Fault Diagnosis of Power Transformer based on Probability-box Theory

In practice, fault diagnosis of power transformer is often performed on the basis of limited data and many various uncertainties factors. Under this circumstance, there are practical difficulties in identifying unique distributions as input for fault diagnosis. In order to solve the problem and improve the diagnosis ability of power transformer by analyzing the dissolved gas, a new method based on probability boxes theory was proposed. Firstly, the raw percentages of dissolved gas were used as the information source to construct the tow p-boxes about H-2 and C2H6 gas content. Then, to take advantage of the complementation of the information source, the tow p-boxes about H-2 and C2H6 gas content were fused. Finally, the SVM features database was established by extracting different types of cumulative uncertainty measures from p-boxes. The analysis result shows that the proposed method has high degree of diagnosis accuracy and is characterized by fast diagnosis and good real-time, demonstrating the model is practical and effective.