A model-based predictive control for FTC for wind turbine wind speed sensor fault

This paper proposes an approach to fault tolerant control (FTC) of variable-speed wind turbine subject to wind speed sensor faults when the turbine is operating below rated wind speed. Both hardware and analytical redundancy are utilized in this approach. The least-Squares Support Vector Machine (LS-SVM) is proposed as the Fault Detection and Isolation (FDI) unit for the wind speed sensor fault. Single and multiple sensor faults are both considered in this strategy. LS-SVM is also used to estimate the effective wind speed (EWS), which is an unmeasurable signal. Based on the FDI unit, good FTC performance is achieved by identifying the faulty sensors and switching to the healthy sensor or the estimated EWS to provide the required controller reference signal. A robust MPC controller is then designed in order to consider the uncertainty due to error of EWS estimation and respect the physical system constraints.