ANFIS-sliding mode control of a DFIG supplied by a two-level SVPWM technique for wind energy conversion system

Received May 29, 2019 Revised Sep 18, 2019 Accepted Feb 19, 2020 A modified adaptive neuro-fuzzy inference system sliding mode control (ANFIS-SMC) by using two-level space vector pulse width modulation (SVPWM) for doubly fed induction generator (DFIG) is proposed in this article. ANFIS-SMC with SVPWM strategy improves the basic SMC performances, which features low stator active and reactive power and also minimize the total distortion harmonic (THD) of stator current. The computer simulation results, in Matlab, demonstrate the effectiveness of the proposed control strategy which improves the performance of the DFIG.


INTRODUCTION
Traditionally, DFIG based wind turbine systems (WTSs) are mainly installed in remote and rural areas [1]. Operation and control of DFIG has been the subject of intense research during last few years. The rotor of the DFIG is connected to AC-DC-AC converter and the stator is connected to the power grid [2]. The principal advantage of the DFIG is that the rotor side converter (RSC) is only sized for 30% of rated power compared to other generators used in variable speed WTSs. Various control strategies have been proposed for studying the behavior of DFIG-based wind energy conversion systems (WECSs) during normal operation. Indirect vector control [3] and a direct vector control [4] have been proposed for DFIG. In these techniques, the decoupling between q-axis and d-axis current is achieved with feedforward compensation, and thus the DFIG model becomes less difficult and PI controllers can be used [5].
In [6], direct vector control (DTC) was proposed to control DFIG. Similar to the DTC method, a direct power control (DPC) of DFIG based wind energy conversion system has proposed recently [7][8][9]. In [10], DPC control based on neural network (NNs) has been proposed. In [11], a modified DTC technique was proposed based on second order sliding mode controller (SOSMC) to regulate the electromagnetic torque and rotor flux. In [12], an reactive and active stator power proportional-integral controllers and space vector pulse width modulation (SVPWM) strategy were combined to replace the traditional hysteresis comparators. In [13], a modified DPC strategy was proposed based on SOSMC technique to regulate the reactive and active stator power of the DFIG-based WECS. SOSMC and fuzzy logic (FL) are combined to control the DFIG [14]. In [15], fuzzy sliding mode controller (FSMC) was designed to regulate the active and reactive power of the DFIG-based WTSs.
Neuro-sliding mode controller (NSMC) is proposed to control electromagnetic torque, reactive and active stator power of the DFIG [16]. Artificial neural networks (ANNs) and SOSMC are combined to control the stator reactive and active power of the DFIG [17]. Backstopping control was proposed to control DFIG [18]. In [19], LQR method was designed to control flux and electromagnetic torque of DFIG. In [20], neural space vector pulse width modulation (NSVPWM) and NSMC technique were combined to control DFIG. Fuzzy SVPWM strategy is proposed to reduce the total harmonic distortion (THD) of current and powers ripples of DFIG [21]. In literature [22], vector control (VC) is the most popular technique used in the DFIG based WECS. This control scheme is simple and easy to implement. On the other hand, this strategy gives more THD value of rotor current and powers ripples of DFIG.
To obtain high performance VC, a simple and robust ANFIS-SMC controller is designed to control the RSC and regulate the stator reactive and active stator power. The ANFIS-SMC controller is designed to avoid the reaching phase stability problem. The proposed scheme preserves the advantages of the conventional SMC such as simplicity, less parameters dependence and fast response. In addition, axes transformation of the stator voltage or current is not required. The stability of the ANFIS-SMC controller is proven using lyapunov stability theorem. Finally, the proposed and conventional SMC techniques performance is verified by the simulation study on 1.5 MW DFIG systems under variation parameters, harmonic distortion of stator current and reference tracking.

MATHEMATICAL MODEL OF DFIG
The general electrical state model of the DFIG obtained using Park transformation is given by (1) where, Lr is the inductance of the rotor, Ls is the inductance of the rotor, M is the mutual inductance, Rr is the resistances of the rotor windings, Rs is the resistances of the stator windings, ψdr and ψqr are the two-phase rotor fluxes, ψds and ψqs are the two-phase stator fluxes.
Rotor and stator voltages: where, Vdr, and Vqr are the rotor voltages, Vqs and Vds are the two-phase stator voltages, Idr, and Iqr are the two-phase rotor currents, Ids and Iqs are the two-phase stator currents.
Reactive and stator active powers: where, Ps is the stator active power, Qs is the stator reactive power.
Electromagnetic torque is done as: where, Te is the electromagnetic torque, p is the number of pole pairs. And its associated motion equation is: (5) where, Tr is the load torque, Ω is the mechanical rotor speed, J is the inertia, f is the viscous friction coefficient.

SVPWM TECHNIQUE
Space vector modulation strategy is widely used in variable speed drive of AC machine. This technique gives 15% more voltage output compare to conventional pulse width modulation (PWM). On the other hand, this strategy minimizes the THD of voltage and powers ripples this technique based on the principal of space vectors and need to calculate of angle and zone [25]. This strategy is detailed in [26]. However, this method of modulation is difficult to implement compared to PWM strategy. To avoid the disadvantages of the SMC technique, a new SVPWM scheme has been discussed in this paper [27]. This proposed strategy based on calculation of maximum (Max) and minimum (Min) of three-phase voltages [28].
The advantages of the proposed SVPWM strategy is not needed to calculate the zone and angle, simple scheme and easy to implement compared to classical SVPWM method. This proposed technique is detailed in [29]. The proposed SVPWM algorithm, which is designed to control the two-level inverter, is shown in Figure 1.

ANFIS-SMC CONTROL
In control system, SMC control is a type of variable structure control (VSV). It is a nonlinear control technique. The SMC method was proposed by Utkin in 1977 [30]. However, this technique is simple control scheme and easy to implement compared to traditional control techniques. Since the robustness is the best advantage of an SMC strategy [31]. But, this strategy has an essential disadvantage, which is the chattering phenomenon caused by the discontinuous control action [32]. This strategy gives more distortion harmonic of rotor voltage, electromagnetic torque ripple and powers ripples.
In order to improve the conventional SMC, a complimentary use of adaptive neuro-fuzzy inference system (ANFIS) controller is proposed. The ANFIS controller has been used in many application and this method is simple scheme and easy to implement. This strategy was proposed by Jang in 1995 [33]. This strategy based on observation and engineering experience. In ANFIS control, does not need a mathematical model of system [34]. On the other hand, the ANFIS controller structure consiste of four blocks that are defuzzification, fuzzification, knowledge base and neural network. One way to improve SMC technique performance is to combine it with ANFIS to form a ANFIS-SMC. The design of a sliding mode controller incorporating ANFIS control helps in achieving reduced chattering, simple control scheme, reduced harmonic distortion of voltage and robustness against disturbances and nonlinearities.

Sliding mode controller
The SMC technique goes through three stages, as follows: a) Choice of switching surface b) Convergence condition c) Control calculation.
The sliding mode reactive and active stator powers controllers are designed to respectively change the d and q-axis voltages ( V qr * and V dr * ) as in (6) [27].
To obtain good performances, dynamic and a commutation around the surface, the command vector is imposed as follows [35]:  (8) where K determine the ability of overcoming the chattering.
The SM will exist only if the following condition is met: 0 .
.  S S (9) Conventional SMC technique as shown in Figure 2.

Anfis-sliding mode controller
The ANFIS-SMC goal is to control the reactive and active stator powers of the DFIG-based WTS. The ANFIS-SMC is similar to a conventional SMC strategy. However, the switching regulators term sat(S(x)), has been replaced by ANFIS controller as given by Figure 3. On the other hand, the ANFIS-SMC control gives more and more minimum THD value of rotor current compared to classical SMC method. Fuzzy system has 49 roles. This roles for the proposed system are given in Table 1 [23]. The membership function definition is shown in Figure 4. We use the next designations for membership functions   The training used is that of the algorithm, Gradiant descent with momentum & Adaptive LR. The convergence of the network in summer obtained by using the value of the parameters grouped in Table 2. For the two proposed ANFIS-sliding mode controllers in Figure 3, the structure of the ANFIS controller is shown in Figure 5. The proposed ANFIS-SMC with SVPWM strategy, which is designed to control reactive stator power and active power of the DFIG-based WTS, is shown in Figure 6.

RESULTS
The ANFIS-SMC control of a DFIG-based WTS is implemented with simulation tools of MATLAB/Simulink. The DFIG attached to a 398 V/50 Hz grid. Both control techniques ANFIS-SMC using SVPWM and conventional SMC using SVPWM technique are simulated and compared regarding reference tracking, rotor current harmonics distortion, and robustness against doubly fed induction generator parameter variations. The DFIG used in this case study is a 1.5MW, 380/696V, two poles, 50Hz; with the following parameters: Rs = 0.012Ω, Rr = 0.021Ω, Ls = 0.0137H, Lr = 0.0136H and Lm = 0.0135H. The system has the following mechanical parameters: J = 1000 kg.m2, fr = 0.0024 Nm/s. Figures 7 and 8 shows the THD of rotor current of the DFIG-based WTS obtained using FFT (Fast Fourier Transform) method for ANFIS-SMC control with SVM (ANFIS-SMC-SVPWM) and conventional SMC with SVPWM one respectively. It can be clearly observed that the THD is minimized for ANFIS-SMC control when compared to conventional SMC with SVPWM technique. Table 3 shows the comparative analysis of THD value.  Figure 11 shows the electromagnetic torque of the both strategies. The simulation result validates the torque ripple is reduced in the proposed control scheme based DFIG drive (See Figure 14).On the other hand, the ANFIS-SMC control scheme using the two-level SVPWM strategy minimized the reactive stator power ripples, electromagnetic ripples and active stator power ripples compared to the traditional SMC using two-level SVPWM technique, see   Table 4 shows the comparative analysis of THD value. Thus it can be concluded that the proposed ANFIS-SMC control with SVPWM strategy is more robust than the conventional SMC with SVPWM strategy.

CONCLUSION
In this paper, the ANFIS-SMC principle is presented and it is shown that with SVPWM technique for a two-level inverter. The simulation results obtained for the ANFIS-SMC with SVPWM technique illustrate a considerable reduction in reactive power ripple, active stator power ripple, electromagnetic torque ripple and THD value of rotor current compared to the conventional SMC utilizing two-level SVPWM strategy.