Aiming at addressing low fault prediction accuracy in the case of inadequate characteristic gas data,this work tentatively proposed a new method to predict of transformer faults based on grey relevance vector machine by combining the grey theory and the relevance vector regression prediction.The main efforts entailed training of the grey relevance vec-tor machine,the establishment of a discrete grey model according to the data sequence of feature sample,the obtaining of the prediction model using prediction value of the discrete grey model as input and the original sample data sequence as output.The trained fault diagnosis method could then be used for fault prediction and fault determination according to rele-vant fault judgment criteria.The proposed method was indicated by comparative case analysis to have obviously superior accuracy compared to BP and SVM prediction methods,and thereby to be potentially effective and feasible.
power transformerfault predictionrelevance vector machinediscrete grey model