Selection of Attribute Indicators in Transformer Fault Diagnosis Based on Fuzzy Clustering

Dissolved Gas Analysis (DGA) in Transformer Oil is one of the most effective methods for real-time diagnosis and analysis of potential faults in oil-filled electrical equipment. The traditional DGA analysis methods include feature gas method, three ratio method, Duavl triangulation method, etc. The fault boundaries of these methods are too absolute, which affects the accuracy of fault recognition. FCM classifies things based on similarity and soft partitions data samples based on membership, overcoming the problem of traditional methods having too absolute fault boundaries and effectively improving fault recognition rate. However, different fuzzy clustering attribute indicators affect the clustering effect of fault data and determine the accuracy of fault identification. In order to find a suitable set of parameters as the attribute indicators of fuzzy clustering, this article compares the clustering results using single gas, three pairs of ratios in the three ratio method, and three sets of parameters in the Duavl triangle as attribute indicators, It was found that using three sets of parameters (CH4%, C2H2%, C2H4%) in the Duavl triangle as fuzzy clustering attribute indicators can effectively distinguish different faults of transformers and have the highest fault recognition accuracy.