Fault-Tolerant Economic Model Predictive Control for Wind Turbines

The operational cost of wind turbines (WT) is remarkably incurred in fatigue loads induced by torsional vibration within the drive-train subsystem and fore-aft bending of the tower subsystem. Under closed-loop control configuration, actuator faults in pitch subsystem and converter subsystem proliferate these fatigue loads, thereby, severely affect the economic operation of WT. In this paper, we present a novel active fault-tolerant control (FTC) methodology for WT, which minimizes the economic cost of WT by achieving the two broad objectives: power maximization and fatigue reduction, possibly under the effect of torque bias faults in converters. The proposed FTC system is composed of two modules: fault detection and diagnosis (FDD), and controller reconfiguration (CR). We develop the CR module using a model-predictive control (MPC) technique where the primary issue is that the constraint set is not convex in decision variables. The novelty of the proposed scheme lies in transforming the original nonconvex optimization problem into a convex problem using some new decision variables. We also develop the FDD module using an unknown-input-residual generator and a suitably designed estimation filter to extract the complete information of the fault. This fault information is subsequently used to reconfigure in realtime the constraints of the MPC to ensure system availability. The effectiveness of the developed scheme is demonstrated on a 2-MW wind turbine system.