Nonlinear control of a class of non-affine variable-speed variable-pitch wind turbines with radial-basis function neural networks

• Published in: ISA transactions 2022-12, Vol.131, p.197-209
• Main Authors: Bagheri, P., Behjat, L., Sun, Q.
• Format: Article
• Language: English
• Subjects: Adaptive control; Computer Simulation; Neural networks; Neural Networks, Computer; Non-affine systems; Nonlinear control; Nonlinear Dynamics; Torque; Uncertainty; Wind turbines
• ISSN: 0019-0578; 1879-2022
• DOI: 10.1016/j.isatra.2022.05.004

Due to complicated dynamics, wind turbines’ governing equations are subject to uncertainties and unknown disturbance sources. Despite uncertainties and disturbance sources, the paper’s focus is to design an adaptive controller that enables trajectory-tracking with a zero-converging tracking error. As the main result of a zero tracking error, the turbine can operate at maximum power efficiency. In addition, novel Lyapunov functions are proposed introducing auxiliary adaptive terms to result in closed-loop asymptotic stability in the presence of a non-affine controller input, uncertainties, and unknown disturbance sources. Considering the turbine dynamics, one can divide the wind turbine control problem into torque and pitch control phases. For addressing the nonlinearities and uncertainties of the dynamics in each phase, RBF neural networks are utilized to develop novel control and adaptive laws. To address the non-affine dynamics stemming from the pitch angle, a neural network alongside the implicit function and mean-value theorems are utilized to transform the dynamics into the control affine form. Several auxiliary adaptive variables are proposed in the transformation procedure, leading to closed-loop asymptotic stability. Moreover, using the Lyapunov stability analysis, closed-loop asymptotic stability is obtained for each phase. In the end, simulation results are presented to verify the analytical results where the proposed controller’s performance is compared to that of an existing method in different scenarios. The proposed controller’s simulation results suggest dramatic improvement over those of the existing method in both trajectory-tracking and required control action. •An adaptive robust controller for non-affine systems, guarantying asymptotic stability.•An adaptive neural network to cope with uncertainties and unknown disturbance sources.•Improvement of the adaptive laws to prevent the possibility of chattering.•Novel cross-term Lyapunov functions, relating the dynamics and approximation errors.•Significant performance improvement of the presented method over existing ones.