Wiley

24
Microgrid dynamic responses enhancement using articial neural network-genetic algorithm for photovoltaic system and fuzzy controller for high wind speeds Alireza Rezvani 1, * ,, Maziar Izadbakhsh 1 and Majid Gandomkar 2 1 Young Researchers and Elite Club, Saveh Branch, Islamic Azad University, Saveh, Iran 2 Department of Electrical Engineering, Saveh Branch, Islamic Azad University, Saveh, Iran ABSTRACT The microgrid (MG) is described as an electrical network of small modular distributed generation, energy storage devices and controllable loads. In order to maximize the output of solar arrays, maximum power point tracking (MPPT) technique is used by articial neural network (ANN), and also, control of turbine output power in high wind speeds is proposed using pitch angle control technic by fuzzy logic. To track the maximum power point (MPP) in the photovoltaic (PV), the proposed ANN is trained by the genetic algorithm (GA). In other word, the data are optimized by GA, and then these optimum values are used in ANN. The simulation results show that the ANN-GA in compar- ison with the conventional algorithms with high accuracy can track the peak power point under different insolation conditions and meet the load demand with less uctuation around the MPP; also it can increase convergence speed to achieve MPP. Moreover, pitch angle controller based on fuzzy logic with wind speed and active power as inputs that have faster responses which leads to have atter power curves enhances the dynamic responses of wind turbine. The models are developed and applied in Matlab/Simulink. Copyright © 2015 John Wiley & Sons, Ltd. Received 28 July 2014; Revised 19 March 2015; Accepted 12 May 2015 KEY WORDS: microgrid; photovoltaic; permanent magnet synchronous generation (PMSG); neural network; genetic algorithm 1. INTRODUCTION The application of distributed energy resources (DER) is proposed to provide efcient and reliable power to electricity customers closer to the point of use. They are usually clean, renewable, small, exible and have become important elements in a diversied set of alternative generation sources. Interconnection networks of distributed energy resources, energy storage systems and loads dene a MG that can operate in stand-alone or in grid-connected mode [1, 2]. The MG is disconnected automat- ically from the main distribution system and change to islanded operation when a fault occurs in the main grid or the power quality of the grid falls below a required standard. MGs are capable to improve the reliability of electrical energy supply if appropriate control techniques are implemented. It can rep- resent a complementary infrastructure to the utility grid due to the rapid change of the load demand. In grid- connected mode, the grid dominates most of the system dynamics and no signicant issues need to be addressed except the power ow control, whereas in the islanding mode, once the isolating switch disconnects the utility from the MG. The MG concept enables high penetration of distributed genera- tion (DG) without requiring re-design or re-engineering of the distribution system itself [3, 4]. Developing photovoltaic energy sources can reduce fossil fuel dependency. PV panels are low- energy conversion efcient; therefore, using the MPPT system is recommended. In other words, the *Correspondence to: A. Rezvani, Young Researchers and Elite Club, Saveh Branch, Islamic Azad University, Saveh, Iran. E-mail: [email protected] INTERNATIONAL JOURNAL OF NUMERICAL MODELLING: ELECTRONIC NETWORKS, DEVICES AND FIELDS Int. J. Numer. Model. (2015) Published online in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/jnm.2078 Copyright © 2015 John Wiley & Sons, Ltd. Downloaded from http://www.elearnica.ir

Transcript of Wiley

Page 1: Wiley

INTERNATIONAL JOURNAL OF NUMERICAL MODELLING: ELECTRONIC NETWORKS, DEVICES AND FIELDSInt. J. Numer. Model. (2015)Published online in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/jnm.2078

Downlo

Microgrid dynamic responses enhancement using artificial neuralnetwork-genetic algorithm for photovoltaic system and fuzzy

controller for high wind speeds

Alireza Rezvani1,*,†, Maziar Izadbakhsh1 and Majid Gandomkar2

1Young Researchers and Elite Club, Saveh Branch, Islamic Azad University, Saveh, Iran2Department of Electrical Engineering, Saveh Branch, Islamic Azad University, Saveh, Iran

ABSTRACT

The microgrid (MG) is described as an electrical network of small modular distributed generation, energy storagedevices and controllable loads. In order to maximize the output of solar arrays, maximum power point tracking(MPPT) technique is used by artificial neural network (ANN), and also, control of turbine output power in high windspeeds is proposed using pitch angle control technic by fuzzy logic. To track the maximum power point (MPP) in thephotovoltaic (PV), the proposed ANN is trained by the genetic algorithm (GA). In other word, the data are optimizedby GA, and then these optimum values are used in ANN. The simulation results show that the ANN-GA in compar-ison with the conventional algorithms with high accuracy can track the peak power point under different insolationconditions and meet the load demand with less fluctuation around the MPP; also it can increase convergence speed toachieve MPP. Moreover, pitch angle controller based on fuzzy logic with wind speed and active power as inputs thathave faster responses which leads to have flatter power curves enhances the dynamic responses of wind turbine. Themodels are developed and applied in Matlab/Simulink. Copyright © 2015 John Wiley & Sons, Ltd.

Received 28 July 2014; Revised 19 March 2015; Accepted 12 May 2015

KEY WORDS: microgrid; photovoltaic; permanent magnet synchronous generation (PMSG); neural network;genetic algorithm

1. INTRODUCTION

The application of distributed energy resources (DER) is proposed to provide efficient and reliablepower to electricity customers closer to the point of use. They are usually clean, renewable, small,flexible and have become important elements in a diversified set of alternative generation sources.Interconnection networks of distributed energy resources, energy storage systems and loads define aMG that can operate in stand-alone or in grid-connected mode [1, 2]. The MG is disconnected automat-ically from the main distribution system and change to islanded operation when a fault occurs in themain grid or the power quality of the grid falls below a required standard. MGs are capable to improvethe reliability of electrical energy supply if appropriate control techniques are implemented. It can rep-resent a complementary infrastructure to the utility grid due to the rapid change of the load demand. Ingrid- connected mode, the grid dominates most of the system dynamics and no significant issues needto be addressed except the power flow control, whereas in the islanding mode, once the isolating switchdisconnects the utility from the MG. The MG concept enables high penetration of distributed genera-tion (DG) without requiring re-design or re-engineering of the distribution system itself [3, 4].

Developing photovoltaic energy sources can reduce fossil fuel dependency. PV panels are low-energy conversion efficient; therefore, using the MPPT system is recommended. In other words, the

*Correspondence to: A. Rezvani, Young Researchers and Elite Club, Saveh Branch, Islamic Azad University, Saveh, Iran.†E-mail: [email protected]

Copyright © 2015 John Wiley & Sons, Ltd.

aded from http://www.elearnica.ir

Page 2: Wiley

A. REZVANI, M. IZADBAKHSH AND M. GANDOMKAR

output power of a PV module varies as a function of the voltage, and also the MPP point is changed byvariation of temperature and sun irradiance [5].

The most prevalent techniques are perturbation and observation (P&O) algorithm [5], incrementalconductance (IC) [6, 7], fuzzy logic [8, 9] and ANN [10–12]. P&O and IC can track the MPP allthe time, regardless of the atmospheric conditions, type of PV panel, by processing real values ofPV voltage and current. Due to the aforementioned inquiries, the profits of P&O and IC methods arelow cost execution and elementary method. One of the drawbacks of these techniques is vast variationof output power around the MPP even under steady state; therefore, it caused the loss of availableenergy more than the other methods [13, 14]. Nevertheless, rapid changing of weather condition affectsthe output power, and these methods cannot track easily the MPP.

Using fuzzy logic can solve the two mentioned problems dramatically. In fact, fuzzy logic controllercan reduce oscillations of output power around theMPP and losses. Furthermore, in this way, convergencespeed is higher than the other two ways mentioned. A weakness of fuzzy logic in comparison with ANNrefers to oscillations of output power around the MPP [15, 16].

Nowadays, artificial intelligence (AI) methods have numerous applications in determining the size ofPV systems, MPPT control and optimal structure of PV systems. In most cases, multilayer perceptron(MLP) neural networks or radial basis function network (RBFN) are employed for modeling PVmoduleand MPPT controller in PV systems [17, 18]. ANN-based controllers have been applied to estimatevoltages and currents corresponding to the MPP of PV module for irradiances and variable temperatures.A review on AI techniques applications in renewable energy production systems has been presented inthese literatures [10, 19].

In [20–22], GA is used for data optimization, and then, the optimum values are utilized for trainingneural networks, and the results show that, the GA technique has less fluctuation in comparison withthe conventional methods. However, one of the major drawbacks in mentioned papers is that theyare not practically connected to the grid in order to ensure the analysis of PV system performance.

As one of the eminent DG sources, wind power generation system (WPGS) is presented [23, 24]. Also,amongst the synchronous and asynchronous generators, permanent magnet synchronous generator (PMSG)ismore favorable due to self-excitation, lower weight, smaller size, less maintenance cost and the eliminationof gearbox have high efficiency and high power factor comparing to Wound Rotor Synchronous Generator(WRSG), Squirrel Cage Induction Generator (SCIG), Doubly Fed Induction Generator (DFIG) and so on.The PMSG does not require a supplementary supply for magnetic field excitation or slip rings and brushes.Moreover, they can operate in a relatively vast range of wind speeds [24, 25]. The main advantage ofvariable wind turbines is the capability of the MPPT from wind energy sources [26].

The major disadvantage of the PMSG is the risk of demagnetization caused by too high temperaturesor high currents. However, in order to obtain the maximum power of wind energy, using aMPPT systemis too indispensable. Variable speed wind turbines operate in two primary regions as below rated powerand above rated power. When power production is below the rated power for the machine, theturbine operates at variable rotor speeds to capture the maximum amount of energy available inthe wind [27, 28]. Generator torque provides the control input to vary the rotor speed, and the bladepitch angle is held constant. In above-rated power conditions, the primary objective is to maintain aconstant power output. This is generally achieved by holding the generator torque constant andvarying the blade pitch angle. MPPT controller somehow changes the rotor speed according tovariations of wind speed that the tip speed ratio (TSR) is maintained in optimum value.

One of the approaches to reach the MPPT is pitch angle control (B) which in small turbines with lowpower delivery is not possible due to mechanical difficulties in production [29]. In high speed wind the extraproduction of active power via wind turbine leads to increased consumption of reactive power in generator,and in which case, we should utilize the reactive power compensator for injecting reactive power that hasextra cost, too. Moreover, in above rated wind speed operation, mechanical erosion and damages will makeus to have more maintenance cost, and this leads us to use controller with fast and suitable response.

The PIDs are used mostly in controllers design, but by the introduction of fuzzy logic instead of PIDcreated a better performance such that it was the best preventative way to eliminate the profoundmathematical understanding of the system. In comparing PIDs and fuzzy logic systems, fuzzy logic hasmore stability, faster and smoother response, smaller overshoot and does not need a fast processor; also itis more powerful than other non-linear controllers [30, 31]. The pitch angle based on fuzzy logic controller

Copyright © 2015 John Wiley & Sons, Ltd. Int. J. Numer. Model. (2015)DOI: 10.1002/jnm

Page 3: Wiley

MICROGRID; PHOTOVOLTAIC; PMSG; NEURAL NETWORK; GENETIC ALGORITHM

is reported in [32–34]. In [34], active power and in [32, 33], both reactive power and rotor speed are imple-mented as input signals and because in mentioned items wind speed is neglected, the controller has not fastresponse which causes mechanical fatigues to the PMSG. Moreover, another drawback in these papers isthat it is not practically connected to grid to investigate the system performance [33–35].

Microturbine generation (MTG) in recent decades because of their small size, relatively low cost,low pollution, fuel diversity, low maintenance cost, relatively simple control and ability to operate inboth grid-connected and stand-alone modes has also received a lot of attention [36, 37], and this modeltaking control, speed, temperature, acceleration and fuel is developed.

Flywheel energy storage system (FESS) is an energy storage technology which can transformelectrical energy into mechanical energy. It has fast response, high dynamics, long life, good efficiencyand characteristics of infinite times of charge and discharge. However it has small storage capacity andhigh initial cost. The flywheel can be used alone to supply loads in the short-term failure of system,which can increase electric reliability, and stabilize the power fluctuations of DGs and loads [38].

In [39], the dynamic characteristics of a grid connected MG associated with power conditioningsystem (PCS) to regulate its power have been investigated. Also, four-quadrant operation of PCSand utilization of PCS to control the power of MG are reported. The MG during grid connected andislanding modes is presented in [40]. In [41], the MG’s grid connected operation during and subse-quent to the islanding mode was investigated; however, the dynamic model of distributed generations(DGs) is not considered, which has a tremendous effect on dynamic responses of the MG subsequent toislanding occurrence. Moreover, DGs (wind, PV, MTG, FC and etc.) are not included in their model.Virtually, in previous references the grid connected process has not reported the influence of windspeed deviations in dynamic responses of the MG, especially the islanding occurrence. In [42], a typ-ical configuration of an MG including three DGs was presented but it has not been analyzed the DGstructure, controllers of each micro source and fault occurrence. The P&O method in PV and wind sys-tem in the MG is addressed in [43], while the P&O method has enormous deviation of output poweraround the MPP. Also, in the aforementioned paper, there is not any controller (pitch angle control)in order to control the output power of WT in high speed which can lead to the damages to PMSG;besides, the P/Q control technique for wind system was not utilized in inverter.

The main objectives of the present study to overcome the disadvantages of the aforementionedreferences are as follows: (i) it is worth to mention that the major part of ANN is the desired datafor training process should be achieved for each PV system and each particular position. First PVsystem is simulated, then GA-based offline trained ANN is applied to provide the reference voltagecorresponding to the maximum power by using Matlab software. Temperature and irradiance as inputdata are given to GA, and optimal voltages (Vmpp) corresponding to MPP are obtained, and then theseoptimum values are used in neural network training. (ii) The FLC (for pitch angle) is proposed tosmooth the output power fluctuations of WT in above rated speed and a comparison of theperformances of the FLC with the conventional PI and GA controller.

The paper is organized as follows: In section 2 the structure of photovoltaic module has beendescribed. In section 3 the steps of implementing genetic algorithm and neural network are discussed.In section 4 PMSG generator and pitch angle controller based on fuzzy logic are discussed. In section 5the MTG system is explained. In section 6 FESS is investigated. In section 7 P–Q, droop and backupcontrollers are described. In section 8 the results are presented based on case studies. Finally, theconclusion is presented in section 9.

2. PHOTOVOLTAIC CELL MODEL

A PV module is a collection of PV panels. A PV cell can be represented by an equivalent circuit, asillustrated in Figure 1. The characteristics of the PV cell can be represented by the following equations[5, 10, 12]:

IPV ¼ Id þ IRP þ I (1)

I ¼ IPV � I0 expVþ RSIVthn

� �� 1

� �� V þ RsI

RP(2)

Copyright © 2015 John Wiley & Sons, Ltd. Int. J. Numer. Model. (2015)DOI: 10.1002/jnm

Page 4: Wiley

Figure 1. Equivalent circuit of one PV array.

A. REZVANI, M. IZADBAKHSH AND M. GANDOMKAR

Vth ¼ NskT

q(3)

I0 ¼ I0;nTn

T

� �3

expq*Eg

n*k1Tn

� 1T

� �� �(4)

where, I is the output current, V is the output voltage, Ipv is the photocurrent of the PV cell (A), Id is thediode current, IRP is the shunt leakage current, I0 is the diode reverse saturation current and n is theideality factor (1.36) for a p–n junction. Vth is known as the thermal voltage. q is the electron charge(1.60217646×10�19 C), k is the Boltzmann constant (1.3806503×10�23 J/K) and T (in Kelvin) is thetemperature of the p–n junction. Eg is the band gap energy of the semiconductor (Eg≈ 1.1 eV for thepolycrystalline Si at 25 °C), and I0,n is the nominal saturation current. T is the cell temperature, andTn is cell temperature at reference conditions. Under normal circumstances, the Rp has a large value,and Rs has a small value. In order to simplify the analysis, Rp and Rs can be neglected [10, 12, 39].Hence, we could assume that series resistance Rs is close to zero and shunt resistance Rp is close toinfinite. This model is simulated by Matlab Simulink. Red sun 90 w is taken as the reference modulefor simulation as well as comparison of parameters of the adjusted model and red sun data sheet valuesat reference conditions is presented in Table I. The arrays of PV modules are established by connecting11 panels in series, and 6 panels in parallel to obtain the power output of 6 kW.

3. MPPT—NEURAL NETWORK AND GENETIC ALGORITHM TECHNIQUE

3.1. The steps of implementing genetic algorithm

The GA-based offline trained network is employed to provide the reference voltage corresponding tothe maximum power. Alongside, GA is utilized for optimum values and then, optimum values are usedfor training network [20–22, 44]. The procedure for exerting GA can be presented as follows [20–22]:(i). assigning the objective function and recognizing the design parameters, (ii). determining the initialproduction population, (iii). evaluating the population using the objective function and (iv). conductingconvergence test stop if convergence is provided.

The objective function of GA is applied for its optimization by the following: finding the optimumX= (X1, X2, X3,…, Xn) to put the F(X) in the maximum value, where the number of design variables isconsidered as 1. X is the design variable equal to array current (Ix) and also, F(X) is the array output

Table I. Comparison of parameters of the adjusted model and red sun data sheet values at reference conditions.

Parameters Model Datasheet

IMP (current at maximum power) 4.84 A 4.94 AVMP (voltage at maximum power) 18.45V 18.65VPMAX (maximum power) 89.3W 90WVOC (open circuit voltage) 22.12V 22.32VISC (short circuit current) 5.04 A 5.24 ANP (total number of parallel cells) 1 1NP (total number of parallel cells) 36 36Series resistance (Rs) .1Ω Not specifiedShunt resistance (Rp) 161.34Ω

Copyright © 2015 John Wiley & Sons, Ltd. Int. J. Numer. Model. (2015)DOI: 10.1002/jnm

Page 5: Wiley

MICROGRID; PHOTOVOLTAIC; PMSG; NEURAL NETWORK; GENETIC ALGORITHM

power which should be maximized [21]. The GA parameters are given in Table II. The relationshipbetween voltage and current of the array is demonstrated by the following equations:

F Xð Þ ¼ VX*IX (5)

VX ¼ ns v0 � RS

npIX þ nk T þ 273ð Þ=qð ÞLn* IPV � IX

nPþ I0

I0

! !(6)

To determine the objective function, the power should be arranged based on the current of array (IX):

F Xð Þ ¼ ns v0 � RS

npIX þ nk T þ 273ð Þ=qð ÞLn*

IPV � IXnpþ I0

I0

! !*IX (7)

0 < IX < ISC (8)

The current constraint should be noted too. By maximizing this function, the optimum values forVmpp and MPP will result in any particular temperature and irradiance intensity.

3.2. MPPT improvement by combination of proposed neural network with genetic algorithm

ANN is the most suitable method for the forecasting of nonlinear systems. Non-linear systems can beapproximated by multi-layer neural networks, and these multi-layer networks have better outcome incomparison to other methods [16]. In this paper, feed forward neural network for MPPT processcontrol is implemented. The major part of ANN is that, the desired data for training process shouldbe achieved for each PV system and each particular position [20, 21]. Based on the PV characteristicswhich depend on PV model and climate changes, neural network should be trained periodically.

Three layers can be considered for the proposed ANN. The input variables are temperature and solarirradiance, and Vmpp corresponding to MPP is the output variable of ANN as depicted in Figure 2.Furthermore, a block diagram of the proposed MPPT scheme is displayed in the Figure 3.

The output characteristic of arrays have changed over time and environmental conditions. Thus,periodic training of the neural network in order to increase precision is essential. Training of theANN is a set of 390 data as demonstrated in Figure 4 (irradiance between 0.05Watts per square meter(W/m2) to 1W/m2 and temperatures between �5 °C and 55 °C), and also, a set of 390 Vmpp

corresponding to MPP is obtained by GA that is depicted in Figure 5.To perform of the ANN for MPPT, the number of layers, number of neurons in each layer,

transmission function in each layer and kind of training network should be assigned. The proposedANN in this paper has three layers which first and second layers have 15 and 12 neurons, respectivelyand third layer has 1 neurons. The first and second layers of the transfer functions are Tansig and thirdlayer is Purelin. The training function is Trainlm. The admissible sum of squares for the ANN isassigned to be 10�9. Training ANN is carried out in 800 iterations that it will converge to a required

Table II. The genetic algorithm parameters.

Number of design variable 1Population size 20Crossover constant 80%Mutation rate 10%Maximum generations 20

Figure 2. Feed forward neural network for MPPT.

Copyright © 2015 John Wiley & Sons, Ltd. Int. J. Numer. Model. (2015)DOI: 10.1002/jnm

Page 6: Wiley

Figure 3. Proposed MPPT scheme.

Figure 4. Inputs data of irradiance and temperature.

Figure 5. Output of Vmpp–MPP optimized by GA.

A. REZVANI, M. IZADBAKHSH AND M. GANDOMKAR

target. After training, the output of training network should be close to optimized output from GA. TheANN training with the target data is illustrated in Figure 6. A set of 80 data are applied for the ANNtest. The ANN test with the target data show trifling training error percentage about 0.4% as depictedin Figure 7.

4. WIND TURBINE SYSTEM CONFIGURATION

The major configuration of a wind turbine based PMSG is displayed in Figure 8. Turbine output is rec-tified by implementing uncontrolled rectifier. Then DC link voltage is adjusted by PI controller until itreaches to the constant value, and then, the constant voltage is inverted to AC voltage using sinusoidalPWM inverter. Inverter regulates the DC link voltage and injected active power by d-axis and injectedreactive power by q-axis using P/Q control method. Furthermore, turbine output is regulated via pitchangle based on FLC in high wind speeds.

Copyright © 2015 John Wiley & Sons, Ltd. Int. J. Numer. Model. (2015)DOI: 10.1002/jnm

Page 7: Wiley

Figure 6. Output of the neural network: (a) the output of the neural network with the amount of target data; (b) the out-put of the neural network of Vmpp with the amount of target data and (c) total error percentage of the Vmpp training data.

MICROGRID; PHOTOVOLTAIC; PMSG; NEURAL NETWORK; GENETIC ALGORITHM

4.1. Wind turbine and PMSG modeling

The value of electricity that turbine is capable to produce depends on the rotor speed and wind speed[45, 46]. The WT mechanical power can be expressed using equation (17):

P ¼ 0:5ρACp λ; βð ÞV3w (9)

λ ¼ WmR

Vw(10)

where P, ρ, A, Vw, Wm and R are power, air density, rotor swept area of the wind turbine, wind speedin m/s, rotor speed in rad/s and radius of turbine, respectively. Also, Cp is the aerodynamic efficiencyof rotor. PMSG voltage equations and other equations of wind turbine are presented in [35, 46].

4.2. Pitch angle based on fuzzy controller

Fuzzy logic controller is made of three parts which is demonstrated in Figure 9. The first part isfuzzification which is the process of changing a real scalar value into a fuzzy set. The second part is

Copyright © 2015 John Wiley & Sons, Ltd. Int. J. Numer. Model. (2015)DOI: 10.1002/jnm

Page 8: Wiley

Figure 7. Output of the neural network test: (a) the output of the neural network test with the amount of target data;(b) the output of the neural network test Vmpp with the amount of target data and (c) total error percentage of the

Vmpp test data.

Figure 8. Block diagram of the system.

A. REZVANI, M. IZADBAKHSH AND M. GANDOMKAR

fuzzy inference motor that combines IF–THEN statements based on fuzzy principle and, finally, it hasdefuzzification which is the process that changes a fuzzy set into a real value in output [38].

Copyright © 2015 John Wiley & Sons, Ltd. Int. J. Numer. Model. (2015)DOI: 10.1002/jnm

Page 9: Wiley

Figure 9. Structure of the fuzzy system.

MICROGRID; PHOTOVOLTAIC; PMSG; NEURAL NETWORK; GENETIC ALGORITHM

The proposed fuzzy logic controller comprises of two input and one output signals. The first input sig-nal is based on the difference between measurement active power and the nominal value in (P.U.) that iscalled as error signal. Therefore, the positive value shows turbine’s normal operation, and thenegative value depicts the extra power generation during the above nominal speed. Also, the fuzzycontroller must change the pitch angle degree by increasing the rated value. The pitch angle degreeis adjusted on zero in a normal condition. The total wind energy can be transformed to mechanicalenergy. The pitch angle starts to increment from the zero value which wind attach angle to the bladeswill be incremented, thereby leading to decrement of aerodynamic power and decreasing the turbineoutput power. Alongside, the second signal is derived from anemometer nacelle [45, 46].

Figure 10. Membership function of fuzzy logic: (a) membership functions of active power (error signal), (b) mem-bership functions of wind speed and (c) membership functions of output (β).

Copyright © 2015 John Wiley & Sons, Ltd. Int. J. Numer. Model. (2015)DOI: 10.1002/jnm

Page 10: Wiley

A. REZVANI, M. IZADBAKHSH AND M. GANDOMKAR

The responses of fuzzy controller will be faster, smoother and more stable while wind speed is appliedas an input signal comparing to the rotor speed and reactive power in wind turbines [32–34]. Although,mechanical fatigues in wind turbines will be alleviated by regulating the fuzzy logic controller. Sketchinga pitch angle based on fuzzy logic controller for wind turbine power regulation in high wind speeds is be-ing presented in this paper. Three gaussian membership functions are applied in this paper. Moreover, theMin–Max method is applied as a defuzzification reference mechanism for centroid. The given member-ship functions are depicted in Figure 10. The three-dimensional curve in fuzzy logic is shown in Figure 11.

In addition, the rules applied to get the needed pitch angle (β) are depicted in Table III. The linguis-tic variables are shown by VG (very great), SG (small great), OP (optimum), SL (small low) and VL(very low) for error signal and VL (very low), SL (small low), OP (optimum), SH (small high) and VH(very high) for wind speed signal and NL (negative large), NS (negative small), Z (zero), PS (positivesmall) and PL (positive large) for output signal, respectively.

5. MTG SYSTEM CONFIGURATION

In Figure 12, the simulation of a single shaft MTG is illustrated. The model includes the speed governor,acceleration control block, temperature control and fuel system control. In [36, 37], MTG details arereported. The power generator is a PMSG, which has two poles and a salient pole rotor. The nominaloutput power is generated by MTG is 25 kW. The rated design speed of the generator is 66 000 rpm.

6. FLYWHEEL ENERGY STORAGE SYSTEM (FESS)

Flywheel has partly fast responses in comparison to other kinds of storage device. Also, flywheel is totallyeffective while there is an imbalance between supply and demand. InMG, the flywheel can handle the powerdemands of the peak load and save the energy at the low load period. The flywheel can chip in to the stabilityof MG voltage amplitude and frequency. For providing instantaneous power desired, the flywheel is con-nected at the DC bus by droop controller. In this article, the storage device utilized is a FESS connectedto the voltage source inverter (VSI). The details of the model of flywheel are mentioned in [38].

Figure 11. Three-dimensional curve in fuzzy logic.

Table III. Fuzzy rules.

Pitch command Active power (error)

Wind speed VG SG OP SL VLVL PL PS Z Z ZSL PL PS Z Z ZOP PL PS Z Z ZSH PL PS PS PS PSVH PL PL PL PL PL

Copyright © 2015 John Wiley & Sons, Ltd. Int. J. Numer. Model. (2015)DOI: 10.1002/jnm

Page 11: Wiley

Figure 12. Simulink implementation of microturbine model.

MICROGRID; PHOTOVOLTAIC; PMSG; NEURAL NETWORK; GENETIC ALGORITHM

7. CONTROL STRATEGY

7.1. P–Q control strategy

Inverter control model has been illustrated in Figure 13. The goal of controlling the grid side is keepingthe dc link voltage in a constant value regardless of production power magnitude. In grid-connectedmode MG must supply local needs to decrease power from the main grid.

One of the main aspects of P–Q control loop is applied in grid connected and stand-alone mode.Higher power reliability and higher power quality are the advantages of this operation mode [47]. idand iq, are active and reactive components of the injected current, respectively. iq current referenceis generally set to zero in order to obtain zero phase angle between voltage and current and so unitypower factor can be attained. For the autonomous controls of both id and iq, the decoupling termsare employed. To synchronize the converter with grid, a three-phase lock loop (PLL) is used. PLL re-duces the difference between grid phase angle and the inverter phase angle to zero using PI controller,thereby synchronizing the line side inverter with the grid.

7.2. Droop control strategy

The VSI is to be coupled with a storage device (Flywheel) to balance load and generation duringislanded operation. Its control is performed using droop concepts [40]. The proposed control strategy

Figure 13. The P–Q control model.

Copyright © 2015 John Wiley & Sons, Ltd. Int. J. Numer. Model. (2015)DOI: 10.1002/jnm

Page 12: Wiley

A. REZVANI, M. IZADBAKHSH AND M. GANDOMKAR

of V/F employs the voltage and current based on conventional V/F droop control which is shown inFigure 14. During islanded operation, when the unbalance of active power and reactive power occur,the frequency and voltage will fluctuate. As a result, the MG will experience a blackout without anyeffective controller. If the system is transferred to the islanded mode when importing power fromthe grid, then the generation needs to increase power to balance power in the islanded mode. Thenew operating point (B) will be set at a frequency (f1) lower than the nominal value (f0). If the systemis transferred to the islanded mode when exporting power to the grid, then the new frequency (f2) willbe higher [42]. Also, the reactive power is injected when voltage (V1) falls from the nominal value (V0)and absorbs the reactive power if the voltage (V2) rises above its nominal value.

7.3. Back up controller

FESS as one of the storage devices has high capacities for injecting power during islanding mode; however,one of the disadvantages is a limited storage capacity. Consequently, it is required the complementary sourcewith appropriate controller to decrement the frequency fluctuation [42]. The structure of back up controller isdepicted in Figure 15. In this paper, MTG is applied for compensating the frequency deviations.

8. SIMULATION RESULTS

In this section, simulation results under different terms of operation in MG are presented usingMatlab/Simulink. System block diagram is shown in Figure 16. The grid voltage and frequency are220V and 60Hz, respectively. Detailed model descriptions are given in Appendix A. In Figures 17, 18and 19, PV, wind system and MTG connected to grid by applying P–Q controller can be seen.Figure 20 also shows FESS connected to grid by applying droop controller.

8.1. Case study 1

In this section, the variation of wind speeds, irradiance, temperature and load fault analysis of MGconnected to the grid is investigated. It is noted that the Sensitive Loads (SLs) are not connected inMG, and DG sources feed only the Non Sensitive Load (NSL). The amount of NSL is 75 kW.

(a)

(b)

Figure 14. Droop control: (a) frequency-droop characteristic; (b) voltage-droop characteristic.

Copyright © 2015 John Wiley & Sons, Ltd. Int. J. Numer. Model. (2015)DOI: 10.1002/jnm

Page 13: Wiley

Figure 15. Back up controller.

Figure 16. Case study system.

Figure 17. PV system connected to grid by applying P–Q controller.

MICROGRID; PHOTOVOLTAIC; PMSG; NEURAL NETWORK; GENETIC ALGORITHM

The MG includes 6-kW photovoltaic system, 88-kW wind turbine system, 25-kW MTG and 25-kWFESS. The P/Q and droop control technique are implemented in MG. The results obtained from PVsystem are illustrated in Figure 21. The main objective of PV system is the comparative study of MPPT

Copyright © 2015 John Wiley & Sons, Ltd. Int. J. Numer. Model. (2015)DOI: 10.1002/jnm

Page 14: Wiley

Figure 19. MTG system connected to grid by applying P–Q and backup controller.

Figure 18. Wind system connected to grid by applying P–Q controller.

Figure 20. Flywheel system connected to grid by applying droop controller.

A. REZVANI, M. IZADBAKHSH AND M. GANDOMKAR

Copyright © 2015 John Wiley & Sons, Ltd. Int. J. Numer. Model. (2015)DOI: 10.1002/jnm

Page 15: Wiley

Figure 21. Simulated results for PV in case 1: (a) irradiance; (b) output voltage of PV (after filter); (c) output cur-rent of PV (after filter) and (d) out power of PV system.

MICROGRID; PHOTOVOLTAIC; PMSG; NEURAL NETWORK; GENETIC ALGORITHM

algorithms under variations of irradiance. Different irradiance levels, according to Figure 21(a)evaluate the PV’s performance. The output voltage and the current of PV are depicted in Figures 21(b) and 21(c). Figure 21(c) illustrates the current of PV system. When irradiance is decreased at

Copyright © 2015 John Wiley & Sons, Ltd. Int. J. Numer. Model. (2015)DOI: 10.1002/jnm

Page 16: Wiley

Figure 22. Simulated results for wind system in case 1: (a) wind speed; (b) variation of pitch angle withpresence of fuzzy controller; (c) turbine output power with absence of controller; (d) turbine output powerwith presence of fuzzy controller; (e) inverter output current with absence of controller; (f) inverter outputcurrent with presence of controller; (g) THD(%); (h) DC link voltage; (i) inverter output voltage and (j)

reactive power.

A. REZVANI, M. IZADBAKHSH AND M. GANDOMKAR

Copyright © 2015 John Wiley & Sons, Ltd. Int. J. Numer. Model. (2015)DOI: 10.1002/jnm

Page 17: Wiley

MICROGRID; PHOTOVOLTAIC; PMSG; NEURAL NETWORK; GENETIC ALGORITHM

t = 1 s and t = 3 s, it causes to decrease the output current of PV. It is worth to mention that theevaluation of the proposed controller is compared and analyzed with the P&O, IC and fuzzy logicalgorithms. It can be derived that the proposed ANN-GA method has smoother power signal line,

Figure 22. Continued.

Copyright © 2015 John Wiley & Sons, Ltd. Int. J. Numer. Model. (2015)DOI: 10.1002/jnm

Page 18: Wiley

A. REZVANI, M. IZADBAKHSH AND M. GANDOMKAR

better stable operating point, better performance and more accuracy for operating at MPP than of P&O,IC and fuzzy logic methods as demonstrated in Figure 21(d).

Figure 23. Simulated results for grid in case 1: (a) grid voltage;(b) grid current with absence of fuzzy controller;(c) grid current with presence of fuzzy controller; (d) active powers with absence of fuzzy controller and (e) active

powers with presence of fuzzy controller.

Copyright © 2015 John Wiley & Sons, Ltd. Int. J. Numer. Model. (2015)DOI: 10.1002/jnm

Page 19: Wiley

MICROGRID; PHOTOVOLTAIC; PMSG; NEURAL NETWORK; GENETIC ALGORITHM

Moreover, in this case, during 0< t< 1 s, the load power is 75 kW, and at t = 1 s, it has 40% stepincrease in load. Wind speed during 0< t<1 s is 11m/s and at t = 2.3 s it is declined to 9m/s. After-ward, during 1< t< 2.3 s, wind speed is 9m/s, and at t = 3.8 s, it is sorely increment to 16m/s. Usingfuzzy logic controllers, when wind speed is more than rated value (12m/s), turbine output power isincremented by sorely incrementing wind speed; although, without fuzzy controller, the power is keptin high level and by using fuzzy logic controller, it is declined to rated power and it was madesmoother, which leads to the prevention of mechanical damages to PMSG.

Figure 22(a) illustrates variations of the wind speed. Figure 22(b) depicts the variations of pitch an-gle in the presence of fuzzy logic controller. In normal condition, pitch angle is adjusted as zero.Figures 22(c) and 22(d) depict the turbine output power in the absence and presence of fuzzy controlleraccording to wind speed. It is clear that, by using fuzzy logic controller, the power curves are smoother.By incrementing the pitch angle degree using fuzzy logic controller, the extra power of wind turbine islimited and reached the rated value. Figures 22(e) and 22(f) show inverter output current in the absenceand presence of fuzzy logic controller, respectively. It depicts the effectiveness of fuzzy logic controllerby incrementing the pitch angle degree. The extra power of wind turbine is more limited and also, theinverter output current is more declined in comparison to PI controller. According to IEEE

Figure 24. Wind system: (a) pitch angle (deg); (b) active power and (c) voltage at terminal (Bus 3).

Copyright © 2015 John Wiley & Sons, Ltd. Int. J. Numer. Model. (2015)DOI: 10.1002/jnm

Page 20: Wiley

A. REZVANI, M. IZADBAKHSH AND M. GANDOMKAR

Std.1547.2003, total harmonic distortion (THD) should be around 5%. In THD curve, it is around 1.5%to 2.5% as illustrated Figure 22(g). DC link voltage remains at a constant value (1050V), which provesthe effectiveness of the established fuzzy controller as displayed in Figure 22(h). Inverter output volt-age as demonstrated in Figure 22(i) is constant. The reactive power produced by wind turbine is ad-justed at zero to keep the power factor as unity as depicted in Figure 22(j). Figures 23(a) and 23(b)illustrate the grid current in the absence and presence of fuzzy controller, respectively. It can be derivedfrom Figures 23(c) and 23(d), that pitch angle based on fuzzy logic controller can limit the extra outputpower of turbine. Then, by the decreasing of the injected output power of wind turbine, the injection ofextra total active power of MG to grid is declined. It is clear that the grid, with the cooperation of wind,PV, MTG systems and FESS, can easily meet the load demand.

8.2. Case study 2

A three-line-to-ground (3LG) fault as network disturbance occurs at the grid. The main objective ofthis section is investigating the MG from grid connected state to the islanding mode. It is supposed thatNSL is not connected in MG and the DG sources feed only SLs. The MG is importing around 15 kWand 11kvar from the upstream MV network, with a local generation of 93 kW and 5kvar and an MGload of 108 kW and 16kvar. Depending on the load, real and reactive power is defined. The VSI is usedto interface the flywheel (storage device) to the MG during and subsequent to islanding occurrence.

The fault events to the system at t = 5 s, which leads to islanding of MG. Moreover, wind power hashigh fluctuations, which causes high deviations in frequency, active and reactive power injected by theVSI, and the voltages of the MG buses. It worth to mention that, FLC operates when wind speed ismore than rated value. The variations of pitch angle (using PI and FLC) are depicted in Figure 24(a)which FLC causes to have more smooth output power and bus voltage of wind generation as illustratedin Figures 24(b) and 24(c), respectively. Smoothing the output power of wind system leads to smoothactive and reactive power injected by the VSI, frequency which shown in Figures 25(a), 25(b) and 26.

Figure 25. Flywheel: (a) active power and (b) reactive power.

Copyright © 2015 John Wiley & Sons, Ltd. Int. J. Numer. Model. (2015)DOI: 10.1002/jnm

Page 21: Wiley

Figure 26. Frequency variation.

Figure 27. Generated active power: (a) MTG; (b) irradiance of PV system and (c) output power of PV system.

MICROGRID; PHOTOVOLTAIC; PMSG; NEURAL NETWORK; GENETIC ALGORITHM

By implementing FLC, the output power of FESS and MTG have higher numerical value and lessoscillation than PI controller. By incrementing the pitch angle of WT using FLC, it leads to smoothingthe output power of FESS and MTG as demonstrated in Figures 25 and 27(a), respectively.

For realistic conditions, it is needed to analyze the noises in a PV system. The performance of theANN controller in PV is compared and analyzed with the conventional techniques such as P&O, ICand fuzzy logic when operating during a cloudy day with rapid irradiance changes. Irradiance of the

Copyright © 2015 John Wiley & Sons, Ltd. Int. J. Numer. Model. (2015)DOI: 10.1002/jnm

Page 22: Wiley

A. REZVANI, M. IZADBAKHSH AND M. GANDOMKAR

PV system is illustrated in Figure 27(b). According to Figure 27(c) the proposed MPPT algorithm cantrack accurately the MPP when the irradiance changes continuously; also, this method has well-regulated PV output power, and it produces extra power rather than aforementioned method. It is worthto mention that the evaluation of the proposed controller shows the better performance in severeconditions than the aforementioned methods. Performance of systems by using proposed method isdesirable even with limited changes in system parameters. However, using conventional methods leadsto wide error in these conditions.

9. CONCLUSION

Detailed dynamic modeling of MG, including WT, PV, MTG, FESS and grid level control strategies,was investigated. Control strategy and precise modeling of DC/AC grid connected converter werepresented. Inverter adjusted the DC link voltage and active power was fed by d-axis and reactive powerwas fed by q-axis using P–Q control method in grid connected mode. When MG operated as anislanding mode, droop control via FESS had to regulate the voltage value at the PCC and also thefrequency of the whole grid.

Based on the PV characteristics which depend on PV model and climate changes, neural networkshould be trained periodically, but one of the main features of ANN-GA controller can dramatically de-crease the weaknesses of the conventional methods. Actually, the proposed method shows smootherpower, less oscillation and better stable operating point than P&O, IC and fuzzy logic methods. Itproduces exceeding power, and it has faster dynamic response rather than mentioned techniques. More-over, the proposed fuzzy logic controller in the wind turbine, by adding wind speed as an input signalof fuzzy logic, could have faster and smoother response. The advantage of fuzzy logic controller is thatit remains the turbine output power in an acceptable value and can prevent more mechanical damages,and, also, the dynamic performance of PMSG can be enhanced. In other words, by increasing pitch angledegree by fuzzy logic controller, the extra power of wind turbine is limited, reaching the rated value anddecreasing inverter output current. Also, by the decreasing of injected output power of wind turbine, theinjection of extra total active power of MG to grid is declined. Finally, by implementing the suitable con-troller, the MG in grid connected and islanding modes could meet the load demand certainly.

APPENDIX

Description of the detailed model

Photovoltaic parameters: output power = 6 kW, carrier frequency in VMPPT PWM generator: 4 kHz andin grid-side controller: 5.5 kHz, boost converter parameters: L=7mH, C=1100μF, PI coefficients ingrid-side controller: KpVdc = 2, KiVdc = 9, KpId = 8 KiId = 300, KpIq = 8, KiIq = 300.

PMSG parameters: output power: 88 kW, stator resistance per phase =2.7Ω, inertia: 0.9e�3 kg-m2,torque constant 12N-M/A, pole pairs = 8, nominal speed =12m/s, Ld =La= 8.9mH. Grid parameters:X/R=7, and other parameters, DC link capacitor = 5300μF, DC link voltage = 1050V. PI coefficientsin grid-side controller: KpVdc = 8, KiVdc = 400, KpId = 0.83, KiId = 5, KpIq = 0.83, KiIq = 5.

MTG parameters: MTG ratings: 25 kW, Rotor speed: 66 000 rpm, T1 = 0.4, T2= 1, K=25. FESSparameters: output power = 25 kW, J= 0.07 kgm2, L=8mH.

REFERENCES

1. Rezvani A, Gandomkar M,Izadbakhsh M, Ahmadi A. Envi-ronmental/economic scheduling ofa micro-grid with renewable energyresources. J Cleaner Production2015; 87:216–226.

Copyright © 2015 John Wiley & Sons, Ltd

2. Izadbakhsh M, Gandomkar M,Rezvani A, Ahmadi A. Short-termresource scheduling of a renewableenergy based micro grid. RenewEnergy 2015; 75:598–606.

3. Mirsaeidi S, Said DM, MustafaMW, Habibuddin MH, GhaffariK. Progress and problems inmicro-grid protection schemes.

.

Renew Sustain Energ Rev 2014;37:834–839.

4. Mirsaeidi S, Said DM, MustafaMW, Habibuddin MH, GhaffariK. An analytical literature reviewof the available techniques for theprotection of micro-grids. Int JElectr Power Energ Syst (IJEPES)2014; 58:300–306.

Int. J. Numer. Model. (2015)DOI: 10.1002/jnm

Page 23: Wiley

MICROGRID; PHOTOVOLTAIC; PMSG; NEURAL NETWORK; GENETIC ALGORITHM

5. Villalva MG, Gazoli JR, Filho E.Comprehensive approach to model-ing and simulation of photovoltaicarrays. IEEE Trans Power Electron2009; 24(5):1198–1208.

6. Kuo YC, Juu Liang JT, ChenJF. Novel maximum-power-point-tracking controller for photovoltaicenergy conversion system. IEEETrans Ind Electron 2001; 48(3):594–601.

7. Eldahab YE, Saad NH, Zekry A.Enhancing the maximum powerpoint tracking techniques for pho-tovoltaic systems. Renew SustainEnergy Rev 2014; 40:505–514.

8. Abd E-S, Nafeh A, Fahmy FH,Abou El-Zahab EM. Maximum-power operation of a stand-alonePV system using fuzzy logic con-trol. Int J Numer Model ElectronNetwork Dev Field 2002; 15(4):385–398.

9. Bouchafaa F, Hamzaoui I,Hadjammar A. Fuzzy Logic Con-trol for the tracking of maximumpower point of a PV system. EnergProcedia 2011; 6(1):152–159.

10. Hong CM, Chen CH. Intelligentcontrol of grid-connected wind-photovoltaic hybrid power systems.Electr Power Energ Syst 2014;55:554–561.

11. Rai AK, Kaushika ND, Singh B,Agarwal N. Simulation model ofANN based maximum power pointtracking controller for solar PVsystem. Sol Energ Mater Sol Cell2011; 95(2):773–778.

12. Hong CM, Ou TC, Lu KH. Devel-opment of intelligent MPPT (maxi-mum power point tracking) controlfor a grid-connected hybrid powergeneration system. Energy 2013;50:270–279.

13. Fangrui L, Duan S, Liu F, Liu B. Avariable step size INC MPPTmethod for PV systems. IEEETrans Ind Electron 2008; 55(7):2622–2628.

14. Esram T, Chapman PL. Compari-son of photovoltaic array maximumpower point tracking techniques.IEEE Trans Energ Convers 2007;22(2):439–449.

15. Simoes MG, Franceschetti NN.Fuzzy optimization based controlof a solar array system. Elec PowerAppl 1999; 14(5):552–558.

16. Lee S, Kim J, Cha H. Design andimplementation of photovoltaicpower conditioning system using acurrent-based maximum powerpoint tracking. J Electr EngTechnol 2010; 5(4):606–613.

17. Hiyama T, Kouzuma S, Imakubo T,Ortmeyer TH. Evaluation of neuralnetwork based real time maximumpower tracking controller far PV sys-tem. IEEE Trans Energ Convers1995; 10(3):543–548.

Copyright © 2015 John Wiley & Sons, Ltd

18. Hiyama T, Kitabayashi K. Neuralnetwork based estimation of maxi-mum power generation from PVmodule using environment infor-mation. IEEE Trans Energ Convers1997; 12(3):241–247.

19. Karatepe E, Boztepe M, Olak MC.Neural network based solar cellmodel. Energ Convers Manag2006; 47(9):1159–1178.

20. Rezvani A, Gandomkar M,Izadbakhsh M, Gandoman FH,Vafaei S. Comparative study ofANN-GA and fuzzy controller forphotovoltaic system in the gridconnected mode. SOP Trans onPower Trans Smart Grid 2014; 1(1):29–43.

21. Vincheh MR, Kargar A,Markadeh GHA. A hybrid controlmethod for maximum powerpoint tracking (MPPT) in photo-voltaic systems. Arab J Sci Eng2014; 39:4715–4725.

22. Ramaprabha R, Gothandaraman V,Kanimozhi K, Divya R, MathurBL. Maximum power point track-ing using GA-optimized artificialneural network for solar PV sys-tem. Electr Energ Syst (ICEES),NewportBeach, Calif, USA 3–5Jan 2011, 264–268.

23. Hayatdavudi M, SaeedimoghadamM, Nabavi MH. Adaptive controlof pitch angle of wind turbine usinga novel strategy for management ofmechanical energy generated byturbine in different wind velocities.J Electr Eng Technol 2013; 8(4):863–871.

24. Izadbakhsh M, Gandomkar M,Rezvani A, Vafaei S. Comparisonof FLC-GA-PI methods to smooththe output power of wind turbinein the grid connected mode. SOPTrans Power Trans Smart Grid2014; 1(1):44–59.

25. Li H, Chen Z. Overview of differentwind generator systems and theircomparisons. IET Renew PowerGenerat 2007; 2(2):123–138.

26. Joo Y, Back J. Power regulation ofvariable speed wind turbines pitchcontrol based on disturbance ob-server. J Electr Eng Technol 2012;7(2):273–280.

27. Simoes MG, Bose BK, Spiegel RJ.Fuzzy logic based intelligent con-trol of a variable speed cage ma-chine wind generation system.IEEE Trans Power Electron 1997;12(1):87–95.

28. Muhandoa EB, Senjyua T, KinjobH, Funabashi T. Augmented LQGcontroller for enhancement of on-line dynamic performance forWTG system. Renew Energy2008; 33:1942–1952.

29. Yuan Lo K, Chen Y, Chang Y.MPPT battery charger for stand-alone wind power system. IEEE

.

Trans Power Electron 2011; 26(6):1631–1638.

30. Cheung JYM, Kamal AS. Fuzzylogic control of refrigerant flow. IntConf Contr 1996; 1(427):125–130.

31. GauravN,Kaur A. Performance eval-uation of fuzzy logic and PID control-ler by using MATLAB/Simulin. Int JInnovat Technol Exploring Eng(IJITEE) 2012; 1(1):84–88.

32. Lingfeng X, Xiyun Y, Xinran L,Daping X. Based on adaptive fuzzysliding mode controller. Intelligentcontrol and automation WCICA7th World Congress China 2008,2970–2975.

33. Amendola CAM, Gonzaga DP.Fuzzy-logic control system of avariable-speed variable-pitch wind-turbine and a double-fed inductiongenerator. Intelligent Systems De-sign andApplications. Seventh Inter-national Conference, Brazil 2007:252–257.

34. Senjyu T, Sakamoto R, Urasaki N,Funabashi T, Sekine H. Outputpower leveling of wind farm usingpitch angle control with fuzzy neu-ral network. In: IEEE 2006 PowerEngineering Society General Meet-ing Japan 2006: 1–8.

35. Van TL, Lee DCH. Output powersmoothening of variable-speedwind turbine systems by pitch an-gle control. Conference on Power& Energy, Ho Chi Minh City China2012: 166–171.

36. Gaonkar DN, Patel RN, Pillai GN.Dynamic model of microturbine ge-neration system for grid-connected/islanding operation. IEEE Int Conf2006:305–310.

37. Pai F. An improved utility interfacefor microturbine generation systemwith stand-alone operation capabil-ities. IEEE Trans Industr Electron2006; 53(5):1529–1537.

38. Sumathi R, Usha M. Pitch and yawattitude control of a rocket engineusing hybrid fuzzy-pid controller.Open Autom Contr Syst J 2014;6(1):29–39.

39. Chen MJ, Wu YC, Huang YX. Dy-namic behavior of a grid-connectedmicrogrid with power conditioningsystem. Int J Numer Model Elec-tron Network Dev Field 2014;27:318–333.

40. PLopes JA, Moreira CL, MadureiraAG. Defining control strategies forMicroGrids islanded operation. IEEETrans Power Syst 2006; 21(2):916–924.

41. Katiraei F, Irvani M, Lehn P.Micro-grid autonomous operationduring and subsequent to islandingprocess. IEEE Trans Power Del2005; 20(1):248–257.

42. Kamel RM, Chaouachi A, NagasakaK. Detailed analysis of micro-gridstability during islanding mode

Int. J. Numer. Model. (2015)DOI: 10.1002/jnm

Page 24: Wiley

A. REZVANI, M. IZADBAKHSH AND M. GANDOMKAR

under different load conditions. En-gineering 2011; 3(5):508–516.

43. Moradian M, Tabatabaei FM,Moradian S.Modeling, control & faultmanagement of microgrids. SmartGrid RenewEnerg 2013; 4(1):99–112.

44. Yang J, Honavar V. Feature subsetselection using a genetic algorithm.IEEE Intell Syst 1998; 13(2):44–49.

Copyright © 2015 John Wiley & Sons, Ltd

45. Hyun SH,Wang J. Pitch angle controland wind speed prediction methodusing inverse input–output relation ofa wind generation system. J ElectrEng Technol 2013; 8(5):1040–1048.

46. Rosyadi M, Muyeen SM, TakahashiR, Tamura J. Transient stability en-hancement of variable speed perma-nent magnet wind generator using

.

adaptive pi-fuzzy controller. PowerTech. Conf, Trondheim Norway2011, 1–6.

47. Blaabjerg F, Teodorescu R, LiserreM, Tim-bus AV. Overview of con-trol and grid synchronization fordistributed power generation sys-tems. IEEE Trans Ind Electron2006; 53(5):1398–1409.

Int. J. Numer. Model. (2015)DOI: 10.1002/jnm