Abstract: Offshore wind farms are located in remote environment and have been constantly corroded by saline alkali. In order to solve the problems of multiple-fault detection and identification in the operation process of wind turbines, a model is established based on the traditional convolution neural network LeNet-5. The model adopts the ReLU function as the activation function, and a convolutional layer, a pooling layer and a full connection layer are incepted. Aiming at the datasets of the wind turbine supervisory control and data acquisition (SCADA) system and the condition monitoring (CM) system, a multi-category fault diagnosis is carried out. A cluster analysis is implemented on several wind turbines, followed by ensemble learning to build a multi-machine wind turbine fault diagnosis model. The experimental results indicate that the diagnostic accuracy of the proposed method is 97% ~ 99%. By comparing the experimental results and other algorithms, the effectiveness of the proposed method is verified.