Research on Feature Extraction Methods for Fault Data of Power Transformers
The machine learning fault diagnosis method for power transformers based on dissolved gas analysis in oil(DGA)is currently the mainstream method for fault diagnosis of power transformers.However,this method has problems such as high dimensionality of fault feature data,significant nonlinear features,information redundancy,and parameter optimization,which can affect the accuracy of fault diagnosis.Using kernel principal component analysis(KPCA)to ex-tract feature parameters can effectively eliminate redundant information and reduce data dimensions.By comparing the classification accuracy and running time of the model before and after feature extraction,the effectiveness of the feature extraction method was highlighted,and the actual effect of the fault feature dataset after feature extraction on improving the accuracy and efficiency of machine learning model fault classification was verified.The superiority of KPCA is of great significance for discovering potential faults in power transformers,timely cutting off potential threats,ensuring the safe and stable operation of power transformers,and maintaining efficient and reliable power supply in the power grid.
power transformerdata extractionfault diagnosisKPCA
NETL
NSTL
万方数据
