Variable Pitch Fault Prediction of Wind Power System Based on LS-SVM of Parameter Optimization

The fault of the wind turbine pitch system is an important factor that causes the wind turbine to stop. In order to improve the accuracy of fault prediction, an intelligent algorithm for fault prediction of turbine pitch system based on Least Squares Support Vector Machines (LS-SVM) parameter optimization is proposed. Firstly, the data of SCADA system are analyzed, and four kinds of parameters, which are closely related to the turbine pitch system fault, are selected as the input of the model, and introduced the minimum output coding (MOC) to construct multiple classifications LS-SVM to realize multi-class classification of pitch fault. Secondly, the algorithm of particle swarm optimization is implemented to select the optimal feature parameters for the multi-class LS-SVM classifiers, and the classification accuracy of the PSO is taken as the fitness function value of the PSO. Finally, the model is applied to a wind farm 1.5 MW turbine. For comparison purpose, three widely used pitch fault prediction methods such as the BP neural network and standard support vector machines are utilized. The results show the proposed approach has a better performance both in training and testing accuracies, and provides an effective method for turbine fault identification and analysis.