Fault diagnosis of the transformer based on QPSO-SVM

Abstract In view of the shortcomings traditional dissolved gas analysis technology low diagnostic veracity and intelligence, this paper proposes to use QPSO optimize nuclear argument in support vector machine (SVM), on basis, (DGA) is used diagnosis transformer faults. Firstly, data preprocessed by DGA technology, processed as input amount fault characteristics. Secondly, for optimization core parameters SVM, algorithm combined with training acquisition. Finally, five kinds feature inputs are added model training, trained multi-classification correlation diagnose test data. After case studies comparative experimental analysis, accuracy method high 94.74%, relatively PSO-SVM, RVM methods, increased 5.11%, 2.12%, respectively.