Grey Wolf Algorithm-Enhanced Sensor-Less Integral Sliding Mode Control of DFIG on Wind Turbine Systems under Real Variable Speeds using ANN/MRAS

This study presents a robust Sensor-Less First Order Integral Sliding Mode (SL/FOISM) strategy, incorporating an innovative observer known as Artificial Neural Network with Model Reference Adaptive System/Adaptive (ANN/MRAS), specifically designed for wind turbine systems. The proposed model is implemented on a Doubly Fed Induction Generator (DFIG) operating under real variable speed conditions in the Adrar region of Algeria. The primary control objective is to independently regulate the reactive and active power of the DFIG stator. This is achieved through decoupling using the field-oriented control technique and control application via FOISM/C. An interesting feature of this methodology is the reduction in both the cost of the control scheme and the size of the DFIG by eliminating the need for a speed sensor. To enhance the Model Reference Adaptive System with Proportional-Integral (MRAS/PI), an ANN is introduced to replace the conventional PI controller in the adaptation mechanism of MRAS. The rotor position estimation is thoroughly examined across various load conditions, encompassing low, zero, and high-speed regions. The optimal parameters for the controller are determined through the application of Grey Wolf Optimization (GWO). The simulation results demonstrate the compelling performance of the proposed observer (ANN/MRAS), with rotor speed estimation errors reduced to less than 0.05% across all speed regions. The methodology ensures finite-time convergence, robust tracking of rotor speed with high accuracy, and resilience against parameter variations and load disturbances. Furthermore, the proposed control scheme achieves stable operation under variable speed conditions, showcasing adaptability and improved performance compared to the conventional MRAS/PI. Consequently, the estimated rotor speed converges to its actual value, demonstrating the capability to accurately estimate position across different speed regions (low/zero/high) while maintaining a maximum estimation error below acceptable thresholds.