Wind energy is generally recognized as one of the nature-friendly energy resources compared with fossil fuels, geothermal energy, hydropower, and solar energy. Wind power is exploited by the wind turbines, which are utilized to convert the kinetic energy in the wind into electric power through the mechanical movement of the blades. Wind turbines mounted worldwide before 2017’s end can cover more than 5% of global electricity demand with overall capacity reached 539.291 GW 1. Globally, the total capacity of wind power added in 2016 and 2017 were 51.402 GW and 52.552 GW respectively 1. Thus, the rapid increase in the growth rate refers to the environmental and economic features of wind energy. From this perspective, improving wind turbines performance has become more crucial to meet the increasing demand for energy. Developing the wind turbine control systems is considered as one of ways to refine the performance as mentioned by 2-4.

The wind turbine control strategy is relatively dependent on wind speed. The installed controller objectives vary according to the wind turbine region of operation as shown in Fig. 1. The wind turbine operation can be divided into three major regions depending on the wind speed 5. If the wind speed is below the cut-in speed (i.e. ?3 m/s), it is considered to be in region 1. In region 2, the wind speed range is between cut-in wind speed and rated wind speed (i.e. 3 m/s to 11.4 m/s). To enhance energy efficiency and decrease mechanical stresses in the second, the control in generator torque is performed by maximizing the tip-speed ratio value (C_p) in-spite of variations in wind speed. In region 3, the wind speed range is above rated wind speed and below the cut-off wind speed (i.e. 11.4 m/s to 25 m/s). By focusing on the last region, regulating the rated generated power and reducing the mechanical loads can be considered as main purposes. Consequently, achieving these two objectives can be attended by designing of the pitch controller, which consisted of collective pitch controller (CPC) and individual pitch controller (IPC) respectively.

Fig. 1. Wind speed regions of operation of a typical 5MW wind turbine

Wind turbine control plays an essential role in modern wind energy conversion systems (WECS). Thus, to improve the quantity and quality of the produced clean energy generated by the wind turbine, various control strategies have been proposed for designing the wind turbine control system and summarized as follows. The conventional PI/PID (Proportional-Integral-Derivative) controller is generally used in the wind turbine industry for designing IPC, which only focuses on decreasing the mechanical loads in the turbine blades 3, 4. Then to handle the nonlinearities, gain scheduling techniques are widely used in the PI/PID parameter tuning. In 6, gain-scheduling PI controllers are designed for pitch control and torque control. By applying the gain-scheduling PI controllers, the performance of wind turbine is improved but still need to be optimized. For example, over-turbulent speed may not be well solved by a PI controller 7, 8. Similarly, gain scheduled-PI is designed using frequency response analysis by achieving a fixed-phase and gain-margins during operation in region 3 9. In 10, two methods of estimating the gains of a PI pitch angle controller are suggested. However, the previous studies based on PI controller provide a comparatively suitable steady-state response because the integral part can remove the steady-state error, they face distortion of the transient performance against variations in wind speed, system nonlinearities, and the wind turbine uncertainties. A nonlinear feedback-linearization is performed to decrease the mechanical loads and also for power regulation 11, 12. Although, these methods concern regarding the robustness, the linearized system is converted to Brunovsky form, which is not a robust solution as whose dynamics are totally different from the prime system, thus it is weak against the uncertainties. A mixed ?H_2 ; H?_? strategy is proposed to perform a trade-off between the pitch actuator rate limits and the wind turbine performance 13. A pitch controller with additional constraints (?H_2 ; H?_? trade-off with pole placement) is designed offline using linear matrix inequalities (LMIs) 14. Adaptive control technique is used to make a trade-off between the load reduction and maximum energy obtained 15. Although, the methods in 13-15 concern regarding the robustness, the pitch angle constraints (actuator and rate constraints) are not considered while designing the controller. The artificial intelligent control strategies (fuzzy logic and neural networks) are proposed to control wind turbines by considering generator speed and power as control inputs for the fuzzy logic control 16. In 17, an artificial neural network is utilized to control the pitch angle. The wind turbine dynamics are modeled by a back-propagation learning algorithm. In 18, an online two-layer neural network is utilized to estimate the unknown wind turbine aerodynamics. A mutual drawback towards controllers offered in 16-18 is that the pitch angle constraints are not considered in these methods. These pitch angle constraints are handled with model-predictive control (MPC) approaches.

Model-predictive control (MPC) is a sophisticated process control method that has been used in the industrial process control. MPC has the advantage of satisfying the system constraints. There are some control strategies use MPC in controlling pitch angle. A multiple linear MPC is presented to obtain maximum power and to regulate the pitch angle in order to keep the rated output power 5. In 19, a multiple MPC with finite control-set is designed to control pitch angle. In 20, 21, a multiple MPC is proposed to control the generator torque and the pitch angle. Takagi-Sugeno (T-S) fuzzy modeling is used in modeling the wind turbines and MPC is used to control the pitch angle as in 22, 23. The studies in 5, 19-23 use multi-models for pitch control design to overcome the nonlinearities of the system and to enhance the performance without guarantying closed-loop stability and the system stability in the transitions among these models. Moreover, at each sample instant, the online computational burden required to solve the MPC optimization problem may complicate the controller’s implementation.