Improving operational stability of large-scale wind turbines through data-driven model predictive individual pitch control

Modern large-scale wind turbines (WTs) with pronounced structural flexibility are subjected to asymmetric and stochastic aerodynamic loads, posing challenges to their stable operation. Accordingly, this study proposes a data-driven model predictive controller-based individual pitch control (DMPC-IPC) for large-scale WTs. DMPC-IPC enhances operational stability through the comprehensive optimization of control objectives, which encompass the reduction of speed/power fluctuations, attenuation of asymmetric loads, and regulation of structural damping. To formulate the cost function for the optimization problem, a data-driven model is developed to predict the impact of pitch control on the nonlinear dynamic responses of WTs arising from structural flexibility. In the data-driven model, the surrogate model is established using arbitrary polynomial chaos expansion, while Gaussian Process regression quantifies the residual uncertainty associated with surrogate model mismatches. A novel First-order Yin-Yang Pair Optimization solver is exploited to compute the cost function across various prediction horizons efficiently. Ultimately, performance validations of the DMPC-IPC are conducted using the DTU-10MW WT under extreme operation gusty and normal turbulent wind conditions across the entire operational region. Results demonstrate promising improvements in speed/power stability and mitigations in structural loads, thereby paving the way for its implementation in large-scale WTs.