• A novel RBF-compensated sliding mode ADRC is proposed for pitch control. • Angular velocity/acceleration saturation ensures safe high-inertia operation. • The method outperforms PID and ADRC in precision and disturbance rejection. • Combines ADRC estimation, SMC robustness, and RBF adaptability effectively. High-precision pitch angle control of wind turbines is critically important for maximizing wind energy conversion efficiency and ensuring stable grid integration, yet it faces significant challenges due to high-inertia blade dynamics and strong nonlinear disturbances inherent in electro-hydraulic servo pump control (EHA) actuators. To address these challenges, this paper proposes a sliding mode active disturbance rejection control (ADRC) strategy with radial basis function (RBF) neural network compensation. Specifically, a velocity-saturated tracking differentiator is designed to respect safe angular velocity and acceleration limits; the conventional nonlinear state error feedback is replaced by a fast terminal sliding mode law to enhance robustness and convergence; and an RBF neural network with adaptive weight update is employed to online approximate unknown system nonlinearities, while the extended state observer estimates external disturbances in real time. Simulation and experimental results on a ground-based test bench show that, under step command and random external disturbances, the proposed method achieves nearly zero overshoot (0.7°), a steady-state accuracy within ± 0.5°, and a rise time of 11.87 s. In comparison, conventional PID control yields 4.1° overshoot and ± 2° accuracy, while traditional ADRC gives 2.2° overshoot and ± 1.3° accuracy. These results lead to the conclusion that the proposed neural-network-compensated sliding mode ADRC significantly outperforms both PID and conventional ADRC in terms of overshoot reduction, steady-state precision, and disturbance rejection, offering a practical and high-performance solution for hydraulic wind turbine pitch systems under strong nonlinearities and high-inertia conditions.
Zhang et al. (Fri,) studied this question.