Continuous-time multi-model predictive control of variable-speed variable-pitch wind turbines

The control objective for wind turbine control systems includes maximum power tracking, speed and power regulation and minimisation of mechanical loads. The operating region of a wind turbine is a function of the incoming wind speed and the control objective in each operating region differ significantly from the other. Therefore, a linear controller designed based on a model obtained at one operating point cannot guarantee stability and satisfactory performance across the whole operating regime of the turbine. In this paper, a continuous-time multi-model predictive controller is proposed for variable-speed variable-pitch wind turbine systems. Four controllers were designed using linearised models obtained at 4, 8, 11 and $ 18\,{\rm m\,s}^{-1} $ 18ms−1. A Bayesian probability inference function was used to make the transition between these controllers based on the errors between the system outputs and each operating point. Simulation studies based on a benchmark 5 MW wind turbine were used to demonstrate the performance of the proposed controller.