ra

Tr

Cana

Received 10 September 2014Accepted 10 April 2015

Multivariable controlWind turbineDC generator

wiiable control of a wind turbine is designed for investigation of theoretical concepts and their physicalimplementation. The control system includes a speed controller and a disturbance estimator for

wind energy and energy extraction, a maximum power point tracking algorithm is developed and in-

wer ge

increase in the related education and research [1e4]. Because it is

whose robustness need to be veried under realistic operatingconditions [7].

Laboratory experiments in the wind energy area includinghardware-in-the loop (HIL) and advanced control systems areimportant for education of future engineers and researchers.

turbine system and extracting mathematical models needed for thejor requirement in

(WECS) is con-ltage in order town, the optimum3]. Variable speedh wind speeds, thewith power elec-d prevent the tur-ic control using

variable pitch blades is usually expensive and complex. It ca alsocause an unnecessarily high activity of the pitch actuator due tosmall uctuations of power during the steady state operation [7].

Control systems implemented in the power electronic interfacerepresent an efcient means to operate a wind turbine at themaximum power extraction. The control is not always aimed atcapturing as much energy as possible. Power generation is limitedduring high wind speeds or when the load demand in an isolatedsystem is low. Model based control strategies, such as feedback,

* Corresponding author. Tel.: 1 9024205712; fax: 1 9024205021.

Contents lists availab

Renewable

els

Renewable Energy 83 (2015) 162e170E-mail address: adel.merabet@smu.ca (A. Merabet).difcult to use a real wind turbine in laboratory environment, asmall turbine that can be used indoors is the best tool for imple-mentation and demonstration of control strategies [5,6]. Windturbine emulators, involving motor-generator set, variable load,and control system, which operate with the power-speed charac-teristics of a wind turbine are frequently used for research andteaching purposes due to their simplicity, low power, and low costdesign [5]. However, neglecting the real wind effects represents animportant aw, especially when dealing with control strategies

design of a model-based control system. One mathe considered wind energy conversion systemtrolling the generator speed and the load vomaximize wind energy extraction. As well knoturbine speed is a function of wind speed [1eWECS are increasingly common. Typically, at higWECS use aerodynamic control in combinationtronics to regulate torque, speed, and power, anbine from damage. However, the aerodynamenergy enjoys signicant interest, which has caused considerable advanced control by analysing the major components of a windSpeed controlMaximum power extraction

1. Introduction

In modern power systems, the pohttp://dx.doi.org/10.1016/j.renene.2015.04.0310960-1481/ 2015 Elsevier Ltd. All rights reserved.turbine laboratory system. A power electronic interface is based on two DCeDC converters: a buckconverter for control of the speed and a boost converter controlling the load voltage. Experimental re-sults demonstrate effectiveness of the multivariable control system for a wind turbine providingmaximum power extraction. The experiment can be recongured for teaching various control conceptsto both undergraduate and graduate students.

2015 Elsevier Ltd. All rights reserved.

neration based onwind

Developing a control system from the model of the wind turbine,and its practical realization would bridge the gap between theoryand practice. It would allow the students to implement theories ofKeywords: tegrated into the control system. The multivariable control system is implemented in a small windAvailable online enhanced robustness of the control system. In order to provide students with deeper understanding ofMultivariable control algorithm for laboenergy conversion

A. Merabet a, *, Md.A. Islam a, R. Beguenane b, A.M.a Division of Engineering, Saint Mary's University, Halifax, NS, B3H 3C3, Canadab Department of Electrical Engineering, Royal Military College, Kingston, ON, K7K 7B4,c Department of Electrical Engineering, University of Nevada, Reno, USA

a r t i c l e i n f o

Article history:

a b s t r a c t

Advanced experimentation

journal homepage: www.tory experiments in wind

zynadlowski c

da

th wind energy conversion systems is described. The real time multivar-

le at ScienceDirect

Energy

evier .com/locate/renene

generator is connected to the load via a power electronic interface

From the electrical and mechanical Equation (4) of the DC ma-chine, a linear state-space equation can be derived as

_x Fx g1V g2Tt (5)

where,

x i u T ; F

26664RL

KbL

KiJ

BJ

37775; g1

264

1L

0

375; g2

2640

1J

375

The controlled output is the rotational speed u, the input is thevoltage V and the disturbance is the turbine torque Tt.

In order to nd a relationship between the output u and theinput V, the mechanical Equation (4b) is differentiated, using

ble Eallowing control of the shaft speed and load voltage.The power delivered by the turbine shaft (neglecting losses in

the drive train) is given by

Pt 0:5prCplr2v3w (1)

where r denotes the air density, r is the length of the turbine blade,vw is the wind speed, and l is the ratio of blade tip speed to windspeed, that is

l urvw

(2)

where u is the angular velocity of the turbine.The power coefcient Cp depends on speeds of the turbine and

wind, and its relation to l is shown in Fig. 1. The power coefcientreaches maximum at a specic optimum value lopt. In order toextract maximum power fromwind, the turbine speed should be socontrolled as to maintain l at the optimum level.

In some wind turbines, the optimum tip speed ratio may beunknown or not well dened and subject to change. Therefore,instant locating of the maximum Cp during the operation of windturbine is very important. An MPPT algorithm based on the varia-tion of the generated power and the shaft speed is proposed, inSection 3.

The torque at the turbine shaft produced by the wind is given bypredictive, and sliding mode, can be employed for speed control inWECS. Quality of control strategies depends on the accuracy ofmathematical model of the system, which is usually not high [8,9].For accurate speed tracking, the controller must maintain highperformance when facing parameter variations and uncertaintiesof the system [10e15].

In this paper, a feedback speed control strategy is developedfrom the mathematical model of a generator connected to a windturbine. Information about the turbine and wind speeds is assumedto be unavailable and their variations will be compensated using atorque estimator integrated in the controller. Performance of theproposed controller will be tested under multivariable controlconditionswith amaximumpower point tracking (MPPT) algorithmand load voltage control. A Quanser's ve-blade wind turbine isemployed in an experimental setup equipped with a power elec-tronic interface. The setup allows to verify efcacy of the proposedcontrol system and to investigate its behaviour with real wind [16].

The rest of the paper is organized as follows: In Section 2, adescription of the experimental system of wind turbine system isgiven. The proposed feedback control method for speed tracking isdetailed in Section 3 followed by the robustness and stabilityanalysis in Section 4. The MPPT algorithm, generating the speedreference needed for maximum extraction of power from wind, isdescribed in Section 5. The experimental setup is described in de-tails in Section 6 and experimental results and their discussion aregiven in Section 7.

2. Wind turbine experimental system

2.1. Wind turbine

Thewind turbine, manufactured by Quanser Inc., is installed in awind tunnel. It has ve blades and drives a DC generator through agearbox of ratio 1:1. The gearbox converts rotation of thehorizontal-axis turbine to that of the vertical-axis generator. The

A. Merabet et al. / RenewaTt 0:5prCtr3v2w (3)where Ct Cp/l is the torque coefcient. Here, the mathematicalmodel of the mechanical structure of wind turbine system isassumed to be unknown. This uncertainty is dealt with in theproposed control system.

2.2. DC generator

The armature of the DC generator is modeled as an RLE circuit,with E representing the back emf (speed voltage). Denoting thegenerated voltage as V, the electrical and mechanical equations ofthe generator can be written as

didt

RLi Kb

Lu 1

LV (4a)

dudt

KiJi B

Ju 1

JTt (4b)

where i is the armature current, Kb is the machine constant, u is therotational speed of the generator, V is the generator voltage, J is therotor inertia, B is the viscous-friction coefcient, and Tt is the un-known turbine torque.

3. Feedback control for speed tracking

3.1. Feedback controller development

Fig. 1. Power coefcient of the wind turbine versus tip-speed ratio.

nergy 83 (2015) 162e170 163Equation (4a), which yields

In this work, the development of a robust controller to deal with

ble EPg2 p0 B k2

1

(14)parameter variations and unknown turbine torque is based on atorque observer. It is derived from the state space model of the DCgenerator, the control law, and the speed tracking error of the windenergy conversion system.

From the model (5), the torque variable is dened as

g2Tt _x Fx g1V (10)Using the torque Equation (10) an observer can be dened as

_bT t Pg2bT t P _x Fx g1V (11)where P, of size 1 2, is the observer gain to be determined.

From (7), (10) and (11), it can be found that the error dynamic ofthe observer is given by

_e _Tt _bT t Pg2Tt bT t (12a)_e Pg2e 0 (12b)

It is exponentially stable if the term Pg2 is a positive constant.Based on (4), (5) and (6), the observer gain P can be dened as

P p0Ki

JBJ

k20 1

(13)

where p0 is a positive constant.From (5) and (13), the term Pg2 is given by

2u KiJ

_i BJ

_u 1J

_Tt

KiJ

R

Li Kb

Lu

B

J

Ki

Ji B

Ju

Ki

JLV B

J2Tt (6)

The dynamic of the wind turbine is more sluggish than theelectric energy conversion system. Therefore, according to [10], itcan be assumed that

_Tt 0 (7)The objective of the control law is minimization of the speed

tracking error expressed by the second-order equation

eu k2 _eu k1eu 0 (8)

where, eu u urefThe following control law is proposed

V JLKi

k1u uref

k2

Ki

Ji B

Ju _uref

Ki

J

R

Li Kb

Lu

B

J

Ki

Ji B

Ju

uref

k21JTt BJ2 Tt

(9)

Implementation of this law requires precise knowledge of pa-rameters of the system and the turbine torque, which is not easilyavailable in the majority of practical wind energy systems.

3.2. Robust control based on torque estimation

A. Merabet et al. / Renewa164J Jwhich is positive if the control gain k2 is such thath dDai Dbu DcV Dd Tt xk2 >BJ

(15)

Thus, the stability of the torque observer is guaranteed.Using the observed torque bT t , the control law (9) becomes

V JLKi

k1u uref

k2

Ki

Ji B

Ju _uref

Ki

J

R

Li Kb

Lu

B

J

Ki

Ji B

Ju

uref

k21JbT t BJ2 bT t

(16)

If Equations (13), (14) and (16) are substituted in (11), a newform of the torque observer is produced

_bT t p0k1u uref k2 _u _uref u uref (17)Integrating (17), the torque observer becomes

bT t p0k1

Z u uref

dt k2

u uref

_u _uref

(18)

The torque observer has a PID structure, which helps the feed-back controller (9) to enhance its capability of speed tracking andcompensation of uncertainties resulting from the lack of detailedknowledge of wind turbine parameters.

4. Robustness and stability analysis

4.1. Robustness analysis

The dynamics of the wind turbine is an uncertain system, wherethe uncertainties include:

a. Structured (parametric) uncertainties characterized by adynamical model of the system with parameter variations.

b. Unstructured uncertainties characterized by omitted quantities,unknown torque, and external disturbances.

The rotor dynamic Equation (6) can be rearranged to

u ai bu cV dTt x (19)

where a KiJRL BJ

; b 1J

KiKbL B

2

J

; c KiJL; d BJ2; and x

represents the effects of the omitted quantities of the electric/po-wer electronics circuit, load, and any external disturbance.

Now, the Equation (19) can be modied to include parameteruncertainties D($) and external disturbance x to

u a Dai b Dbu c DcV d DdTt x (20)Equation (20) is now reorganized to include all uncertainties in a

common term as

u ai bu cV dTt h (21)

where

1

nergy 83 (2015) 162e170The uncertainties term, h, includes parametric uncertainty,omitted quantities, and external disturbances in the system.

2 3 2 3

Implementation of (34) has been carried out numerically using a

ble Ex x1x2x3

4 5 eu_eue

4 5 (27)From (26) and (27), the state space representations are

8>>>>>>>>>>>>>>:

l1 p0k3

l2 k2

k22 4k1

q2

l3 k2

k22 4k1

q2

(32)

It can be seen that eigenvalues li satisfy the condition Re (li) < 0.Therefore, x(t) / 0 as t / for all x(0), and the stability isguaranteed.

5. Maximum power point tracking algorithm

For maximum power extraction, if the wind turbine character-istics are available, the speed reference prole is derived from (2)using the optimum tip speed ratio lopt such that

uref loptvw

r(33)

However, practical implementation of the MPPT algorithmbased on (33) requires measurement of the wind speed andknowledge of the wind turbine characteristics. It makes this algo-rithm unreliable due to inaccuracies of the wind speed determi-nation and aws in modelling the wind turbine. Here, an MPPTmethod based on the variation of the generated power P and theturbine-generator speed u is proposed. The variation of the speedreference uref is given by

durefdt

a,u,dPdt

(34)

where the generated power is found from measurements of thevoltage and current at the generator, that is,

P Vi (35)

and a is a constant. Correct choice of awill improve speed tracking

nergy 83 (2015) 162e170 165discretisation method as fellows

uref kDt uref k 1Dt a$ukDt$PkDt Pk 1Dt(36)

where, Dt is the sampling time and k is an integer.

6. Wind turbine experimental setup

The wind turbine experimental setup consists of a ve-bladewind turbine and a DC generator, as shown in Fig. 2. The gener-ator provides power to a set of LEDs, which form a variable load.Control of the turbine speed and of the load voltage is realizedthrough a power electronic interface between the generator andload. The interface consists of two DCeDC converters as shown inFig. 3. The DCeDC buck converter controls the rotational speed ofthe generator and the DCeDC boost converter controls the loadvoltage. Such a cascade of two DCeDC converters would beimpractical and in a commercial system. However, it allows inde-pendent manipulations, valuable for teaching purposes. A detailed

Voltage and current sensors (VS and CS) are available in theexperimental setup, as shown in Fig. 3, to provide measurements ofvoltages of the generator, the buck converter, and the load, as wellas the generator and load currents. The measurements are deliv-ered to the computer (PC) through a real-time data acquisitionboard Q8-USB. The results were analysed using software packagesQuarc with Matlab/Simulink. Quarc is a rapid control prototypingtool, which signicantly accelerates the control system design andimplementation [16].

Test 1: A randomwind speed, shown in Fig. 5, was generated bychanging the blower speed in order to test the MPPT algorithm and

P Robust Feedback Speed Control

Blower

+V*_

DC Generator

DC-DC Buck

r Wind Turbine

Tunnel

_ref

i

vw

GG

MPPT

G2 G6G1

A. Merabet et al. / Renewable Energy 83 (2015) 162e170166description of the interface exceeds the scope of this paper andmore information can be found in Ref. [17].

The variable electronic load, based on LEDs and shown in Fig. 4,consists of six parallel equal banks of two LEDs in each bank inseries with a resistance. The load banks can be switched auto-matically ON/OFF from Matlab/Simulink, through the real-timedata acquisition board Q2-USB [16], by sending a signal to theMOSFET gates of the banks.

The load side converter control system is used to regulate thevoltage across the load in order to maintain a proper functioning ofthe LEDs load. The PI voltage controller produces signal U* given by

U* kpVL V*L

ki

Z VL V*L

dt (37)

The output of the voltage controller is the ring signal to bedelivered to the gate of the MOSFET in the DCeDC buck converteras shown in Fig. 3. Rotational speed is measured by an encodermounted on DC generator rotor shaft. Voltage and current sensorsare available to measure the armature current and the generatorand load voltages. The measurements are calibrated and sent tothe computer (PC) through a real-time data acquisition board Q8-USB, to be analysed by using the software package QUARC withMATLAB/Simulink. QUARC is a powerful rapid control prototypingtool that signicantly accelerates control system design andimplementation [16].Fig. 2. Experimental set-up of the wind energy conversion system.The wind is generated by a DC blower motor. An incrementalencoder mounted on blower rotor shaft measures the motor speed,which is proportional to the wind speed. The advantage of theproposed control system is that it does not require knowledge ofthe wind speed for evaluation of the turbine torque.

7. Experimental results

Experiments were carried out to validate the proposed controlstrategy under different scenarios of operation. The choice of thefeedback controller gains (k1 and k2), the estimator gain (p0) andthe load controller (kp and ki) were determined by trial and error toachieve high-quality performance.

VL* DC-DC Boost

Converter

Voltage Control

U*L*_

VLV+

Fig. 3. Multivariable control strategy for the experimental wind turbine-generatorsystem.Fig. 4. Variable electronic load.

ed t

A. Merabet et al. / Renewable Energy 83 (2015) 162e170 167the rotor speed tracking. The experimental wind turbine systemwas operating with maximum power extraction, nominal values ofthe parameters of themodel (4), and xed load (all banks were ON).

Fig. 5. MPPT based speIt can be seen that the speed reference generated from MPPT andthe rotor speed follow the variation of the wind speed. Estimationsof the extracted power and torque are illustrated in Fig. 6. The es-timators enhance the control, which results in a zero steady-state

Fig. 6. Estimated torque, extracted pospeed error. The voltage regulation does not affect the speedtracking response.

Test 2: The wind turbine and control system were tested with a

racking for xed load.variable load, the banks being turned ON at different time intervalsas shown in Fig. 7. It can be observed in Fig. 8 that speed trackingand voltage control are accurate in spite of the varying wind andload. This high performance is attained by precise torque

wer and load voltage regulation.

Fig. 7. Variable load banks.

A. Merabet et al. / Renewable Energy 83 (2015) 162e170168estimation, which allows compensation of external disturbancesarising from the wind and load variations. The generated powerand current follow the wind speed, as shown in Fig. 9, proving thatthe turbine extracts themaximumpower fromwind. The calibratedcontrol input signal shown in Fig. 10 makes the turbine to operatewithin a safe region.

Test 3: The experimental wind turbine system was tested withmismatched parameters. The perceived values of generator pa-

rameters had been increased by 10% and the load was varying

Fig. 8. MPPT based speed tragain. The results of speed tracking and torque estimation areshown in Fig. 11. The performance is still satisfactory, demon-strating robustness of the developed controller.

8. Conclusions

A robust feedback control strategy has been proposed to track aspeed prole, generated by an MPPT algorithm, to operate a labo-

ratory wind energy system based on a DC generator. Information

acking for variable load.

Fig. 9. Generated power and current, and load voltage regulation.

A. Merabet et al. / Renewable Energy 83 (2015) 162e170 169about the wind turbine and wind speed is not needed in theimplementation of this strategy, as their effects are compensatedthrough a torque estimator integrated into the controller. Param-eter, wind, and load variations, and component omitted in themathematical model, such as the power electronics interface, do

not signicantly spoil the speed tracking performance.

Fig. 10. Calibrated conStability and robustness of the feedback controller have beenanalysed, and the developed control algorithmwas tested in a smallscale laboratory wind turbine system. The systemwill be used as animportant tool for teaching control systems theory and training inthe eld of wind energy.trol input signal.

Fig. 11. MPPT based speed tracking for variable l

A. Merabet et al. / Renewable Energy 83 (2015) 162e170170Acknowledgements

This study was partially supported by the Natural Sciences andEngineering Research Council of Canada (NSERC) under the Engage

Grant 432302-12.

Appendix

Wind turbine: r 14 cm, r 1.14 kg/m3DC generator: R 3.705U, L 575 mH, Kb 10.575 mV/rpm,

Ki 100.95 mNm/A, B 0.001833, J 165 g/cm2Feedback controller and torque estimator: k1 3.33 105,

k2 250, p0 0.001.MPPT algorithm: a 0.5.PI voltage controller: Kp 20, Ki 0.5.

References

[1] Rubio JO, Aguilar LT. Maximizing the performance of variable speed windturbine with nonlinear output feedback control. Procedia Eng 2012;35:31e40.

[2] Merabet A, Thongam T, Gu J. Torque and pitch angle control for variable speedwind turbines in all operating regimes. In: Proc 10th int conf. on environ-mental and electrical engineering; 2011.

[3] Neammanee B, Sirisumrannukul S, Chatratana S. Development of a windturbine simulator for wind generator testing. Int Energy J 2007;8:21e8.

[4] Mayo-Maldonado JC, Salas-Cabrera R, Cisneros-Villegas H, Castillo-Ibarra R,Roman-Flores J, Hernandez-Colin MA. Maximum power point tracking controlfor a DC-generator/multiplier-converter combination for wind energy appli-cations. In: Proc world congress on engineering and computer Sc; 2011. p. 1.[5] Kariyawasam K, Karunarathna K, Karunarathne R, Kularathne M, Hemapala R.Design and development of a wind turbine simulator using a separatelyexcited DC motor. Smart Grid Renew Energy 2013;4:259e65.

[6] Merabet A, Keeble R, Rajasekaran V, Beguenane R, Ibrahim H, Thongam J.Power management system for load banks supplied by pitch controlled wind

oad and mismatched generator parameters.turbine system. Appl Sci 2012;2:801e15.[7] Neammanee B, Sirisumrannukul S, Chatratana S. Control strategies for

variable-speed xed-pitch wind turbines. In: Muyeen SM, editor. wind power.Intech; 2010. p. 209e32.

[8] Magri AE, Giri F, Besanon G, Fadili AE, Dugard L, Chaoui FZ. Sensorlessadaptive output feedback control of wind energy systems with PMS genera-tors. Control Eng Pract 2013;21:530e43.

[9] Beltran B, Ahmed-Ali T, Benbouzid MEH. Sliding mode power control ofvariable-speed wind energy conversion systems. IEEE Trans Energy Convers2008;23:551e8.

[10] Li Z, Chen J, Zhang G, Gan MG. Adaptive robust control for DC motors withinput saturation. IET Control Theory & Appl 2011;5:1895e905.

[11] Qiang H. The design of human simulated intelligent controller for DC motor.In: Proc int conf on automatic control and articial intelligence; 2012.p. 2088e91.

[12] Son YI, Choi DS, Lim S, Kim KI. Robust current control for speed sensorless DCmotor drive using reduced-order extended observer. Electron Lett 2012;48:1112e4.

[13] Hilairet M, Auger F. Speed sensorless control of a DC-motor via adaptive l-ters. IET Electr Power Appl 2007;1:601e10.

[14] Sankar R, Ramareddy S. Closed loop controlled PMDC motor drive. In: Proc IETint conf on information and communication technology in electrical sciences;2007. p. 329e32.

[15] Duran MJ, Barrero F, Pozo-Ruz A, Guzman F, Fernandez J, Guzman H. Un-derstanding power electronics and electrical machines in multidisciplinarywind energy conversion system courses. IEEE Trans Educ 2013;56:174e82.

[16] Wind turbine experiment. Quick-start guide. Quanser Inc; 2010.[17] Merabet A, Kerr J, Rajasekaran V, Wight D. Power electronics circuit for speed

control of experimental wind turbine. In: Proc 24 int conf on microelectronics;2012.

Multivariable control algorithm for laboratory experiments in wind energy conversion1. Introduction2. Wind turbine experimental system2.1. Wind turbine2.2. DC generator

3. Feedback control for speed tracking3.1. Feedback controller development3.2. Robust control based on torque estimation

4. Robustness and stability analysis4.1. Robustness analysis4.2. Closed loop stability analysis

5. Maximum power point tracking algorithm6. Wind turbine experimental setup7. Experimental results8. ConclusionsAcknowledgementsAppendixReferences