Transformer Fault Diagnosis Based on the Multiple Reconstruction Prediction and LS- SVR
QU Feng-chen · Electrical Measurement & Instrumentation · 2014
Transformer faults can be found through prediction of the dissolved gas in the transformer oil. With the state variables in the multivariate time series reconstruction as the inputs of the LS- SVR model,a transformer fault prediction model was proposed. Firstly,the prediction principle based on multiple reconstruction and LS- SVR theory were introduced. Then,the effects of the reconstruction parameters and LS- SVR parameters on predicting errors were discussed. The parameters were reasonably chosen to ensure prediction accuracy. Finally,the proposed method was used in the actual transformer fault diagnosis in order to verify the applicability of multiple reconstruction and support vector machine prediction. Compared with other predicting approaches,the proposed transformer fault combination predicting model based on the LS- SVR theory has higher prediction accuracy than any single predicting model or any other combination predicting model.