliedEnergy, vol. 248, pp. 446–462, 2019.

[20] M. Venkateshkumar, R. Raghavan, and N. Kumarappan,“Design of a new multilevel inverter standalone hybrid PV/FCpower system,” Fuel Cells, vol. 15, no. 6, pp. 862–875, 2015.

[21] T. Hemanand, N. P. Subramaniam, and M. Venkateshkumar,“Comparative analysis of intelligent controller based micro-grid integration of hybrid PV/wind power system,” Journal ofAmbient Intelligence and Humanized Computing, pp. 1–20,2018.


Power Management Controller for Microgrid Integration of ...

Documents

Transcript of Power Management Controller for Microgrid Integration of ...