Wind turbines have been widely installed in many areas, especially in remote locations on land or offshore. Routine inspection and maintenance of wind turbines has become a challenge in order to improve reliability and reduce the energy of cost; thus adopting an efficient condition monitoring approach of wind turbines is desirable. This paper adopts extreme learning machine
(ELM) algorithms to achieve condition monitoring of wind turbines based on a model-based condition monitoring approach. Compared with the traditional gradient-based training algorithm widely used in the single-hidden layer
feed forward neural network, ELM can randomly choose the input weights and hidden biases and need not be tuned in the training process. Therefore, ELM algorithm can dramatically reduce learning time. Models are identified
using supervisory control and data acquisition (SCADA) data acquired from an operational wind farm, which contains data of the temperature of gearbox oil sump, gearbox oil exchange and generator winding. The results show that the proposed method can efficiently identify faults of wind turbines.