Model Predictive Pitch Control of PMSG-Based WECS Using Fuzzy-MPC and Deep Learning-Based Wind Prediction

Maintaining steady output power in Wind Energy Conversion Systems (WECS) under varying wind speeds, especially above the rated wind velocity, poses a significant control challenge. To address this, a Model Predictive Control (MPC) approach for pitch angle regulation in WECS is proposed. The nonlinear dynamics of the wind turbine are first linearized and formulated using the Takagi-Sugeno (T-S) fuzzy modelling technique, allowing effective handling of system nonlinearity. The fuzzy-based linearized model is then integrated into the MPC framework, which incorporates both pitch angle and generator torque constraints to ensure safe and optimal turbine operation. To eliminate the reliance on physical wind speed sensors and enhance control performance, a Recurrent Neural Network (RNN) model is developed to predict future wind velocity based on turbine dynamics. This model uses Long Short-Term Memory (LSTM) cells to retain temporal patterns over extended periods, significantly improving the wind speed forecasting accuracy. The integration of this deep learning-based wind prediction into the MPC enhances robustness and adaptability to fluctuating wind conditions. The effectiveness of the proposed predictive pitch angle control strategy is validated on a Permanent Magnet Synchronous Generator (PMSG)-based WECS. Simulation results demonstrate that the proposed method improves power output stability and control responsiveness while reducing mechanical stress on the turbine components. This hybrid approach combining fuzzy modelling, predictive control, and deep learning offers a promising solution for advanced wind turbine control under real-world dynamic conditions.