A fault diagnosis based on LSSVM and Bayesian probability for wind turbines

Abstract

The wind turbine is affected by complex and changeable load and bad working environment, which leads to the randomness and uncertainty of the fault, so it is difficult to diagnose the fault accurately. To solve this problem, a fault diagnosis method based on LSSVM and Bayesian probability is proposed. Firstly, three inferences of Bayesian evidence framework are used to train LSSVM model and optimize parameters to improve the accuracy of fault diagnosis. Then, combined with Bayesian formula, the posterior probability of fault classification results is calculated, and the maintenance sequence of wind turbine is obtained. Finally, the proposed method is applied to the benchmark model of the 5-MW offshore wind turbine of the National Renewable Energy Laboratory (NERL), and compared with the traditional LSSVM and PSO-LSSVM methods. The results show that the proposed fault classification model has better classification accuracy, combined with Bayesian formula can accurately calculate the probability of fault occurrence and give the fault ranking results.