A fault diagnosis method for new energy power plants based on an improved SVM

New energy power plants generate vast amounts of operational data and experience highly variable operating conditions， posing a significant challenge when it comes to diagnosing faults in generator sets. For this reason， a fault diagnosis model based on an improved SVM （support vector machine） algorithm is introduced. Firstly， the concept and principle of SVM for new energy power plants are analyzed. A multivariate SVM classifier is used to optimize the SVM. Secondly， the fault signal extraction method and fault characterization method of photovoltaic power stations and wind farms are studied. Moreover， a fault diagnosis model is proposed. Finally， sample data are obtained from new energy power plants， and an improved SVM fault diagnosis model based on decision level fusion is built. Additionally， the model is trained using fault feature vectors. The results show that the fault diagnosis accuracy for photovoltaic power plants reaches 97.5%， while the fault diagnosis accuracy for wind farms reaches 98.09%， which verify the accuracy of the method.