IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 26, NO.

1, JANUARY 2011

165

Application of Adaptive Network-Based Fuzzy Inference System for Sensorless Control of PMSG-Based Wind Turbine With Nonlinear-Load-Compensation Capabilities
Mukhtiar Singh, Student Member, IEEE, and Ambrish Chandra, Senior Member, IEEE
AbstractThe precise information of permanent-magnet synchronous generator (PMSG) rotor position and speed is essentially required to operate it on maximum power points. This paper presents an adaptive network-based fuzzy inference system (ANFIS) for speed and position estimation of PMSG, where an ANFIS-based model reference adaptive system is continuously tuned with actual PMSG to neutralize the effect of parameter variations such as stator resistance, inductance, and torque constant. This ANFIS-tuned estimator is able to estimate the rotor position and speed accurately over a wide speed range with a great immunity against parameter variation. The proposed system consists of two back-to-back connected inverters, where one controls the PMSG, while another is used for grid synchronization. Moreover, in the proposed study, the grid-side inverter is also utilized as harmonic, reactive power, and unbalanced load compensator for a three-phase, four-wire (3P4W) nonlinear load, if any, at point of common coupling (PCC). This enables the grid to always supply/ absorb a balanced set of fundamental currents at unity power factor. The proposed system is developed and simulated using MATLAB/SimPowerSystem (SPS) toolbox. Besides this, a scaled laboratory hardware prototype is developed and extensive experimental study is carried out to validate the proposed control approach. Index TermsActive-power lter, adaptive neuro-fuzzy systems, distributed generation, grid interconnection, permanent-magnet synchronous generator (PMSG), power quality, renewable energy, sensorless control, wind energy.

I. INTRODUCTION

HE GROWING concern about global warming and the harmful effects of fossil-fuel emissions has created new demand for cleaner and sustainable renewable energy sources. Among renewable sources, wind energy is one of the fastest

Manuscript received October 16, 2009; revised March 25, 2010; accepted June 11, 2010. Date of current version December 27, 2010. This work was supported by the Government of India and the High Commission of India, Ottawa, ON, Canada. Recommended for publication by Associate Editor M. G. Simoes. M. Singh is with the Department of Electrical Engineering, Ecole de Technologie Superieure, Montreal, QC H3C 1K3, Canada, on leave from the Department of Electrical Engineering, Deenbandhu Chhoturam University of Science and Technology, Sonepat 131 039, India (e-mail: smukhtiar_79@yahoo.co.in). A. Chandra is with the Department of Electrical Engineering, Ecole de Technologie Superieure, Montreal, QC H3C 1K3, Canada (e-mail: ambrish. chandra@etsmtl.ca). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TPEL.2010.2054113

growing and lowest-priced renewable energy technologies available today, costing between 4 and 6 cents/kWh, depending upon the size of a particular project. Modern wind turbines have lot of commercially available topologies based on induction generator (xed speed) and doubly fed induction generator/synchronous generator/permanent-magnet synchronous generator (PMSG) (variable speed) [1][3]. The variable-speed wind turbines are more attractive, as they can extract maximum power at different wind velocities, and thus, reduce the mechanical stress on WECS by absorbing the wind-power uctuations. Recently, PMSG-based directly driven variable-speed WECS are becoming more popular due to the elimination of gear box and excitation system. Since, the maximum power is the cubic function of generator speed for a given tip speed ratio, the continuous information of generator position and speed is essentially required. For this purpose, generally shaft-mounted speed sensors are used, resulting in additional cost and complexity of the system. To alleviate the need of these sensors, several speed-estimating algorithms based on motional electromotive force (EMF), ux-linkage variation, and Kalman lter have been introduced in the past. However, the precise estimation of rotor position and speed is very difcult as most of these suffer because of simplied computations based on several assumptions, ignorance of parameter variations, and inaccuracy involved with low-voltage signal measurement at lowerspeed, especially in case of directly driven PMSG [4][8]. To overcome these problems, several indirect methods for maximum power extraction have also been introduced. Bhowmik and Spee [9] have proposed power-coefcient polynomial to estimate wind velocity. In this method, an iterative algorithm is employed to determine the roots of a polynomial. As the polynomial may be up to seventh order, therefore, the exact calculation of the roots is very complex and time-consuming task. Tan and Islam [10] reported to apply a 2-D lookup table of power-coefcient or power-mapping method to estimate the wind velocity. The mapping-power technique may occupy a lot of memory space. While saving the memory space by reducing the size of table, the control accuracy will be affected. Li et al. [11] have also suggested a similar articial neural network (ANN) based power-coefcient compensation method. In this paper, a novel adaptive network-based fuzzy inference system (ANFIS) based position and speed estimator of PMSG has been proposed for wide range of speed operation. The ANFIS architecture has well-known advantages of

0885-8993/$26.00 2010 IEEE

166

IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 26, NO. 1, JANUARY 2011

modeling a highly nonlinear system, as it combines the capability of fuzzy reasoning in handling uncertainties and capability of ANN in learning from processes [12]. Thus, the ANFIS is used to develop an adaptive model of variable-speed PMSG under highly uncertain operating conditions, which also automatically compensates the distortion in measuring signal and any variation in parameters such as inductance, resistance, etc. A detailed comparison of simulation results supported by experimental results obtained using proposed ANFIS estimator is also provided with EMF-based method under variable-speed PMSG operation. Both the simulation and experimental studies are also carried out for 20% variation in PMSG resistance and inductance to validate the robustness of proposed ANFIS algorithm against generator parameter variation. An error-gradientbased dynamic back-propagation method has been used for the online tuning of ANFIS architecture. This estimated rotor speed is further utilized to nd out the maximum possible power using powerspeed characteristic of PMSG. An IGBT-based rectier is utilized to implement the proposed control strategy. Moreover, in the proposed study, the grid-side inverter rating is optimally utilized by incorporating the power-quality improvement features. Normally, the grid-interfacing inverter has very low utilization factor of 20%30% with a possible peak of 60% of rated output due to the intermittent nature of wind [13], [14]. Therefore, if the same inverter is utilized for solving power-quality problem at point of common coupling (PCC) in addition to its normal task, then the additional hardware cost for custom power devices like active power lter (APF), static compensator (STATCOM), or VAR compensator can be saved. Thus, the authors have proposed a very simple and cost-effective solution by using the grid-side inverter as a load harmonics, load reactive power, and load unbalance compensator of a 3P4W nonlinear unbalanced load at PCC in a distribution network, in addition to its normal task of wind-power injection into the grid. The paper is organized as follows. The system under consideration is discussed in Section II. The proposed control strategy for generator and grid-side converters is explained in Section III. The MATLAB/SPS-based simulation results are given in Section IV and the experimental results are discussed in Section V. Finally, Section VI concludes the paper. II. SYSTEM DESCRIPTION AND CONTROL The proposed system consists of a PMSG-based variablespeed WECS consisting two back-to-back inverters with a common dc link. The generator-side inverter controls its speed to extract maximum power at different speeds, while the gridside inverter delivers the renewable power to grid with 3P4W nonlinear-load compensation simultaneously. The block diagram of proposed variable-speed WECS is shown in Fig. 1. A. Generator Position/Speed Estimation and Control The information about generator position and speed is very essential for optimal control of PMSG, in order to extract maximum power at different wind velocities. However, due to nonideal rotor-ux distribution, parameter variation, and

Fig. 1.

Block diagram of proposed system.

Fig. 2.

Control diagram of proposed system.

rapidly changing operating conditions, it is extremely difcult to reproduce the exact model of PMSG for control purpose. Therefore, in order to reject any such kind of external disturbances, an adaptive model is continuously tuned in parallel of PMSG with the help ANFIS structure. The complete schematic diagram of proposed system is shown in Fig. 2. 1) Modeling of PMSG: Since the back-EMF is the function of rotor position in stationary reference frame, therefore, it is convenient to model PMSG in this frame. The voltage equations of PMSG are as follows: V = Rs i Ls di + E dt di + E V = Rs i Ls dt (1) (2)

where V are stator terminal voltages, Rs is stator resistance, Ls is stator inductance, i are output currents, and E are back EMFs, which can be given as E = E E = r m sin(r ) cos(r ) . (3)

Here, r , r , and m are rotor speed, rotor position, and magnetic-ux linkage, respectively.

SINGH AND CHANDRA: APPLICATION OF ADAPTIVE NETWORK-BASED FUZZY INFERENCE SYSTEM

167

Fig. 3.

State-space model of PMSG.

On rearranging (1) and (2), and rewriting them in matrix form, we have i i = Rs /Ls 0 + 1/Ls 0 0 Rs /Ls 0 1/Ls i i V V E E . (4)
Fig. 4. Block diagram of ANFIS-based speed and position estimation.

Using the same structure for adaptive model, the current estimator can be designed, if the accurate value of back EMF is known in the following equation: = A + B (V E ) i i where the cap indicates the estimated variables and A= Rs /Ls 0 0 Rs /Ls = i i i , B= 1/Ls 0 V V 0 1/Ls , and , (5)

= i , i i E = E E .

V =

Fig. 5.

Schematic of ANFIS architecture.

The state-space-equivalent diagram of (5) is shown in Fig. 3. 2) ANFIS-Based Rotor Position and Speed Estimation: Since the system is highly nonlinear and working under different operating conditions, it is extremely difcult to estimate the exact value of back EMF, which also carries the information about the rotor position and speed. To solve this problem, an ANFIS-based dynamic model of PMSG is developed. For ANFIS training, the explicit mathematical model of plant is not required. Only the order of the system with adequate input/output data from actual plant is needed to develop its adaptive model [15]. The target output from the actual plant is compared with the output of ANFIS-based dynamic model in parallel and error is used to update the weights. In the proposed study, the dynamic model is continuously tuned by forcing its estimated currents to track the actual currents of PMSG using ANFIS structure with dynamic back-propagation method. Once the estimated currents are equal to actual currents, the adaptive model exactly represents the real model irrespective of any kind of parameter variations and external disturbances. This results in the accurate estimation of back EMFs given as follows: E = E E = r m sin(r ) cos(r ) . (6)

Now, the position and speed can be easily estimated using the estimated value of back EMFs as r = tan1 (E /E ) and r /dt), where p is the number of pole pairs. The r = (1/p)(d estimated back EMFs are passed through low-pass lter to lter the noise. The schematic diagram of rotor position and speed estimation is shown in Fig. 4. 3) ANFIS Architecture: An ANFIS based on Takagi SugenoKang (TSK) method having 2:6:3:3:2 architecture with two inputs (estimated currents and ) and two outputs (estii i mated back EMFs E and E ) are used to develop the dynamic model of PMSG. The errors between the actual and estimated i currents ( = i ) are used to tune the precondition and consequent parameters. The diagram of ANFIS architecture is shown in Fig. 5 and the node functions of each layer are described as follows. Layer 1: This layer is also known as fuzzication layer, where each node is represented by a square. Here, three membership functions are assigned to each input. The trapezoidal and triangular membership functions are used to reduce the computation burden, as shown in Fig. 6, and their corresponding node

168

IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 26, NO. 1, JANUARY 2011

and
B B B Y2 = 1 f1 + 2 f2 + 3 f3 .

Fig. 6.

Fuzzy membership functions.

equations are as given as follows: X 1, X a 1 A 1 (X ) = B 1 (X ) = , b1 b1 a1 0, X 1 X a2 , 0.5b2 A 2 (X ) = B 2 (X ) = 0,

b1 X a1 |X a2 | 0.5b2 |X a2 | 0.5b2 (8) a1 (7)

The output from this layer are now multiplied with the normalizing factor and passed through the low-pass lter to nd the estimated value of back EMFs E and E . 4) Online Training of ANFIS Architecture: For the proposed ANFIS architecture, a gradient-descent technique is used to reduce the error (usually a cost function given by the squared error), where the weights are iterated by propagating the error from output layer to input layer. This backward trip of such a calculation is termed as back propagation. The training algorithm is completed in two stages, known as preconditionparameter tuning and consequent-parameter tuning, where the objective function to be minimized is dened as
2 i = (i )2 .

(14)

Precondition-parameter tuning: The precondition parameters are required to update the fuzzy membership functions as discussed in previous section for Layer 1. To minimize the 2 error function by gradient-descent method, the change in each precondition parameter must be proportional to the rate of change of the error function w.r.t. that particular precondition parameter, i.e., aA i =
2 , aA i

X a3 0, X a 3 A 3 (X ) = B 3 (X ) = , a3 X b3 b3 a3 1, X b3

i = 1, 2, 3

(15)

(9)

where is the constant of proportionality dened as the learning rate. Therefore, the new value of the consequent parameter is given as aA i (n + 1) = aA i (n) + aA i , or aA i (n + 1) = aA i (n) Similarly, bA i (n + 1) = bA i (n)
2 , bA i 2 , aA i

where the value of parameters {ai , bi } changes according to the error, and accordingly generates the linguistic value of each membership function. Parameters in this layer are referred as premise parameters or precondition parameters. Layer 2: Every node in this layer is a circle labeled as , which multiplies the incoming signals and forwards it to next layer i = A i (X ) B i (X ), i = 1, 2, 3. (10)

i = 1, 2, 3

(16)

i = 1, 2, 3.

(17)

i = 1, 2, 3.

(18)

Here, the output of each node represents the ring strength of a rule. Layer 3: Every node in this layer is represented as a circle. This layer calculates the normalized ring strength of each rule given as follows: i i = , i = 1, 2, 3. (11) 1 + 2 + 3 Layer 4: Every node in this layer is a square node with a node function given by i = 1, 2, 3 and j = , . (12) where the parameters {ai , ai } are tuned as the function of inputs 0 1 { , }. Parameters in this layer are also referred as consequent i i parameters. Layer 5: This layer is also called output layer, which computes the output as follows:
A A A Y1 = 1 f1 + 2 f2 + 3 f3

Now, the partial derivative of (17) can be found by the chain rule of differentiation as follows: 2 2 E 1 1 A 1 i = . (19) E 1 1 A 1 aA 1 aA 1 i On computing all the terms of (19) and putting in (17), we can nd the updated value of parameter aA 1 as follows:
A aA 1 (n + 1) = aA 1 (n) 2 (n) k f1 (n) B 1 (n)

Oi = i fij = i (ai + ai Xj ), 0 1

A 1 (n) 1 2 (n) + 3 (n) . 1 (n) + 2 (n) + 3 (n) bA 1 (n) aA 1 (n) (20)

Similarly,
A bA 1 (n + 1) = bA 1 (n) + 2 (n) k f1 (n) B 1 (n)

A 1 (n) 2 (n) + 3 (n) . 1 (n) + 2 (n) + 3 (n) bA 1 (n) aA 1 (n) (21)

(13)

SINGH AND CHANDRA: APPLICATION OF ADAPTIVE NETWORK-BASED FUZZY INFERENCE SYSTEM

169

In the same manner, the precondition parameters for the remaining fuzzy membership functions can be derived as follows:
A bA 2 (n + 1) = bA 2 (n) 2 (n) k f2 (n) B 2 (n)

1 A 2 (n) 1 (n) + 3 (n) 1 (n) + 2 (n) + 3 (n) bA 2 (n)

(22)

A aA 3 (n + 1) = aA 3 (n) 2 (n) k f3 (n) B 3 (n)

A 3 (n) 1 1 (n) + 2 (n) 1 (n) + 2 (n) + 3 (n) bA 3 (n) aA 3 (n) (23)

A bA 3 (n + 1) = bA 3 (n) + 2 (n) k f3 (n) B 3 (n)

A 3 (n) 1 (n) + 1 (n) . 1 (n) + 2 (n) + 3 (n) bA 3 (n) aA 3 (n) (24)

Fig. 7.

Powerspeed characteristic of PMSG.

Similarly, the precondition parameters for the fuzzy membership functions of second input can be found just by interchanging the subscripts and A B in (20)(24). Consequent-parameter tuning: To tune the consequent parameters, as discussed in Layer 4, the following updated laws are developed:
2 ai (n + 1) = ai (n) c , 0 0 a0 i

where is the air density, V is the wind velocity, and r is the radius of circular swept area of rotor blades. However, for a given turbine, Cp is not always constant and heavily depends on the tip speed ratio , which is given as = r r Tip speed = . Wind speed V (32)

i = 1, 2, 3 i = 1, 2, 3

(25) (26)

ai (n + 1) = ai (n) c 1 1

2 , ai 1

where c is the learning rate for consequent parameters. The derivative terms in (25)(26) can be found by the chain rule as already discussed in case of precondition parameters as follows:
2 2 E i , = a0 i i E a0 i 2 2 E i = , E ai a1 i i 1

i = 1, 2, 3 i = 1, 2, 3.

(27) (28)

Consequently, different wind speeds will require the optimal values of tip speed ratio (optim al ) to achieve the optimum value of power coefcient (Cp,optim al ). The aforementioned aspects make it very clear that to extract maximum power out of the varying wind, we need to have a WECS that allows the change in rotor speed to achieve optimal aerodynamic conditions. Under optimal aerodynamic conditions, the maximum power output at different wind velocities is almost the cubic function of generator speed, as shown in Fig. 7. Thus, to extract maximum power, the generator speed is needed to be controlled according to powerspeed characteristic [16]. The reference speed obtained from powerspeed characteristic is compared with speed estimated in previous section and the difference is applied to proportionalintegral (PI) controller. The output of PI results into a torque-controlling current component given as follows: i = q KP + KI S
(r r )

On substituting the terms derived in (27)(28) into (25)(26), the updated value of consequent parameters can be derived as follows: i ai (n + 1) = ai (n) 2c k , 0 0 1 + 2 + 3 i = 1, 2, 3 (29) i , i ai (n + 1) = ai (n) 2c k 1 1 1 + 2 + 3 i = 1, 2, 3. (30) Similarly, the consequent parameter for second input can be found just by replacing by in above equations. 5) Control of PMSG: The main function of variable-speed WECS is to extract maximum power at different wind velocities. The wind power captured by wind turbine depends on its power coefcient (Cp ) as per the relation given by PTurb = Pwind Cp = 1 3 r2 V Cp 2 (31)

(33)

where KP and KI are proportional and integral gains for generator speed control. The direct-axis reference current component can be set to zero in order to obtain maximum torque per amperage (MTPA), as the generator is always supposed to run below the base speed. The control action is implemented with the help of IGBT-based rectier. B. Grid-Side Inverter Control The proposed system employs a four-leg, grid-interfacing inverter, where the fourth leg of inverter is utilized to compensate the neutral current of 3P4W distribution network. Here, the inverter is a key element, since it delivers the power from renewable to grid, and also solves the power-quality problem arising due to unbalanced nonlinear load at PCC. The duty ratio of inverter switches is varied in a power cycle such that the combination of load- and inverter-injected power appears as

170

IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 26, NO. 1, JANUARY 2011

Fig. 8.

Control diagram of grid-side converter.

balanced resistive load to the grid, resulting in the unity power factor (UPF) grid operation. The schematic of grid-interfacing inverter control with 3P4W nonlinear unbalanced load compensation is shown in Fig. 8. Usually, the load currents, and either of the grid or inverter currents are required to compensate the load unbalance, reactive power, and harmonics [17][21]. This makes it necessary to have at least eight current sensors for a 3P4W distribution network. However, in the proposed control strategy, only gridcurrent sensing is required to force the grid currents to be purely sinusoidal at UPF without having any knowledge of inverterand load-current prole. The control of grid-side inverter comprises of two loops. The outer dc voltage control loop is used to set the current reference for active-power control. The output of dc control loop (im ) is multiplied with in-phase unity vector templates (Ua , Ub , and Uc ) to generate the reference grid currents (i , i , and i ). The reference grid neutral current (i ) is a c n b set to zero, being the instantaneous sum of balanced grid currents. The in-phase unit vector templates Ua , Ub , and Uc can be easily computed from three-phase grid voltages va , vb , and vc using phase-locked loop (PLL) [22]. III. SIMULATION RESULTS AND DISCUSSION The proposed system is simulated in MATLAB/Simulink environment using SimPowerSystem (SPS) blockset. The performance of ANFIS-based position and speed estimation for sensorless control of PMSG is demonstrated through the extensive simulation results. Besides this, a detailed comparison of proposed method is also provided w.r.t. the conventional EMF observer-based results under variable-speed operation, with different value of PMSG parameters. The variable torque is applied to the PMSG, and accordingly its speed and position are estimated. The traces of applied torque, generated voltage, actual and estimated speed, actual and estimated position, speed error, and generated output power are mainly shown in Figs. 9 and 10. In Fig. 11, the traces of grid voltages, grid currents, load currents, and inverter currents are shown. The traces of phase-a currents and neutral currents for grid, load, and inverter are shown in Fig. 12 w.r.t. phase-a grid voltage, just to demonstrate the load harmonics, and unbalanced- and neutral-current-compensation capabilities.

Fig. 9. Simulation results with nominal PMSG parameters. (a) Torque. (b) PMSG voltage. (c) PMSG actual and estimated speed. (d) PMSG actual and estimated position. (e) Speed error. (f) PMSG output power.

Fig. 10. Simulation results with 20% variation in PMSG resistance and inductance. (a) Torque. (b) PMSG voltage. (c) PMSG actual and estimated speed. (d) PMSG actual and estimated position. (e) Speed error. (f) PMSG output power.

In Fig. 9, the simulation results with nominal value of PMSG parameters are shown. A variable torque available from wind turbine is applied on the PMSG, as shown in Fig. 9(a). The corresponding change in amplitude and frequency of generator voltage can be easily seen in Fig. 9(b). The Fig. 9(c) shows the actual

SINGH AND CHANDRA: APPLICATION OF ADAPTIVE NETWORK-BASED FUZZY INFERENCE SYSTEM

171

Fig. 13.

Schematic of experimental setup.

Fig. 11. Simulation results showing the performance of grid-side inverter. (a) Grid voltages. (b) Grid currents. (c) Unbalanced load currents. (d) Inverter currents.

Fig. 12. Simulation results: grid, load, inverter currents of phase-a and neutral w.r.t. phase-a grid voltage.

and estimated speeds based on ANFIS and EMF algorithms, and their corresponding actual and estimated positions are shown in Fig. 9(d). The ANFIS estimator estimates the PMSG position and speed accurately, as evident from the trace of speed error in Fig. 9(e), whereas the EMF method estimates the speed with lot of oscillations and deviation from actual speed. The corresponding PMSG output power is shown in Fig. 9(f). The simulation results with 20% variation in PMSG resistance and inductance are shown in Fig. 10. The performance of EMF-based algorithm deteriorates with the variation in PMSG parameters, while the ANFIS estimator shows promising immunity to any such kind of variation, as evident from Fig. 10(e). Actually, the back EMF method, as in (1) and (2), requires the explicit mathematical model of PMSG. It can give satisfactory performance under ideal conditions, i.e., exact value of parameter must be known and the input signal should be distortion free. All these requirements make it unsuitable for wind-power application where the parameter of machine may vary with the passage of time. Moreover, the directly driven wind turbines are supposed to operate at lowerspeeds, where the output voltage may be very small with lower signal-to-noise ratio. On the other hand, the proposed ANFIS algorithm requires less-accurate mathematical model of PMSG, where it sees the real PMSG as a black box and tries to estimate the back EMF

adaptively by forcing the estimated currents equal to actual currents. Once the estimated currents are equal to actual currents for the same set of input voltages, the ANFIS model exactly replicates the actual model of PMSG, and hence, does not require the exact model of PMSG parameters. The proposed model automatically takes the parameter variation and distortion in signal into account and modies the estimated signal adaptively. The aforementioned facts are veried both in simulation and also with experimental results, where the ANFIS algorithm outperforms the conventional back-EMF method under dynamic operating conditions with nominal and 20% variation in PMSG parameter. The performance of grid-side inverter is shown in Fig. 11, where the grid-side currents, as shown in Fig. 11(b), are always kept as a balanced set of sinusoidal currents at UPF, even in the presence of 3P4W unbalanced nonlinear current shown in Fig. 11(c). The inverter supplies the load harmonics, load unbalance, load-neutral current, and generated power from PMSG, as shown in Fig. 11(b). Initially, the inverter current is constant and it starts increasing at t = 0.2 s in response to increase in PMSG input torque. As the inverter-injected current is more than the load-current demand, the difference of these is being injected into the grid with a sinusoidal and balanced prole. At t = 0.425 s, the input torque starts decreasing, which results in corresponding decrease in inverter and grid currents. In Fig. 12, it is shown that inverter current consists of loadcurrent harmonics component, load-current reactive component, and the active-current component in proportion to generated wind power. Moreover, the inverter is fully able to supply the load-neutral current locally, resulting in the zero-grid neutral current, and hence, balanced grid operation. On visualizing the prole of all traces carefully in Fig. 12, it can be easily noticed that the grid phase-a current is in phase opposition to grid phasea voltage. This indicates that the power is being injected into the grid at UPF. IV. EXPERIMENTAL RESULTS AND DISCUSSION A scaled hardware prototype is developed and implemented, as shown in Fig. 13. The PMSG is mechanically driven by a three-phase induction motor coupled with ACS800 drive from ABB. The variable torque caused by variable wind speed

172

IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 26, NO. 1, JANUARY 2011

Fig. 14.

Experimental results for PMSG voltage and current.

Fig. 16. Experimental results with 20% variation in PMSG resistance and inductance: (a) ANFIS estimator-based results and (b) EMF-estimator-based results.

command generated from MATLAB/Simulink le is applied to the drive, which, in turn, rotates the PMSG at variable speed. The output of PMSG is connected to grid by two back-to-back connected inverters with a common dc link. The proposed ANFIS speed and position estimation and control algorithm is implemented using dSPACE 1104. An extensive experimental study is carried out to validate the proposed strategy. The experimental results are shown in Figs. 1418. A. Generator-Side Control Performance The generator-side experimental results are shown in Figs. 1416. The variable amplitude and frequency of PMSG voltage, as shown in Fig. 14, indicate its variable-speed operation, where the speed is varied from 12% to 22% of its rated speed keeping in mind the low-speed operation of real wind turbine. The PMSG speed and position is estimated in real time. The experimental results based on ANFIS estimator and EMF estimator with nominal and varying parameters are shown in Figs. 15 and 16, respectively. The ANFIS estimator is able to

Fig. 15. Experimental results with nominal PMSG parameters. (a) ANFIS estimator-based results. (b) EMF-estimator-based results.

SINGH AND CHANDRA: APPLICATION OF ADAPTIVE NETWORK-BASED FUZZY INFERENCE SYSTEM

173

Fig. 18. (a) Traces of grid voltage, grid current, load current, and inverter current. (b) Traces of grid voltage, grid current, load current, and inverter current.

parameters and also loses control at regular interval under dynamic operating conditions, as shown in Fig. 16(b). B. Grid-Side Control Performance The performance of grid-side inverter is shown in Figs. 17 and 18, where the inverter is utilized as multifunction device. In Fig. 17, the experimental results are shown only for active ltering application of grid-interfacing inverter. The grid current prole is shown in Fig. 17(a), just before and after compensation. Here, it can be noticed that the grid is supplying a highly nonlinear 3P4W unbalanced current before compensation, which becomes a perfectly balanced set of sinusoidal currents once after the inverter is connected in the circuit. The inverter-injected currents are shown in Fig. 17(b), where the inverter not only supplies the nonlinear unbalanced component of load current, but also compensates the load-neutral current demand locally. This allows the grid to supply only the active balanced component of load current. Moreover, the traces of grid voltage, grid current, load current, and inverter currents are shown in Fig. 17(c), where the grid currents becomes perfectly sinusoidal

Fig. 17. (a) Traces of grid currents just before and after compensation. (b) Traces of inverter currents just before and after compensation. (c) Traces of grid voltage, grid currents, load current, and inverter current just before and after compensation.

estimate the PMSG position and speed accurately under nominal and varying parameter conditions, as shown in Figs. 15(a) and 16(a), respectively, while the EMF method shows poor estimation performance, as shown in Fig. 15(b). The performance of EMF estimator further deteriorates with the variation in PMSG

174

IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 26, NO. 1, JANUARY 2011

TABLE I SUMMARY OF INVERTER COMPENSATION CAPABILITIES

inverter can act as multioperation device in order to utilize its maximum rating. The current unbalance, current harmonics, and load-reactive power demand of an unbalanced 3P4W nonlinear load at PCC are compensated effectively such that the grid currents are always maintained as a balanced set (0% UF) of sinusoidal currents (2.7% THD) at UPF. Moreover, the load-neutral current is compensated locally by the fourth leg of inverter to keep the zero neutral current on grid-side, and hence, resulting in balanced grid operation. The grid-interfacing inverter injects the wind-generated power, load-reactive power, load unbalance, and load harmonics simultaneously, and hence, enables the grid to supply/absorb only fundamental current at UPF. APPENDIX SYSTEM PARAMETERS

and in phase with the grid voltage once after the inverter starts injecting the nonlinear, unbalanced, and reactive component of load current. In Fig. 18, the performance of grid-side inverter is shown, both in active ltering mode and renewable power injection mode simultaneously. The grid current is purely sinusoidal and starts decreasing with the increase in inverter-injected current, as shown in Fig. 18(a). This indicates that the grid is partially supporting the active-power load demand, while the inverter is supplying the rest of active load demand, load harmonics, and load reactive-power demand simultaneously. This fact can be easily veried from Fig. 18(b), where the grid current starts decreasing and becomes purely sinusoidal and in phase with grid voltage, once after the inverter starts injecting the current in the network. A comparative table showing the total harmonic distortions (THDs) and unbalance factor (UF) before and after compensation is given (see Table I), where the percentage UF is calculated separately for each phase as follows: %UFabc = |iabc iavg. | 100. iavg.

REFERENCES
[1] F. Blaabjerg, R. Teodorescu, M. Liserre, and A. V. Timbus, Overview of control and grid synchronization for distributed power generation systems, IEEE Trans. Ind. Electron., vol. 53, no. 5, pp. 13981409, Oct. 2006. [2] J. Morren and S. W. de Hann, Ride-through of wind Turbines with doublyfed induction generator during a voltage dip, IEEE Trans. Energy Convers., vol. 20, no. 2, pp. 435441, Jun. 2005. [3] G. Saccomando, J. Svensson, and A. Sannino, Improving voltage disturbance rejection for variable-speed wind turbines, IEEE Trans. Energy Convers., vol. 17, no. 3, pp. 422428, Sep. 2002. [4] T. Senjyua, S. Tamakia, E. Muhandoa, N. Urasakia, H. Kinjoa, T. Funabashib, H. Fujitac, and H. Sekinea, Wind velocity and rotor position sensorless maximum power point tracking control for wind generation system, in Proc. 30th Annu. Conf. IEEE Ind. Electron. Soc. (IECON 2004), vol. 2, pp. 19571962. [5] S. Taniguchi, S. Mochiduki, T. Yamakawa, S. Wakao, K. Kondo, and T. Yoneyama, Starting procedure of rotational sensorless PMSM in the rotating condition, IEEE Trans. Ind. Appl., vol. 45, no. 1, pp. 194202, Jan./Feb. 2009. [6] S. Bolognani, L. Tubiana, and M. Zigliotto, Extended Kalman lter tuning in sensorless PMSM drives, IEEE Trans. Ind. Appl., vol. 39, no. 6, pp. 17411747, Nov./Dec. 2003. [7] Z. Chen, M. Tomita, S. Doki, and S. Okuma, An extended electromotive force model for sensorless control of interior permanent-magnet synchronous motors, IEEE Trans. Ind. Electron., vol. 50, no. 2, pp. 288295, Apr. 2003. [8] P. P. Acarnley and J. F. Watson, Review of position-sensorless operation of brushless permanent-magnet machines, IEEE Trans. Ind. Electron., vol. 53, no. 2, pp. 352362, Apr. 2006. [9] S. Bhowmik and R. Spee, Wind speed estimation based variable speed wind power generation, in Conf. Rec. IEEE IECON 1998, Aachen, Germany, pp. 596601.

V. CONCLUSION The paper has presented a novel ANFIS-based speed- and position-estimation algorithm for the sensorless control of PMSG. The simulation and experimental results have been provided under variable-speed operation. The estimator is able to estimate the rotor speed and position accurately under both steady-state and dynamic conditions. A detailed comparison is also provided in between conventional back EMF and proposed ANFIS algorithm under nominal and 20% variation in PMSG parameters. The proposed ANFIS algorithm outperforms the back-EMF method, where ANFIS estimator shows great immunity against parameter variation. It has also been demonstrated that the grid-side inverter can perform all the duties of shunt APF while maintaining the smooth bidirectional power ow simultaneously. The simulation results supported by experimental results are provided to validate the fact that the grid-side

SINGH AND CHANDRA: APPLICATION OF ADAPTIVE NETWORK-BASED FUZZY INFERENCE SYSTEM

175

[10] K. Tan and S. Islam, Optimum control strategies in energy conversion of PMSG wind turbine system without mechanical sensors, IEEE Trans. Energy Convers., vol. 19, no. 2, pp. 392399, Jun. 2004. [11] H. Li, K. L. Shi, and P. G. McLaren, Neural-network-based sensor-less maximum wind energy capture with compensated power coefcient, IEEE Trans. Ind. Appl., vol. 41, no. 6, pp. 15481556, Nov. 2005. [12] J. Shing and R. Jang, ANFIS: Adaptive-network-based fuzzy inference system, IEEE Trans. Syst., Man, Cybern., vol. 23, no. 3, pp. 665685, May 1993. [13] N. Boccard, Capacity factor of wind power realized values vs. estimates, Elsevier Trans. Energy Policy, vol. 37, pp. 26792688, 2009. [14] Capacity Factors at Kansas Wind Farms Compared With Total State Electrical Demand, July 2007 to June 2008. (2009, Jun.). [Online]. Available: http://www.kcc.state.ks.us/energy/charts/Wind Capacity Factors at Kansas Wind Farms Compared with Total State Electrical Demand.pdf. [15] B. K. Bose, Modern Power Electronics and AC Drives. Upper Saddle River, NJ: Prentice-Hall, 2002. [16] M. Singh and A. Chandra, Power maximization and voltage sag/swell ride-through capability of PMSG based variable speed wind energy conversion system, in Proc. 34th Annu. Conf. IEEE Ind. Electron. Soc. (IECON 2008), Orlando, FL, pp. 22062211. [17] S. Rahmani, A. Hamadi, N. Mendalek, and K. Al-Haddad, Implementation and performance of cooperative control of shunt active lters for harmonic damping throughout a power distribution system, IEEE Trans. Ind. Electron., vol. 56, no. 8, pp. 29042915, Aug. 2009. [18] V. Khadkikar, A. Chandra, and B. N. Singh, Generalised single-phase p-q theory for active power ltering: Simulation and DSP-based experimental investigation, IET Trans. Power Electron., vol. 2, no. 1, pp. 6778, Jan. 2009. [19] P. Rodrguez, J. I. Candela, A. Luna, L. Asiminoaei, R. Teodorescu, and F. Blaabjerg, Current harmonics cancellation in three-phase four-wire systems by using a four-branch star ltering topology, IEEE Trans. Power Electron., vol. 24, no. 8, pp. 19391950, Aug. 2009. [20] J. Dai, D. Xu, and B. Wu, Implementation and performance of cooperative control of shunt active lters for harmonic damping throughout a power distribution system, IEEE Trans. Power Electron., vol. 24, no. 4, pp. 963 972, Mar./Apr. 2009. [21] J. H. R. Enslin and P. J. M. Heskes, Harmonic interaction between a large number of distributed power inverters and the distribution network, IEEE Trans. Power Electron., vol. 19, no. 6, pp. 15861593, Nov. 2004. [22] A. Timbus, M. Liserre, R. Teodorescu, P. Rodriguez, and F. Blaabjerg, Evaluation of current controllers for distributed power generation systems, IEEE Trans. Power Electron., vol. 24, no. 3, pp. 654664, Mar. 2009.

Mukhtiar Singh (S08) received the B.Tech. and M.Tech. degrees in electrical engineering from National Institute of Technology (formerly known as R.E.C. Kurukshetra), Kurukshetra, India, in 1999 and 2001, respectively. He is currently working toward the Ph.D. degree at Ecole de Technologie Superieure, Universite du Quebec, Montreal, QC, Canada, under the National Overseas Scholarship, funded by the Government of India. He was a faculty member at B.M.I.E.T., Sonepat, India, and K.I.E.T, Ghaziabad, India, during 2000 2002. Since 2002, he has been an Assistant Professor in the Department of Electrical Engineering, Deenbandhu Chhoturam University of Science and Technology, Sonepat, India. Currently, he is on study leave. His research interests include renewable energy sources, power quality, energy storage systems, electric vehicles, and power electronics and drives.

Ambrish Chandra (SM99) was born in India in 1955. He received the B.E. degree from the University of Roorkee [currently the Indian Institute of Technology (IIT)], Roorkee, India, in 1977, the M.Tech. degree from the IIT, New Delhi, India, in 1980, and the Ph.D. degree from the University of Calgary, AB, Canada, in 1987. He was a Lecturer, and later, was a Reader at the University of Roorkee. Since 1994, he has been a Professor in the Department of Electrical Engineer ing, Ecole de Technologie Sup rieure, Universi du e e Qu bec, Montreal, QC, Canada. His current research interests include power e quality, active lters, static reactive-power compensation, exible ac transmission systems (FACTS), and renewable energy resources. Dr. Chandra is a member of the Ordre des Ing nieurs du Qu bec, Canada. e e