Performance improvement of off-grid hybrid renewable energy system using dynamic voltage restorer Wael S. Hassanein a , Marwa M. Ahmed b , M. Osama abed el-Raouf c , Mohamed G. Ashmawy d , Mohamed I. Mosaad e, * a Department of Industrial Engineering, Faculty of Engineering, King Abdulaziz University, Saudi Arabia b Department of Electrical Engineering, Faculty of Engineering, King Abdulaziz University, Saudi Arabia c Building Physics and Environmental Research Institute, Housing and Building, National Research Center (hbrc), Cairo, Egypt d Department of Electrical Power and Machine Engineering, El-Shorouk Academy, Cairo, Egypt e Yanbu Industrial College (YIC), Saudi Arabia Received 28 December 2019; revised 21 March 2020; accepted 28 March 2020 KEYWORDS Artificial Neural Network; Dynamic voltage restorer; Fuel cell; Hybrid system; Low voltage ride through; Photovoltaic; Wind turbine Abstract This article proposes an Artificial Neural Network (ANN) controller of Dynamic Volt- age Restorer (DVR) to improve the performance of a stand-alone hybrid renewable energy system that is feeding a new community located in Egypt. The hybrid system consists of three renewable energy sources, namely, solar PV cells, a wind turbines based-permanent magnet synchronous gen- erator, and fuel cells. These three sources are tied to a common DC link by three boost converters, one for each source. The common DC link is connected to the AC side via a DC/AC inverter. The optimal size of the three proposed renewable sources is calculated using the HOMER software package. The DVR control is attained through regulating the load voltage at different anomalous working conditions. These conditions are three-phase fault, voltage sag/swell, and unbalanced load- ing. Two ANNs are utilized to adjust the IGBT pulses of the voltage source inverter (VSI) used to control DVR by regulating the D-Q axes voltage signals. These D-Q axes components at any load- ing condition represent the inputs to the two ANNs. The outputs of the two ANNs represent the IGBT pulses. The input/output data used for training ANNs are obtained by two optimized PI con- trollers, introduced for regulating the load voltage through DVR-VSI pulses at different abnormal operating conditions, and accordingly convert the static optimized PI controller into adaptive one based ANN. The system performance with the proposed ANN-DVR controller is enhanced through improving the current, voltage, and power waveforms of each generating source. With compensation of the faulty line voltage, the system retains an uninterrupted operation of the three renewable sources during fault events and consequently increases the low voltage ride through (LVRT) capability. Moreover, the total harmonic distortion is reduced. Ó 2020 Faculty of Engineering, Alexandria University. Production and hosting by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). * Corresponding author. E-mail address: [email protected](M.I. Mosaad). Alexandria Engineering Journal (2020) xxx, xxx–xxx HOSTED BY Alexandria University Alexandria Engineering Journal www.elsevier.com/locate/aej www.sciencedirect.com https://doi.org/10.1016/j.aej.2020.03.037 1110-0168 Ó 2020 Faculty of Engineering, Alexandria University. Production and hosting by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Please cite this article in press as: W.S. Hassanein et al., Performance improvement of off-grid hybrid renewable energy system using dynamic voltage restorer, Alex- andria Eng. J. (2020), https://doi.org/10.1016/j.aej.2020.03.037
15
Embed
Performance improvement of off-grid hybrid renewable energy …€¦ · Performance improvement of off-grid hybrid renewable energy system using dynamic voltage restorer Wael S.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
https://doi.org/10.1016/j.aej.2020.03.0371110-0168 � 2020 Faculty of Engineering, Alexandria University. Production and hosting by Elsevier B.V.This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Please cite this article in press as: W.S. Hassanein et al., Performance improvement of off-grid hybrid renewable energy system using dynamic voltage restorerandria Eng. J. (2020), https://doi.org/10.1016/j.aej.2020.03.037
Wael S. Hassanein a, Marwa M. Ahmed b, M. Osama abed el-Raouf c,
Mohamed G. Ashmawy d, Mohamed I. Mosaad e,*
aDepartment of Industrial Engineering, Faculty of Engineering, King Abdulaziz University, Saudi ArabiabDepartment of Electrical Engineering, Faculty of Engineering, King Abdulaziz University, Saudi ArabiacBuilding Physics and Environmental Research Institute, Housing and Building, National Research Center (hbrc), Cairo, EgyptdDepartment of Electrical Power and Machine Engineering, El-Shorouk Academy, Cairo, EgypteYanbu Industrial College (YIC), Saudi Arabia
Received 28 December 2019; revised 21 March 2020; accepted 28 March 2020
KEYWORDS
Artificial Neural Network;
Dynamic voltage restorer;
Fuel cell;
Hybrid system;
Low voltage ride through;
Photovoltaic;
Wind turbine
Abstract This article proposes an Artificial Neural Network (ANN) controller of Dynamic Volt-
age Restorer (DVR) to improve the performance of a stand-alone hybrid renewable energy system
that is feeding a new community located in Egypt. The hybrid system consists of three renewable
energy sources, namely, solar PV cells, a wind turbines based-permanent magnet synchronous gen-
erator, and fuel cells. These three sources are tied to a common DC link by three boost converters,
one for each source. The common DC link is connected to the AC side via a DC/AC inverter. The
optimal size of the three proposed renewable sources is calculated using the HOMER software
package. The DVR control is attained through regulating the load voltage at different anomalous
working conditions. These conditions are three-phase fault, voltage sag/swell, and unbalanced load-
ing. Two ANNs are utilized to adjust the IGBT pulses of the voltage source inverter (VSI) used to
control DVR by regulating the D-Q axes voltage signals. These D-Q axes components at any load-
ing condition represent the inputs to the two ANNs. The outputs of the two ANNs represent the
IGBT pulses. The input/output data used for training ANNs are obtained by two optimized PI con-
trollers, introduced for regulating the load voltage through DVR-VSI pulses at different abnormal
operating conditions, and accordingly convert the static optimized PI controller into adaptive one
based ANN. The system performance with the proposed ANN-DVR controller is enhanced
through improving the current, voltage, and power waveforms of each generating source. With
compensation of the faulty line voltage, the system retains an uninterrupted operation of the three
renewable sources during fault events and consequently increases the low voltage ride through
(LVRT) capability. Moreover, the total harmonic distortion is reduced.� 2020 Faculty of Engineering, Alexandria University. Production and hosting by Elsevier B.V. This is an
open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
G irradiance (W/m2)ID saturation current of the diode of equivalent pho-
tovoltaic cell circuit
Id & Iq d and q axes currentsIL photocurrent of equivalent photovoltaic cell cir-
cuit
Isc PV cell short-circuit current at 25 �C and 1000W/m2
ISCðT1Þ short circuit current per cell at temp T1
ISCðT2Þ short circuit current per cell at temp T2
k Boltzmann’s constantK0 constant which determined from ISC vs. TKp proportional gain
Ki integral gainNpv number of PV panels
n ideality factor
Ppv power of photovoltaic panelq charge of an electronSr solar radiationSTC Standard Test Conditions
T temperature in �CVg band gap voltagevocðT1Þ open circuit voltage per cell at temperature T1
vocðT2Þ open circuit voltage per cell at temperature T2
voc open circuit voltageVd & Vq d and q axes voltages
Vki velocity of particle i at iteration k
W weighting functionx number of populationsXk
i current position of particle i at iteration k
g PV system efficiency
2 W.S. Hassanein et al.
1. Introduction
Renewable energy plays a major role in feeding electricity to
remote areas. It is the ideal alternative to traditional energy
as it is inexhaustible, and it has proved its importance and
achieved many important privileges as an alternative energy
for the future instead of fossil fuels, which will be exhausted
one day. Many developed and developing countries have
turned to renewable energy to reduce harmful emissions to
the atmosphere, reduce global warming, and save costs.
Renewable energy is available continuously and each country
can rely on the most available sources [1]. For example, Arab
countries can use solar energy to generate electricity and water
heating and others as the sun shines most months of the year.
Recently, many developed and developing countries use
renewable energy as unpolluted and renewable alternative, in
place of using contaminated materials [1]. Hybrid systems
are important for several reasons, the most important of which
are: to ensure the continuous feeding of the electrical loads at
various operating conditions, and to optimize the size of the
resources in remote areas [2]. It is preferable to use a hybrid
system, because the total capacity of more generation source
is smaller than the capacity of a system consisting of only
one source, whether solar or wind, as well as for the storage
capacities of batteries. One of the main advantages of using
hybrid systems in remote areas is that securing the electrical
supply from different power sources increases the reliability
and durability of the system [2,3].
A hybrid system (PV-Wind-fuel cell) tied to the grid is uti-lized to feed residential thermal and electrical loads with higheconomic efficiency as indicated in (A. Maleki et al. 2017 a) [4].
Another application for grid-connected PV systems consider-ing economic issues and investigating different technical solu-tions was presented in (O. Ayadi et al. 2018) [5]. Artificial
intelligence techniques have significant role in controlling therenewable energy systems. (A. Maleki et al. 2017) [6], presentedtwo heuristic approaches based-PSO which, are applied to PV-fuel cell-natural gas connected to grid to supply residential
Please cite this article in press as: W.S. Hassanein et al., Performance improvement oandria Eng. J. (2020), https://doi.org/10.1016/j.aej.2020.03.037
loads. (G. Zhang et al. 2018) [7], studied another applicationfor desalination plants that is dependent on the renewableenergy supplies such as wind and solar energies where simu-
lated annealing-chaotic search algorithm were utilized as anoptimization technique. (Shaimaa Barakat et al. 2018) [8], pre-sented a techno economic studies for hybrid renewable energy
systems with different optimization techniques such as opti-mizing a PV/Biomass hybrid system with different batterytechnologies.
There are some problems associated with integrated hybrid
power systems, such as harmonics issue generated from invert-ers that need adding extra equipment with more cost [3,9], inhybrid systems, there are different generation sources, and
any sudden change in the load or in the power generated fromany source can affect the system stability [9]. Load sharingissue in renewable energy hybrid system due to dependence
on renewable generation sources. The power economic dis-patch and calculating cost per unit generation are complicated[10]. Moreover, the short circuit occur in hybrid renewable
energy system sometimes lead to outage of the generationsource from system, especially if the control is not accurate.This short circuit, in case of wind systems may accelerate thegenerator speed to higher values especially without using con-
trollers that lead to disconnecting the generator from the sys-tem or damage the shaft [11].
Some devices were added to overcome these problems for
such hybrid systems [12]. Dynamic Voltage Restorer (DVR)and Static Transfer Switch were used for enhancing the perfor-mance of wind energy system (WES) and the system has the
ability to mitigate the disturbances applied to these systemsas in [13]. A. Prajyusha [14] presented a DVR that is capableof mitigating the power quality issues. Other devices including
Unified Power Quality Conditioner (UPQC) and D-STATCOM were used to enhance the performance of therenewable systems as in [15].
The control of these frameworks is considered as an imper-
ative issue [9,10]. The control process of these devices espe-cially when integrated to the hybrid systems is complicatedas the nonlinearity and complexity. Many control schemes
f off-grid hybrid renewable energy system using dynamic voltage restorer, Alex-
Performance improvement of off-grid hybrid renewable energy system 3
are presented to drive such devices to enhance the performanceof the hybrid systems [16].
PI controller method is introduced in many power system
control applications. PI controller is introduced to regulatethe inverter used to tie Fuel Cell (FC) to the electrical network[17]. PI controller parameters are optimized by many optimiza-
tion techniques from them particle swarm optimization (PSO).This optimization technique for finding the PI controllerparameters is performed off-line and for the purpose of online
control, some adaptations are needed as the optimal controllerparameters and optimal performance are not guaranteed whenchanging the operating conditions. The solution is using adap-tive control techniques for the offline optimal control. Many
adaptive control schemes are introduced such as adaptive PIbased ANN that was introduced to convert the static PI con-trollers into adaptive one as in [18]. Model reference predictive
control of STATCOM for enhancing the WES performancewas presented in [11].
The main contribution of this paper is using adaptive ANN
controller of DVR to enhance the effectiveness of off-gridhybrid system bolstering a new community in Egypt-case studyNew El-Farafra Oasis. The proposed controller (ANN-DVR)
regulates the voltage between DVR and the load to alleviatethe disturbances applied to the system and consequentlyimprove the system performance. A comparison between PItuned by PSO and adaptive ANN for controlling DVR is
summarized.
Fig. 1 (a) Block diagram of the proposed
Please cite this article in press as: W.S. Hassanein et al., Performance improvement oandria Eng. J. (2020), https://doi.org/10.1016/j.aej.2020.03.037
The paper is organized as follows. Section 2 describes thesystem with all its components. Section 3 represent the model-ing of the three renewable sources while Section 4 describes the
controlled design. Section 5 represents the results and discus-sions. Finally, Section 6 is the conclusions.
2. System description
The proposed system consists of a hybrid PV/WES/FC for bol-stering an off-grid community placed in Egypt, new Farafra
oasis [19,20]. The system consists of PV system tied to the firstDC-DC boost converter; WES is tied to the second DC-DCconverter while FC is tied to the third DC-DC converter.
The three DC-DC converters are connected to a DC link of780 V. Community AC loads are fed through an inverter.Two types of transformers are used, step up transformers are
used to feed off-grid network of the proposed communitywhich stepped up to 11 kV, and step-down transformers areused for supplying the AC loads of the community,11/0.380 kV, 50 Hz. DVR controlled by ANN, is connected
to the system at point of common coupling (PCC) to mitigatethe voltage swell, sag, unbalance and improve the power qual-ity of the system. Any distortion in supply voltage Vs (feeder
F1) may be compensated at the load voltage (feeder F2) byinjecting appropriate voltage from voltage source inverter(VSI) through an injection transformer, Fig. 1. The inverter
hybrid system and (b) DVR component.
f off-grid hybrid renewable energy system using dynamic voltage restorer, Alex-
output voltage is calculated according to system ratings. Allsystem parameters are given in details in [2].
3. System modeling
3.1. PV modeling
PV equivalent circuit depicted in Fig. 2, contains photo currentIL, and RS & Rsh series and parallel resistances respectively
[21].The modeling of the PV cell can be described as:
I ¼ IL � ID eq vþIRsð Þ
nkT � 1n o
ð1Þ
IL ¼ ILðT1Þð1þ K0 T� T1ð ÞÞ ð2Þ
ILðT1Þ ¼ G� ISC T1;nomð Þ=G nomð Þ ð3Þ
K0 ¼ ðISC T2ð Þ � ISC T1ð ÞÞ=ðT2 � T1Þ ð4Þ
ID ¼ IDðT1Þ � T
T1
� �3n
� e�qvg
nk� 1
T� 1T1
� �ð5Þ
IDðT1Þ ¼ ISCðT1Þ=ðeqvoc T1ð ÞnkT1 � 1Þ ð6Þ
RS ¼ �dv=dIvoc � 1=XV ð7Þ
XV ¼ I0ðT1Þ � q
nkT1
� eqvoc T1ð ÞnkT1 ð8Þ
where k and q are 1.38e�23 and 1.60e�19 respectively.
The PV power at a time t is determined as [21]:
Ppv tð Þ ¼ Sr tð Þ � a� g ð9ÞThe total power of the PV system can be defined as:
PtðtÞ ¼ Npv� PpvðtÞ ð10Þ
3.2. Wind turbine modeling
Kinetic energy from wind is converted to mechanical powerthrough wind turbine. Many types of generators were usedand implemented in WES. Induction generators with their
common types namely self-excited [22] and doubly-fed induc-tion generators [23,24] were used in many WES applications.Switched reluctance generators were also used in WES in both
distribution and generation levels [18,25]. PMSG were used inWES due to their ability of operation at variable wind speedconditions and gearless operation [26]. The WES-based PMSG
was presented in off-grid hybrid renewable system in [26].
Fig. 2 The equivalent circuit of solar cell.
Please cite this article in press as: W.S. Hassanein et al., Performance improvement oandria Eng. J. (2020), https://doi.org/10.1016/j.aej.2020.03.037
3.3. Fuel cell model
Fuel cell changes chemical energy into electrical energy. Theprinciple of its operation is the opposite of the principle ofthe work of the water electrolysis vessel, which analyzes the
water into its compound’s hydrogen and oxygen, consuminga certain amount of electrical energy. The Solid Oxide FuelCell (SOFC) unit used is depicted in Fig. 3. The fuel cell isfed with hydrogen from the side of the anode, so this electrode
is called a ‘‘fuel electrode”. Electrons that are not passed by theelectrolyte have to flow through an external circuit (e.g., theload) to land on the other pole of the cell, the cathode, which
forms the passage of electricity. At the cathode, the cell is fedwith air to provide the oxygen needed to make it work, thisoccurs in one of the poles which determined by cell type.
The electrochemical reaction within the cell consists of twoparts: the first is at the anode and the second is at the cathode.FC modeling and data are given in [2,27].
3.4. Optimal size of the PV/Wind/FC hybrid system
The Egyptian Code for electrical installations in buildingsstates that” electricity distribution companies in Egypt prepare
schedules to estimate the total electrical loads of all types ofbuildings”. These schedules, give the minimum total electricalcapacity that companies can agree with the owners of the
buildings under which they are fed by electric current. Thecapacity depends on the total area of all the floors of the build-ing and the nature of the activity in the building (residential,
educational, administrative, commercial, etc.. . .) and whetherthe building is located in an urban or rural area and the clas-sification of its location in those areas, etc. In this work, a newcommunity is proposed in new Farafra oases western desert of
Egypt, as a case study [30]. It consists of three loads (residen-tial load with services, light and medium industries loads andagriculture loads) these loads are now existing loads in Egypt,
so these loads are used in a recently found region in a desert ofEgypt as a case study. The new community is planned for20,000 inhabitants where, residential loads with services con-
sumptions represent 11,000 kWh/day, light and medium indus-tries load with services consumptions reflect a 174.8 MWh/day,
Fig. 3 Scheme of solid-oxide fuel cell.
f off-grid hybrid renewable energy system using dynamic voltage restorer, Alex-
Performance improvement of off-grid hybrid renewable energy system 5
and deferred load includes the agricultural load of thecommunity. The climbed yearly average deferred load is5772 kWh/day and with maximum load of 1443 kW. Agricul-
ture loads (irrigation using pumping water systems) represent5772 kWh/day.
HOMER software is utilized to determine the optimal size
of PV/Wind/FC hybrid system to feed the AC loads of a newcommunity based on some economic criteria [28,29]. The aver-age wind speed at the chosen site is 5.45 m/s and the radiation
is 8 kWh/m2/day.
Fig. 4 (a) Flowchart of DVR operation, (b) Control block
diagram, (c) PI control circuit.
Please cite this article in press as: W.S. Hassanein et al., Performance improvement oandria Eng. J. (2020), https://doi.org/10.1016/j.aej.2020.03.037
4. Controller design
4.1. Dynamic voltage restorer
As any disturbance is applied to the system, the DVR willintroduce a controlled voltage over an injection transformer
to mitigate that fault. Different disturbances are applied tothe system such as voltage sag, swell, unbalanced operationand three-phase short circuit. The DVR will mitigate these dis-
turbances. The flow chart of DVR process is given in Fig. 4a.The control circuit of DVR along with PI controller aredepicted in Fig. 4b and c. The ABC three-phase coordinatesystem is changed over into dq0 coordinate system as
follows [2]:
Vd ¼ 2
3VaSinwtþ VbSinVc sin wtþ 2p
3
� �� �ð11Þ
Vq ¼ 2
3VaCoswtþ VbCos wt� 2p
3
� �þ VcCos wtþ 2p
3
� �� �
ð12Þ
V0 ¼ 1
3Vaþ Vbþ Vc½ � ð13Þ
Table 1 DVR DATA.
Description Symbol Specifications
DVR DC voltage Vdc 700 V
DVR capacitor DC link C DC 5000 mF
Fig. 5 Flow Chart of PSO.
f off-grid hybrid renewable energy system using dynamic voltage restorer, Alex-
The load voltage (VLoad) is sensed and transformed intodq0 coordinates. The difference between the dq and the refer-ence voltages of dq coordinates represent the input to the PI
controllers, the d-axis reference is 1p.u and q-axis referenceis 0, Fig. 5c. Two PI controllers are utilized for d and q errorsignals PId, PIq respectively. The errord represents the input to
the D-axis PI controller while the errorq is for Q-axis PIcontroller.
Fig. 6 ANN controller for d-q coordinates of DVR.
Fig. 7 Nordal grid code.
Table 2 Optimal PI controller’s parameters determined by PSO.
Cases D-axis Q-
Kpd Kid Kp
3-phase fault 9.6 200.03 78
sag voltage
sag 30% 79.2 195.90 85
sag 60% 11.3 210.03 82
sag 90% 9.36 201.63 77
Swell voltage
Swell 30% 17.8 215.1 83
Swell 60% 21.94 217.92 85
Swell 90% 25.84 220.84 90
Unbalance voltage
Unbalance 4.003 154.32 52
Please cite this article in press as: W.S. Hassanein et al., Performance improvement oandria Eng. J. (2020), https://doi.org/10.1016/j.aej.2020.03.037
The integral of the square of error (ISE) index is utilized tostudy the system effectiveness as:
ISE ¼Z 1
0
e2 tð Þdt ð14Þ
e ¼ errord þ errorq ð15Þ
errord ¼ Vdref � Vd ð16Þ
errorq ¼ Vqref � Vq ð17ÞThe output of the PI controller is converted into ABC then
to PWM to activate the VSI-IGBT. The DVR data are shownin Table 1.
4.2. PSO algorithm for optimal tuning of PI control parameters
There are different optimization techniques for adapting PI
controller. PSO is one of them, where it is developed by search-ing on swarm. The solution of the problem represented by aparticle. Each particle has its own flying which it is adjusted
according to its flying background. In the D- dimensionalspace, the particle treated as a point. The ith particle is denotedas XI = (xi1, xi2. . .). Best value which called (pbest) is calcu-
lated according to minimum fitness value and represented asPI = (PI1, PI2. . .). In PSO technique the velocity of the particleis referred as VI = (VI1, VI2,. . .), the particles are updatedaccording
Vkþ1i ¼ WVk
i þ C1rand1 � pbesti � Xki
� þ C2rand2
� gbest� Xki
� ð18Þ
where Vki : Velocity of particle i at iteration k, W: weighting
function, C: weighting factor, j rand: random number between
0 and 1, Xki : Current position of particle i at iteration k, pbesti:
pbest of particle i, gbest: gbest of the group. The PI controlleroptimized parameters are determined for each test case. The
PSO flowchart for adapting the PI controller parameters isshown in Fig. 5.
axis D-axis Q-axis
q Kiq Vd Vq
.52 191.72 263.3 272.8
.00 124.91 270.4 265.9
.41 296.02 273.7 263.8
.83 190.84 265.3 272.2
.02 266.01 261.8 274
.40 269.03 273.1 265
.83 280.93 268.8 269.8
.43 180.52 263.7 272.6
f off-grid hybrid renewable energy system using dynamic voltage restorer, Alex-
Fig. 8 PV system with and without DVR at three-phase fault.
Performance improvement of off-grid hybrid renewable energy system 7
Please cite this article in press as: W.S. Hassanein et al., Performance improvement of off-grid hybrid renewable energy system using dynamic voltage restorer, Alex-andria Eng. J. (2020), https://doi.org/10.1016/j.aej.2020.03.037
Fig. 9 FC system with and without DVR at three-phase fault.
8 W.S. Hassanein et al.
Please cite this article in press as: W.S. Hassanein et al., Performance improvement of off-grid hybrid renewable energy system using dynamic voltage restorer, Alex-andria Eng. J. (2020), https://doi.org/10.1016/j.aej.2020.03.037
Fig. 10 WES with and without DVR at three-phase fault.
Performance improvement of off-grid hybrid renewable energy system 9
4.3. Adaptive control techniques (ANN controller)
As discussed in previous section, the difficulty and more run-ning time are considered as a disadvantage in using offline con-trol techniques, so that adaptive control plays an important
role to increase the system effectiveness with more rapidresponse. In the present application, two ANNs are used,
Please cite this article in press as: W.S. Hassanein et al., Performance improvement oandria Eng. J. (2020), https://doi.org/10.1016/j.aej.2020.03.037
one for D-axis and the other for the Q-axis. Each ANN hasthree layers namely input layer representing the D-axis andQ-axis values of the load voltage change for first and second
ANN respectively, a hidden layer including 20 neurons andan output layer representing the D-axis and Q-axis voltagesfor the first and second ANN respectively with supervised
training via a back-propagation as depicted in Fig. 6. The out-put voltage signals of the two ANN in the D-Q are trans-formed into (Va, Vb, Vc) to drive PWM of voltage source
inverter of DVR. The ANN training is obtained by eightabnormal operating conditions counting three phase fault,voltage sag/swell and unbalanced operation. At each conditionthe load voltage is sensed and transformed to D-Q frame that
represent the input to the two ANNs. Tuning of the two PIcontroller proposed by PSO is introduced at each time (eighttimes) for optimal control of DVR to increase the hybrid sys-
tem execution and the D-Q frame of the voltage control signalsare used as the output of the two ANNs. The D-Q frame-controlled voltage signals are converted into Vabc components
to drive the PWM of IGBT used in the VSI of DVR system. Inthis work ISE performance function has been used to representthe error between the target and input values. In this paper, the
number of epochs required for training D coordinate ANN is500 and the best validation performance is 0.51137 at epoch500 and the number of epochs required for training Q coordi-nate ANN is 450 and the best validation performance is
0.40037 at epoch 400.
4.4. Assessment of integrating DVR into hybrid system throughLVRT
The DVR is integrated into the hybrid PV/wind/fuel cell tosupport the dynamic performance of the proposed system dur-
ing abnormal operating conditions. Through using LVRTcapability, the hybrid system especially for wind generators,the system will be able to remain in service during abnormal
operating conditions. Nordal grid code [31] is used in this arti-cle, as an example, as shown in Fig. 7.
5. Results and discussions
The proposed system bolstering a new community in Egypt,given in Fig. 1 is introduced for testing the validity of the pro-posed ANN-DVR controller over PI controller optimized by
PSO considering sag, swell, unbalance voltages and three-phase short circuit.
5.1. Homer results
Homer software [28] was used for determining the optimal rat-ing of each generating source introduced in this paper. With
eight different scenarios for all possible combinations of threesources, the results show that, the optimal combination ofthree renewable sources are 30.687, 13.5 and 46 MW for PV,
WES and FC systems respectively, generic 1 kWh lead-acidbattery (117,089 strings) and system converter (32.911 MW)with a dispatch strategy of load. Moreover, the added to netdisplay cost is $ 265 M, the capital price is $ 193 M and the
cost of electricity is $ 0.293/kWh.
f off-grid hybrid renewable energy system using dynamic voltage restorer, Alex-
In this part, the DVR integrated into the hybrid renewableenergy system is regulated by two PI controllers optimizedby PSO to keep to the Nordal grid code and to improve the
overall system performance. Eight test cases were presentedcounting three-phase fault, sag/swell with three levels of volt-age for each, and unbalanced voltage. These different casesare simulated for obtaining the PI control parameters required
for control DVR and the corresponding D/Q voltage signalsrequired for training the ANNs. The PSO optimized PI con-troller parameters for all the above-mentioned cases are given
in Table 2. Through these cases, the regulated voltage andeffects of DVR on load voltages are studied. Also, the effectson rotor wind speed before and after using a DVR are illus-
trated. Moreover, the role of connecting DVR to the systemto keep the normal operation of PV/Wind/Fuel cell without
(a) Output load voltage
WW
WiWiLV
(a)
(b)
Fig. 11 Output voltage with and w
Please cite this article in press as: W.S. Hassanein et al., Performance improvement oandria Eng. J. (2020), https://doi.org/10.1016/j.aej.2020.03.037
disconnection during voltage dips due to several fault eventsare taken place at F1.
5.3. Test case 1: three-phase fault at F1
In this case three-phase fault is applied at feeder F1 with abetween 0.5 and 0.75 s. This fault affects the output of the gen-
eration sources. The voltage of PV decreases, the currentincreases and as a result the power increases system as shownin Fig. 8.
With this fault event, a decrease is taken place in the FCvoltage. While increase in the FC current to 150% that will dis-connect FC from the system, Fig. 9.
Due to this fault, wind system current increases as shown inFig. 10b. Through this fault, the electrical torque throbs incor-porates more motions with low-frequency esteem and this maylead to failure of the system due to large vibrations on the
, (b) Voltage at PCC
ithout DVR ith DVR PI-PSO
thout DVR th DVR PI-PSO RT
ithout DVR at three-phase fault.
f off-grid hybrid renewable energy system using dynamic voltage restorer, Alex-
Without DVR DVR with static PI DVR with retuned PI
Without DVR DVR with static PI DVR with retuned PI
(a)
(b)
Fig. 12 RMS of the load voltage with and without DVR.
(a) Voltage at PCC, (b) RMS output load voltage
Without DVR DVR with static PI DVR with retuned PI
Without DVR DVR with static PI DVR with retuned PI LVRT
(a)
(b)
Fig. 13 Phase-a voltage with and without DVR.
Without DVR DVR with static PI DVR with retuned PI DVR with adapted ANN LVRT
Fig. 14 Voltage at PCC without and with DVR based PI-PSO,
based ANN.
Performance improvement of off-grid hybrid renewable energy system 11
mechanical part of the wind system. The generator speed isincreased to high values (20%) without using DVR that willlead to disconnecting the generator from the system as the
speed surpasses greatest permissible cutoff of 17% [11]. Thewind turbine rotor speed is depicted with and without DVRin Fig. 10d.
The output voltage when using controlled DVR shows bet-
ter performance than without using DVR, Fig. 11a during thisthree-phase fault. DVR has the capacity to regulate the voltageat PCC with less harmonics during this fault conditions using
PSO-PI controller and consents to the Nordal grid code pro-posed, Fig. 11b.
5.4. Test case 2: Voltage sag/swell conditions
In this case, 30% voltage sagging condions were considered toassess the effictivness of the tunned PI controller for DVR to
amelorate the stsyem effectivness. Moroever, a comaprisonbetween staic tunned PI parameters obtained for the three-phase fault case (test case1) and the retuned PI control param-etes obtained for the 30% voltage sagging condition is intro-
duced as given in Table 2. RMS load voltage enhancement isshown in Fig. 12a. Although the load voltage levels areimproved using DVR with both static and retuned PI con-
trollers, the retuned PI controllers give better performancedue to the PSO is tunned at the 30% sagging condition. Con-sequently and in order to get the best (optimal) performance at
Please cite this article in press as: W.S. Hassanein et al., Performance improvement oandria Eng. J. (2020), https://doi.org/10.1016/j.aej.2020.03.037
each abnormal condition, PI controller parameters should beretuned at each condition that can not be achieved as this tun-ning process is performed off-line rather than long time taken
for each tuning. The adaptive control is the key solution forthe retunned PI control paramerets hence, adaptive ANN con-trol will be used. Eight test cases summerized in Table 2 are
f off-grid hybrid renewable energy system using dynamic voltage restorer, Alex-
used for training the ANN. On the other hand, with 30% swellconditions, the RMS output load voltage enhancement at swellvoltage is shown in Fig. 12b. With DVR and the PI controllo-
ers are not retuned for this condition (vaules of case study1),the load voltage at this condition reaches to 430 V. With theretuned PI controller parameters, the value is 380 V. This
ensure the need for adaptive control rather than optimizedPI controller.
(a) Voltage at F1, (b)(d) Rotor wind turb
WWW
(a)
(d)
Fig. 15 System performance during three-phase
Please cite this article in press as: W.S. Hassanein et al., Performance improvement oandria Eng. J. (2020), https://doi.org/10.1016/j.aej.2020.03.037
5.5. Test case 3: Unbalanced voltage condition
A single line to ground fault at phase-a, is applied to the sys-tem in this case. Without using DVR, the PCC voltage reaches0.37p.u with more oscillations, but with DVR and static PI
controller, the PCC voltage reaches 0.58p.u while with retunedPI controller parameters, the value is 0.68p.u., Fig. 13a. Theload voltage with and without DVR at unbalance voltage is
Load (c) Injected voltages ine speed
ithout DVR ith DVR PI ith DVR ANN
(b)
(c)
short-circuit with DVR based adapted ANN.
f off-grid hybrid renewable energy system using dynamic voltage restorer, Alex-
Performance improvement of off-grid hybrid renewable energy system 13
shown in Fig. 13b. Although voltage enhancement at PCC andRMS load voltage, this tunning process is performed off-linerather than long time taken for this tuning, so that adaptive
ANN controller will be used.
5.6. ANN adaptive control
Now ANN adaptive control technique is applied to the hybridsystem. As mentioned before, two ANN networks are used, thefirst for D-axis and the second for Q-axis. The input to ANN is
the change of load VL in D-Q coordinates and the output isupdated three phase voltages (Vd, Vq, V0) control parametersused to drive PWM of voltage source inverter of DVR. Differ-
ent test cases are used to test the effectiveness of PI controller.Eight test cases summerized in Table 2 are used for training theANN with obtaining an online tuning for control parameters.With three-phase fault and without using DVR, the PCC volt-
age raches 0.39p.u, Fig. 14. With DVR and the PI controlloersare not retuned for this condition (vaules of case study1), theminimu PCC volage at the sagging interval is 0.58p.u while
with retuned PI controller parameters, the minimum value is0.78p.u, while adaptive ANN controller, the minimum valuereaches to 0.9 p.u. the voltage levels at PCC are improved
using DVR with adaptive ANN controller more than both sta-tic and retuned PI controller, with adaptive ANN controllers,there is no need for retuning PI parameters.
Fig. 16 THD with DVR based
Please cite this article in press as: W.S. Hassanein et al., Performance improvement oandria Eng. J. (2020), https://doi.org/10.1016/j.aej.2020.03.037
DVR based ANN could regulate the voltage at PCC withdiminished harmonics through faulty conditions as comparedto the PI-PSO as shown in Fig. 14. Fig. 15a to c show the volt-
age at F1, load and injected voltage waveforms during three-phase short circuit with ANN controller, which has a veryrapid response with smoothing waveforms than PI controller.
With DVR based PI-controller, the generator speed isincreased to values (10%), but with using ANN adaptive con-troller where, the generator rotor speed reaches to values (8%),
Fig. 16d. A comparison of Wind rotor speed without using aDVR, with using PI-PSO and with using adapted ANN isshown in Fig. 16.
5.7. Total harmonic distortions
The Total Harmonic Distortion (THD) is measured by usingFFT analysis. THD is calculated as, [17]
where V1: is the main harmonic magnitude in RMS, nth har-monics magnitude. Fig. 16a and b show the total harmonic dis-tortions with DVR based PI-PSO (4.31%) and ANN (0.86%).
The simulation is carried using FFT analysis in MATLABsoftware.
(a) PI-PSO and (b) ANN.
f off-grid hybrid renewable energy system using dynamic voltage restorer, Alex-
The application of a DVR based ANN connected to hybridPV/Wind/FC supplying the electrical power to a new off-grid
community proposed in Egypt is investigated in this paper.ANN controller of DVR has the ability to overcome somefault events include, voltage sag/swell, unbalanced operation
and three-phase faults at PCC. The proposed controller suc-ceeded at keeping the three renewable sources (PV/Wind/fuelcell) in continues operation during these fault events withoutdisconnection from the system. The main idea behind this pro-
posed controller is to regulate the voltage between DVR andconsequently improve the voltage profile at PCC. A compar-ison between this proposed scheme and PI controller tuned
by PSO was introduced. The proposed controller showed abetter performance rather than static and retuned PI con-troller. Moreover, the proposed ANN controller of DVR
showed has a rapid response with minimum total harmonicdistortion level.
Declaration of Competing Interest
The authors declare that they have no known competing
financial interests or personal relationships that could haveappeared to influence the work reported in this paper.
Acknowledgment
This project was funded by the Deanship of Scientific Research(DSR), king Abdulaziz University, Jeddah, under grant
No. (DF-202-135-1441). The authors, therefore, gratefullyacknowledge DSR technical and financial support.
References
[1] M.O.A. El-Raouf, M.I. Mosaad, M.A. Al-Ahmar, F.M. El
Bendary, MPPT of hybrid solar-wind-grid power generation
system, Int. J. Ind. Electron. Driv. 2 (4) (2015) 234–241.
[2] Mohamed I. Mosaad, M. Osama Abed El-Raouf, Mahmoud A.
Al-Ahmar, Fahmy M. El Bendary, Optimal PI Controller of
DVR to Enhance the Performance of Hybrid Power System
Feeding a Remote Area in Egypt, Sustain. Cities Soc. 47(May)
(2019) 101469–101483.
[3] M. Osama abed el-Raouf, Mohamed I. Mosaad, Adel
Mallawany, Mahmoud A. Al-Ahmar and Fahmy M. El
Bendary, Optimal Control of DVR to Enhance The Power
Quality of Pv/Wind/Fuel Cell Hybrid System Feeding a New
Community, in: 25th International Conference on Electricity
Distribution CIRED, Madrid, 3–6 June 2019, pp. 599–603.
[4] A. Maleki, M. Rosen, F. Pourfayaz, Optimal Operation of a