To address the low accuracy and low stability of a single algorithm for transformer fault diagnosis, this dissertation is based on multi feature fusion diagnosis algorithm by combing support vector machine (SVM) and D-S evidence theory, The way to construct the basic probability assignment (BPA) of evidence has been improved by calculating the correct recognition rate and misdiagnosis probability of the SVM classification results. Simulation results show that this method can obtain more reliable belief function of the evidence, and further improve the accuracy of multi-feature fusion fault diagnosis.
展开▼
