This paper presents a fault diagnosis method based on CS and LS-SVM .Firstly, the main component analysis method is used to reduce the dimension of generator operation data, so as to reduce the complexity of parameter data and facilitate the acquisition of principal component features. Secondly, the least-squares support vector machine is used to construct training samples, obtain decision functions, and create a fault classifier model for the acquired principal component feature data to realize the classification of generator running state. Among them, the parameters c and g of the least squares support vector machine are optimized by cuckoo search algorithm. To verify the accuracy of the diagnosis model is created, two turbines were measured using a wind farm in Ningxia run data based on CS and LS-SVM validation of wind turbine fault diagnosis model, validated shows that after dealing with the dimension reduction of least squares support vector machine classifier can effectively identify generators of different fault, to verify the effectiveness of the design method.