Abstract:

In the light of transformer fault diagnosis based on dissolved gas analysis (DGA) with a small sample size, poor information and the fault diagnosis results is easily affected by the noise in the sample, we proposed an intuitionistic fuzzy least squares support vector machine algorithm (IFLS-SVM). First we derived the related algorithm, and designed the multiclass classifier based on the IFLSSVM. Then we implemented the power transformers’ fault
diagnosis using the Matlab software. At last we compared the diagnostic result of the algorithm we proposed with the diagnostic results of the several LS-SVM multiclassification algorithms and BP neural network diagnostic result. Experiments results show that the IFLSSVM diagnosis is better, with stronger noise immunity.

Key words:  power transformers, fault diagnosis, intuitionistic fuzzy, least squares support vector machine (LS-SVM)