This article presents the design of a hybrid fuzzy sliding mode loss-minimisation control for the speed of a permanent magnet synchronous generator (PMSG) and a high-performance on-line training recurrent neural network (RNN) for the turbine pitch angle control. The back-propagation learning algorithm is used to regulate the RNN controller. The PMSG speed uses maximum power point tracking below the rated speed, which corresponds to low- and high-wind speeds, and the maximum energy can be captured from the wind. The sliding mode controller with an integral-operation switching surface is designed, in which a fuzzy inference mechanism is utilised to estimate the upper bound of uncertainties.