lable at ScienceDirect

Energy 64 (2014) 868e874

Contents lists avai

Energy

journal homepage: www.elsevier .com/locate/energy

Adaptive neuro-fuzzy maximal power extraction of wind turbinewith continuously variable transmission

Dalibor Petkovi�c a, �Zarko �Cojba�si�c a, Vlastimir Nikoli�c a, Shahaboddin Shamshirband b,*,Miss Laiha Mat Kiah b, Nor Badrul Anuar b, Ainuddin Wahid Abdul Wahab b

aUniversity of Ni�s, Faculty of Mechanical Engineering, Department for Mechatronics and Control, Aleksandra Medvedeva 14, 18000 Ni�s, SerbiabDepartmentofComputer SystemandTechnology, FacultyofComputerScienceand InformationTechnology,UniversityofMalaya, 50603KualaLumpur,Malaysia

a r t i c l e i n f o

Article history:Received 19 June 2013Received in revised form28 October 2013Accepted 30 October 2013Available online 26 November 2013

Keywords:Wind turbinePower coefficientContinuously variable transmissionIntelligent controlANFIS controller

* Corresponding author. Tel.: þ60 146266763.E-mail addresses: dalibortc@gmail.com (D. Petko

(S. Shamshirband).

0360-5442/$ e see front matter � 2013 Elsevier Ltd.http://dx.doi.org/10.1016/j.energy.2013.10.094

a b s t r a c t

In recent years the use of renewable energy including wind energy has risen dramatically. Because of theincreasing development of wind power production, improvement of the control of wind turbines usingclassical or intelligent methods is necessary. To optimize the power produced in a wind turbine, thespeed of the turbine should vary with the wind speed. Variable-speed operation of wind turbines pre-sents certain advantages over constant-speed operation. In this paper, in order to maintain the maximaloutput power of wind turbine, a novel intelligent controller based on the adaptive neuro-fuzzy inferencesystem (ANFIS) is designed. To improve the wind energy available in an erratic wind speed regime, awind generator equipped with continuously variable transmission (CVT) was proposed. In this model theANFIS regulator adjusts the system speed, i.e. CVT ratio, for operating at the highest efficiency point. Theperformance of proposed controller is confirmed by simulation results. Some outstanding properties ofthis new controller are online implementation capability, structural simplicity and its robustness againstany changes in wind speed and system parameter variations. Based on the simulation results, theeffectiveness of the proposed controllers was verified.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

Renewable energies such as wind and solar energy conversionsystems have driven attention during the past decade due to theenvironmental concerns. Wind is a natural resource that featuresmany advantages since it is clean and considered reliable in someareas.

A wind turbine system is a system that converts the wind tur-bine’s mechanical energy obtained fromwind into electrical energythrough a generator and can be categorized by the types of gen-erators used, power control methods, constant- or variable-speedoperation, and methods of interconnecting with the grid [1].Variable-speed operation of a wind turbine is generally more ad-vantageous over constant-speed operation since a variable-speedoperation is able to track the maximum power of the wind tur-bine with wind speed changes. Modern high-power wind turbinesare equipped with adjustable speed generators [2]. It was shownthat the control strategies have a major effect on the wind turbine

vi�c), shahab1396@gmail.com

All rights reserved.

and whatever the kind of the wind turbine, the control strategyremains a key factor [3e6].

As wind energy becomes more dominant there is growing in-terest in controlling wind turbines or wind plants in an intelligentmanner to minimize the cost of wind energy. This can be done bycontrolling the turbines to extractmore energy from thewind. In thewind energy conversion systems, the control problem consists ofdelivering the maximum power available from the wind to ensurethe system reliability and security in order to deal with the variablenature of the generated energy [7e9]. Wind power conversion de-pends essentially on the power coefficient, “Cp” of the machinewhich transforms the efficiency of converting wind power to elec-trical power. In order to implement maximum wind power extrac-tion, the wind turbine generator must be operated at variable-speedmode. The power coefficient is characterized as a function of both tipspeed ratio and the blade pitch angle. The tip speed ratio is the ratioof linear speed at the tip of blades to the speed of the wind. Optimalperformance of thewind turbine can be obtained if the transmissionratio could change with the wind speed [10,11]. In this paper acontinuously variable transmission (CVT) has been installed be-tween a wind turbine and a generator to make the turbine operatealong the maximum efficiency. The aim of the investigation was tochange the transmission ratio between the wind turbine and the

D. Petkovi�c et al. / Energy 64 (2014) 868e874 869

generator at different wind speeds so that the turbine may be keptrunning atmaximum efficiency levels at all wind speeds. It is knownabout automatic CVT regulation to adjust and stabilize their trans-mission ratio according to transmitted torques without relying onother regulating or control mechanisms [12,13]. Considering allthese, it is interesting to explore the feasibility of installing anautomatic CVT in a wind turbine since such a solution may optimizethe efficiency of these systems by means of simple technologies.

To improve the control of the wind turbines, fuzzy logic (FL) [14e19] or artificial neural network (ANN) control has attracted much

Cpðb;Ve;Ur;RÞ ¼ 0:5176

0B@ 116

1RUr

Ve� 0:08b

� 0:035

b3 þ 1

� 0:4b� 5

1CAe

�211

RUrVe

�0:08b�0:035b3þ1 þ 0:0068

RUr

Ve

attention in recent years [20e27]. As a non-linear function [28e31],ANNs can be used for identifying the extremely non-linear systemparameters with high accuracy. Neural networks can learn fromdata. However, understanding the knowledge learned by neuralnetworks has been difficult. In contrast, fuzzy rule based models areeasy to understand because they use linguistic terms and thestructure of IF-THEN rules. Unlike neural networks, however, fuzzylogic by itself cannot learn [32]. Since neural networks can learn, it isnatural to merge these two techniques. This merged technique ofthe learning power of the ANNs with the knowledge representationof FL has created a new hybrid technique, called neuro-fuzzy net-works or adaptive neuro-fuzzy inference system (ANFIS) [33]. ANFIS,as a hybrid intelligent system that enhances the ability to auto-matically learn and adapt, was used by researchers for modeling[34e37], predictions [38e40] and control [41e45] in various engi-neering systems. The basic idea behind these neuro-adaptivelearning techniques is to provide a method for the fuzzy modelingprocedure to learn information about data [46e53].

In this paper, the application of ANFIS is proposed to control theCVT ratio to extract the maximal wind energy through the windturbine. As inputs in the controller, current wind speed and currentwind turbine rotor speed are used. The output should be optimalgenerator speed.

2. Wind turbine power extraction and continuously variabletransmission

The major components of a typical wind energy conversionsystem include a wind turbine, a generator, interconnection appa-ratus, and control system. Therefore, the design of a wind energyconversion system is complex. The most important part of a windenergy conversion system is the wind turbine transforming thewind kinetic energy into mechanical or electric energy. The systembasically comprises a blade, a mechanical part and an electric enginecoupled to each other. The kinetic energy of wind is the function ofwind speed, the specific mass of air, the area of air space where thewind is captured and the height at which the rotor is placed. Thepower available in a uniform wind field can be expressed as

Pw ¼ 12rAy3

where Pw is the power [W] of the wind with air density r [kg/m3]and wind speed n [m/s] is passing through the swept area A [m2] of

a rotor disk that is perpendicular to thewind flow. Thewind turbinecan only capture a fraction of the power available from the wind.The ratio of captured power to available power is referred to as thepower coefficient

Cpðb;Ve;Ur;RÞ

which is a function of the collective blade pitch angle b, effectivewind speed Ve, rotor speedUr and rotor radius R. The value of Cp canbe expressed according to Ref. [54] as:

A characterization of the power coefficient Cp for the windturbine used in this study is optimized to achieve maximum value.The optimization procedures is expressed as

maxCp ¼ Cpðb;Ve;Ur;RÞ45

� � b � 0�

50 m=s � Ve � 8 m=s30 rpm � Ur � 60 rpmR ¼ 25 m

(1)

In this paper a new approach to a CVT power transmission sys-tem is presented. It is added just before the generator, avoiding theneed to change the main gearbox and the aerodynamic tip brakecontrol pipes. Fig.1 shows awidely used power transmission systemof a wind turbine with the proposed CVT system installed. The po-wer flows from the rotor hub through the input shaft to the maingearbox. It is the same unit that is used in a fixed speed wind tur-bine. This gearbox could consist of a planetary stage and two simplespur gear stages. The disc brake is conventionally installed after themain gearbox. In a fixed speed design the power would flow fromthe main gearbox directly to the generator. This is the point wherethe proposed CVT system is installed. It is suggested to use for CVTsystem two spring-loaded pulleys, one at the driving shaft and oneat the driven shaft. With such a simple and inexpensive solution, theCVT was automatically regulated and adjusted its transmission ratioto the torque applied on the driving pulley. A layout of the drive traincomponents of the wind turbine is illustrated in Fig. 2.

The general speed ratio iCVT is given by:

iCVT ¼ uAuB

where uA is the angular velocity of the power input shaft and isconnected to the output shaft of the main gearbox. The uB is theangular velocity of the output shaft connected to the generator.Finally there is adjustment shaft with angular velocity uC which isconnected to the hydraulic system. Variation of the speed of theadjustment shaft leads to variation of the total transmission ratio ofthe gearbox. The angular velocities of the three shafts uA, uB, uCfulfill the following relationship:

uC ¼ x$uA þ y$uB

where x and y are constants defined by the numbers of teeth of eachgear and the overall gearbox system. In the special case where

Fig. 1. Power transmission system of a CVT regulated wind energy converter.

Fig. 2. Wind turbine drive train components.

D. Petkovi�c et al. / Energy 64 (2014) 868e874870

uC¼ 0 the angular velocities of the input and output shaft are uA�0and uB�0 and the speed ratio is

iCVT ¼ uA�0uB�0

¼ �yx

A new variable f is introduced to describe the difference of thecurrent angular velocity of the input shaft uA from uA�0.

f ¼ uA � uA�0uA�0

(2)

f¼ 0 means that uA¼uA�0. In terms of f the speed ratio of thegearbox is given by:

iCVT ¼ �yx$ð1þ fÞ ¼ iCVT�0$ð1þ fÞ

The speed ratio iR between the adjustment shaft uC and theinput shaft uA can also be expressed in terms of f:

iR ¼ uAuC

¼ 1x$1þ f

f

In quasi steady state conditions the power and moment equi-libriums of the black box is:

PA þ PB þ PC ¼ 0 (3)

TA þ TB þ TC ¼ 0 (4)

By replacing P¼u$T in Eq. (3) and combining with Eqs. (4) and(2) and adjustment power ratio can be obtained in terms of f:

PCPA

¼ � f

1þ f

The above equation directly relates the power PC required tochange the speed uA by a factor of (1þ f) to the input power PA.

Fig. 3. Block diagram of the ANFIS control scheme for the wind turbine rotor radiuscontrol.

3. ANFIS controller design

A controller is a device which controls each and every operationin a decision-making system. From the control system point of

view, it brings stability to the system when there is a disturbance,thus safeguarding the equipment from further damage. It may be ahardware-based controller or a software-based controller or acombination of both. In this section, the development of the controlstrategy for control of thewind turbine rotor radius and rotor speedis presented using the concepts of ANFIS control scheme, the blockdiagrams of both the control schemes are shown in Fig. 3. The fuzzylogic controller provides an algorithm, which converts the lin-guistic control, based on expert knowledge, into an automaticcontrol strategy. Linguistic variables, defined as variables whosevalues are sentences in a natural language (such as large or small),may be represented by the fuzzy sets. A fuzzy set is an extension ofa ‘crisp’ set where an element can only belong to a set (full mem-bership) or not belong at all (no membership). Fuzzy sets allowpartial membership, which means that an element may partiallybelong to more than one set. Therefore, the fuzzy logic algorithm ismuch closer in spirit to human thinking than traditional logicalsystems. The main problem with the fuzzy logic controller gener-ation is related to the choice of the regulator parameters. For thisreason, we apply the ANFIS methodology to adapt the parametersof the fuzzy controller according to real data about the problem.

Fig. 5. A trapezoidal membership function.

Table 1ANFIS input parameters.

Wind speed [m/s] Rotor speed [rpm]

8 3610 3612 4014 4716 54

D. Petkovi�c et al. / Energy 64 (2014) 868e874 871

The ANFIS structure is tuned automatically by least-squareestimation and the back propagation algorithm. ANFIS controllersin general have six modules, which are as follows.

1. Preprocessing: during preprocessing different values of effectivewind speed and rotor speed are chosen.

2. Fuzzification: process by which a particular input is rated interms of its belongingness to a certain membership function(MF). Here, the crisp variables were converted into fuzzy vari-ables or the linguistic variables.

3. Fuzzy inference engine: values of linguistic variables are acquiredusing fuzzification, which makes it easier to implement. Duringthe second fuzzification stage, the set point is converted intolinguistic variables.

4. Rule base: a rule base was acquired for the functioning of fuzzycontroller. The developed fuzzy rules were obtained during theconstruction of ANFIS controller.

5. Defuzzification: done on the output data that is achieved afterthe data has been passed through the module.

6. Post-processing: the output of the input fuzzy system here wasthe prescribed numerical optimal CVT ratio and obtained poweroutput of the wind turbine.

Fig. 4 shows an ANFIS structure for two inputs, effective windspeed and rotor speed and for one output, the optimal wind turbineCVT ratio. According to these inputs, and the training input/outputdata pairs, the ANFIS network could make decision for to achievemaximal wind energy conversion from the wind turbine. Traininginput/output data pairs were collected from the presented opti-mization procedures (1) and from CVT ratio related expressions.

In this work, the first-order Sugeno model with two inputs andfuzzy IF-THEN rules of Takagi and Sugeno’s type is used:

if x is A and y is C then f1 ¼ p1xþ q1yþ r1

The first layer consists of input variables (MFs), input 1 and input2. This layer just supplies the input values to the next layer. In thefirst layer every node is an adaptive node. In this study, triangleMFswith maximum equal to 1 and minimum equal to 0 are chosen(Fig. 5), such as

mðxÞ ¼ triangleðx; ai; bi; ciÞ ¼

8>>>>>>>>><>>>>>>>>>:

0; x � ai

x�aibi�ai

; ai � x � bi

ci�xci�bi

; bi � x � ci

0; ci � x

where {ai, bi, ci} is the set of parameters set that in this layer arereferred to as premise parameters. In this layer x and y are the

Fig. 4. ANFIS structure.

inputs to nodes and they are effective wind speed and rotor speed.Table 1 summarizes relation between effective wind speed androtor speed used in this study as the ANFIS inputs.

The second layer (membership layer) checks for the weights ofeachMFs. It receives the input values from the first layer and acts asMFs to represent the fuzzy sets of the respective input variables.Every node in the second layer is non-adaptive and this layermultiplies the incoming signals and sends the product out likewi¼ m(x)*m(y). Each node output represents the firing strength of arule.

The third layer is called the rule layer. Each node (each neuron)in this layer performs the pre-condition matching of the fuzzyrules, i.e. they compute the activation level of each rule, the numberof layers being equal to the number of fuzzy rules. Each node ofthese layers calculates the weights which are normalized. The thirdlayer is also non-adaptive and every node calculates the ratio of therule’s firing strength to the sum of all rules’ firing strengths likew*

i ¼ wi=w1 þw2; i ¼ 1;2;. The outputs of this layer are callednormalized firing strengths.

The fourth layer is called the defuzzification layer and it pro-vides the output values resulting from the inference of rules. Everynode in the fourth layer is an adaptive node with node functionO4i ¼ w*

i xf ¼ w*i ðpixþ qiyþ riÞwhere {pi,qi, r} is the parameter set

and in this layer is referred to as consequent parameters.The fifth layer is called the output layer which sums up all the

inputs coming from the fourth layer and transforms the fuzzyclassification results into a crisp (binary). The single node in the

18 6020 6022 6024 5826 5328 50.630 49.432 48.534 48.536 48.538 48.540 48.542 48.544 48.546 48.548 48.550 48.5

Fig. 7. ANFIS predicted relationships for the optimal CVT ratio between (a) effectivewind speed (input 1), rotor speed (input 2) and generator speed (output) e ANFIS 2.

Fig. 8. ANFIS predicted relationships between (a) optimal CVT ratio (input 1), gener-ator speed (input 2) and wind turbine power output (output) e ANFIS 3.

D. Petkovi�c et al. / Energy 64 (2014) 868e874872

fifth layer is not adaptive and this node computes the overall outputas the summation of all incoming signals

O4i ¼

Xi

w*i xf ¼

PiwifPiwi

This type of adaptive network is functionally equivalent to atype-3 fuzzy inference system. The hybrid learning algorithmswere applied to identify the parameters in the ANFIS architectures.In the forward pass of the hybrid learning algorithm, functionalsignals go forward until Layer 4 and the consequent parameters areidentified by the least-squares estimate. In the backward pass, theerror rates propagate backwards and the premise parameters areupdated by the gradient descent.

4. Results

In this paper ANFIS training and checking data are extractedusing above mentioned optimization procedure (1) and by CVTratio related expressions. Power coefficient is used as numericalindicator for the wind turbine energy estimation.

The final decision surfaces after ANFIS training is shown inFigs. 6e8.

The wind turbine power coefficient as function of the effectivewind speed and rotor speed is implemented in MATLAB Simulinkblock diagrams as shown in Fig. 9. It shows block diagram forestimation of the optimal wind turbine CVT ratio to achievemaximal power coefficient while the rotor speed is variable. Thisapproach is very useful for fast estimation of the maximal windturbine power coefficient according to the main wind turbine pa-rameters and wind speed variation as well.

5. Conclusion

In summary, wind energy is a rapid growing industry, andthis growth has led to a large demand for better modeling andcontrol of wind turbines. The uncertainties and difficulties inmeasuring the wind inflow to wind turbines makes the controldifficult, and more advanced modeling via system identificationtechniques and a number of advanced control approaches shouldbe explored to reduce the cost of wind energy. The windresource available worldwide is large, and much of the world’sfuture electrical energy needs can be provided by wind energy

Fig. 6. ANFIS predicted relationships between (a) effective wind speed (input 1), rotorspeed (input 2) and CVT ratio (output) e ANFIS 1.

alone if the technological obstacles are overcome. The applica-tion of advanced controls for wind energy systems is still in itsinfancy, and there are many fundamental and applied issues thatcan be addressed by the systems and control community tosignificantly improve the efficiency, operation, and lifetimes ofwind turbines.

Variable-speed operation of wind turbine is necessary to in-crease power generation efficiency. The presented research workdeals with variable-speed wind control design, in order to achievethe objectives of maximizing the extracted energy from the wind.This paper has suggested coupling a wind turbine rotor to agenerator by means of a continuously variable transmission tomaximize turbine efficiency. The CVT is added just before thegenerator, avoiding the need to change the main gearbox and theaerodynamic tip brake control pipes. It allows for varying the speedof the rotor according to the current wind speedwhile retaining thespeed of the generator constant, leading to a better exploitation ofthe available wind energy potential. The implementation of thesystem does not require a new main gearbox. Instead, it can bemounted just before the generator.

Fig. 9. Simulink block diagram for estimation of the optimal wind turbine CVT ratio.

D. Petkovi�c et al. / Energy 64 (2014) 868e874 873

An intelligent controller has been suggested. The ANFIScontroller was developed in order to implement a maximum powertracking scheme for a wind turbine. This method is based on theregulation of CVT ratio. We found that under the same operatingconditions the proposed method is able to gain more power if thewind turbine is operated at variable-speed mode by the proposedscheme. As the parameter for measuring performance of the windturbine power coefficient Cp was used. Two Simulink models weredeveloped in MATLAB with the ANFIS networks. The main advan-tage of designing the ANFIS coordination scheme is to achievemaximal wind turbine power coefficient as the main turbineparameter according to optimal CVT ratio. Simulations were run inMATLAB and the results were observed on the correspondingoutput blocks. The main advantages of the ANFIS scheme arecomputationally efficient, well-adaptable with optimization andadaptive techniques. The developed strategy is not only simple, butalso reliable and may be easy to implement in real time applica-tions using some interfacing cards like the dSPACE, data acquisitioncards, NI cards, etc. for control of various parameters. This can alsobe combined with expert systems and rough sets for other appli-cations. ANFIS can also be used with systems handling morecomplex parameters. Another advantage of ANFIS is its speed ofoperation, which is much faster than in other control strategies; thetedious task of training membership functions is done in ANFIS.Using the CVT is more efficient in areas with turbulent winddistribution.

The research is at an early stage; hence cost and performance ofthe system are unknown. Being the analytical behavior of the sys-tem was limited to the steady state only, the application of theproposed system has to be further investigated, i.e. dynamicalsimulations of the proposed system should be interesting in orderto investigate the response to a gust.

Acknowledgment

The corresponding author would like to acknowledge thefinancial support of the Bright Spark Program at University ofMalaya. This paper is supported by Project Grant TP35005“Research and development of new generation wind turbines ofhigh-energy efficiency” (2011e2014) financed by Ministry of Edu-cation, Science and Technological Development, Republic of Serbia.The last author work is partly funded by the Malaysian Ministry ofHigher Education under the University of Malaya High ImpactResearch Grant UM.C/625/1/HIR/MOHE/FCSIT/17.

References

[1] Marques J, Hey H. A survey on variable-speed wind turbine system. Proc BrazConf Electron Power 2003;1:732e8.

[2] Boukhezzar B, Siguerdidjane H. Review of state of the art in smart rotorcontrol research for wind turbines. Prog Aerosp Sci 2010;46:1e27.

[3] Boukhezzar B, Siguerdidjane H. Nonlinear control with wind estimation of aDFIG variable speed wind turbine for power capture optimization. EnergyConvers Manage 2009;50:885e92.

[4] Laks JH, Pao LY, Wright AD. Control of wind turbines: past, present, and future.In: American control conference, 2009, ACC ’09 2009. pp. 2096e103.

[5] Song YD, Dhinakaran B, Bao XY. Variable speed control of wind turbines usingnonlinear and adaptive algorithms. J Wind Eng Ind Aerodyn 2000;85:293e308.

[6] Aho J, Buckspan A, Laks J, Jeong Y, Dunne F, Pao L. Tutorial of wind turbinecontrol for supporting grid frequency through active power control. In:American control conference Montreal, Canada June 27e29, 2012 2012.

[7] Sedaghati R. A novel control strategy study for DFIG-based wind turbine. In-dian J Sci Technol December 2012;5(12). ISSN: 0974-6846.

[8] Mihet-Popa L, Blaabjerg F. Wind turbine generator modeling and simulationwhere rotational speed is the controlled variable. IEEE Trans Ind Appl January/February 2004;40(1).

[9] Muljadi E, Butterfield CP. Pitch-controlled variable-speed wind turbine gen-eration. In: Presented at the 1999 IEEE Industry Applications Society AnnualMeeting Phoenix, Arizona October 3e7, 1999 1999.

[10] Hall JF, Chen D. Performance of a 100 kW wind turbine with a variable ratiogearbox. Renew Energy 2012;44:261e6.

[11] Ribeiro F, Teixeira JC. Continuously variable transmission in medium-sizedwind generatorsIn ABCM symposium series in mechatronics, vol. 5; 2012.pp. 1329e38. Section VIII e Sensors and Actuators.

[12] Mihailidis A, Karaoglanidis G, Nerantzis I. A CVT system for wind energyconverters. In: The second international conference “power transmissions2006 2006. pp. 411e6.

[13] Gorla C, Cesana P. Efficiency models of wind turbines gearboxes with hy-drostatic CVT. Balkan J Mech Transm 2011;1(2):17e24.

[14] Zadeh LA. Fuzzy sets. Inf Control 1965;8:338e53.[15] Zadeh LA. Fuzzy sets and systems, system theory. Brooklyn, NY: Polytechnic

Press; 1965. pp. 29e39.[16] Mamdani EH. Application of fuzzy algorithms for control of simple dynamic

plant. Proc Inst Electr Eng 1974;121:1585e8.[17] Shamshirband S, Kalantari S, Daliri ZS, Shing Ng L. Expert security system in

wireless sensor networks based on fuzzy discussion multi-agent systems. SciRes Essays 2010;24(5):3840e9.

[18] Shamshirband S. A distributed approach for coordination between trafficlights based on game theory. Int Arab J Inf Technol March 2012;9(2):148e53.

[19] Feizollah A, Shamshirband S, Anuar NB, Salleh R, Mat Kiah ML. Anomalydetection using cooperative fuzzy logic controller, intelligent robotics sys-tems: inspiring the NEXT. Commun Comput Inf Sci 2013;(376):220e31.

[20] Barlas TK, van Kuik GAM. Application of neural network controller formaximum power extraction of a grid-connected wind turbine system. ElectrEng 2005;88:45e53.

[21] Kassem AM. Neural control design for isolated wind generation system basedon SVC and nonlinear autoregressive moving average approach. WSEAS TransSyst February 2012;11(2):39e49.

[22] Li H, Shi KL, McLaren P. Neural network based sensorless maximum windenergy capture with compensated power coefficient. IEEE Trans Ind Appl2005;41(6):1548e56.

[23] Rajaji L, Kumar C. Neural network controller based induction generator forwind turbine applications. Indian J Sci Technol Feb. 2009;2(2). ISSN: 0974-6846.

[24] Ricalde LJ, Cruz BJ, Sánchez EN. High order recurrent neural control for windturbine with a permanent magnet synchronous generator. Comput Sist2010;14(2). ISSN: 1405-5546:133e43.

[25] Sedighizadeh M, Rezazadeh A. Adaptive PID control of wind energy conver-sion systems using rasp1 mother wavelet basis function networks. Proc WorldAcad Sci Eng Technol February 2008;27. ISSN: 1307-6884.

[26] Qiao W, Liang J, Venayagamoorthy GK, Harley R. Computational intelligencefor control of wind turbine generators. In: 2011 IEEE power and energy so-ciety general meeting 2011. http://dx.doi.org/10.1109/PES.2011.6039778.

[27] Shamshirband Shahaboddin, Anuar NB, Kiah MLM, Patel A. An appraisal anddesign of a multi-agent system based cooperative wireless intrusion detection

D. Petkovi�c et al. / Energy 64 (2014) 868e874874

computational intelligence technique. Eng Appl Artif Intell October2013;26(9):2105e27.

[28] Shiraz Muhammad, Gani Abdullah, Hafeez Khokhar Rashid, Buyya Rajkumar.A review on distributed application processing frameworks in smart mobiledevices for mobile cloud computing. IEEE Commun Surv Tutorials November2012;15(3):1294e313. http://dx.doi.org/10.1109/surv.2012.111412.00045.

[29] Ahmed Ejaz, Shiraz Muhammad, Gani Abdullah. Spectrum-aware distributedchannel assignment in cognitive radio wireless mesh networks. Malaysian JComput Sci September 2013;26(3):232e50.

[30] Anuar NB, Sallehudin H, Gani A, Zakari O. Identifying false alarm for networkintrusion detection system using hybrid data mining and decision tree.Malaysian J Comput Sci 2008;21(2):101e15.

[31] Mansoori M, Zakaria O, Gani A. Improving exposure of intrusion deceptionsystem through implementation of hybrid honeypot. Int Arab J Inf Technol2012;9(5):436e44.

[32] Enayatifar Rasul, Sadaei Hossein Javedani, Abdullah Abdul Hanan,Gani Abdullah. Imperialist competitive algorithm combined with refinedhigh-order weighted fuzzy time series for short term load forecasting. EnergyConvers Manage Dec 2013;76:1104e16.

[33] Jang J-SR. ANFIS: adaptive-network-based fuzzy inference systems. IEEE TransSyst Man Cybern 1993;23:665e85.

[34] Petkovi�c D, Pavlovi�c ND. Applications and adaptive neuro-fuzzy estimation ofconductive silicone rubber properties. Strojarstvo 2013;54(3).

[35] Petkovi�c D, Pavlovi�c ND, �Cojba�si�c �Z, Pavlovi�c NT. Adaptive neuro fuzzy esti-mation of underactuated robotic gripper contact forces. Expert Syst Appl2013;40(1). ISSN: 0957-4174:281e6.

[36] Petkovi�c D, Issa M, Pavlovi�c ND, Pavlovi�c NT, Zentner L. Adaptive neuro-fuzzyestimation of conductive silicone rubber mechanical properties. Expert SystAppl 2012;39(10). ISSN: 0957-4174:9477e82.

[37] Petkovi�c D, �Cojba�si�c �Z. Adaptive neuro-fuzzy estimation of automatic nervoussystem parameters effect on heart rate variability. Neural Comput Appl2012;21(8):2065e70.

[38] Petkovi�c D, �Cojba�si�c �Z, Luki�c S. Adaptive neuro fuzzy selection of heart ratevariability parameters affected by autonomic nervous system. Expert Syst ApplSeptember 2013;40(11):4490e5.

[39] Khajeh A, Modarress H, Rezaee B. Application of adaptive neuro-fuzzy infer-ence system for solubility prediction of carbon dioxide in polymers. ExpertSyst Appl 2009;36:5728e32.

[40] Sivakumar R, Balu K. ANFIS based distillation column control. IJCA J 2010:67e73 (Special issue on Evolutionary Computation for Optimization Techniques).

[41] Petkovi�c D, Issa M, Pavlovi�c ND, Zentner L. Intelligent rotational directioncontrol of passive robotic joint with embedded sensors. Expert Syst Appl2013;40(4). ISSN: 0957-4174:1265e73.

[42] Petkovi�c D, Issa M, Pavlovi�c ND, Zentner L, �Cojba�si�c �Z. Adaptive neuro fuzzycontroller for adaptive compliant robotic gripper. Expert Syst Appl2012;39(18). ISSN: 0957-4174:13295e304.

[43] Areed FG, Haikal AY, Mohammed RH. Adaptive neuro-fuzzy control of aninduction motor. Ain Shams Eng J 2010;1:71e8.

[44] Altin N, Sefa I, Mohammed RH. dSPACE based adaptive neuro-fuzzy controllerof grid interactive inverter. Energy Convers Manage 2012;56:130e9.

[45] Tian L, Collins C. Adaptive neuro-fuzzy control of a flexible manipulator.Mechatronics 2005;15:1305e20.

[46] Aldair AA, Wang WJ. Design an intelligent controller for full vehicle nonlinearactive suspension systems. Int J Smart Sens Intell Syst 2011;4(2):224e43.

[47] Dastranj MR, Ebroahimi E, Changizi N, Sameni E. Control DC motorspeed withadaptive neuro-fuzzy control (ANFIS). Austr J Basic Appl Sci 2011;5(10):1499e504.

[48] Wahida Banu RSD, Shakila Banu A, Manoj D. Identification and control ofnonlinear systems using soft computing techniques. Int J Model Optim2011;1(1):24e8.

[49] Grigorie TL, Botez RM. Adaptive neuro-fuzzy inference system-based con-trollers for smart material actuator modelling. J Aerosp Eng 2009:655e68.

[50] Akcayol MA. Application of adaptive neuro-fuzzy controller for SRM. Adv EngSoftw 2004;35:129e37.

[51] Moustakidis SP, Rovithakis GA, Theocharis JB. An adaptive neuro-fuzzytracking control for multi-input nonlinear dynamic systems. Automatica2008;44:1418e25.

[52] Peymanfar A, Khoei A, Hadidi K. Design of a general proposed neuro-fuzzycontroller by using modified adaptive-network-based fuzzy inference sys-tem. Int. J Electron Commun 2010;64:433e42.

[53] Omar BAA, Haikal AYM, Areed FFG. Design adaptive neuro-fuzzyspeed controller for an electro-mechanical system. Ain Shams Eng J 2011;2:99e107.

[54] Heier S. Wind energy conversion systems. In: Grid integration of windenergy conversion systems. Chichester, UK: John Wiley & Sons Ltd.; 1998.pp. 34e6.