On-Line Diagnosis Method for Transformer Winding Deformation Based on Running Voltage and Current Correlation Mining

摘要

When the transformer is subjected to a short-circuit or transport collision, the windings may be deformed under the action of electric power or mechanical force. Commonly used winding deformation diagnosis methods are all off-line diagnostic methods, which have the disadvantages of requiring the transformer to be out of service and high professional skills of the operator. To improve the accuracy and efficiency of winding deformation diagnosis, this paper proposed an online intelligent diagnosis method for winding deformation combined with information entropy (IE) and support vector machine (SVM). The method uses the permutation entropy (PE) and wavelet entropy (WE) to extract the characteristics of current and voltage signals, and comprehensively analyses the variation of the monitoring indicators in the complexity and time-frequency domain. The machine learning algorithm automatically learns the diagnosis logic from the fault features to realize the intelligent diagnosis of winding deformation. This study carried out training and testing on the real operating data of 29 transformers. The correct rate of diagnosis for deformation was 93.10%, and the correct rate of diagnosis for specific deformation positions reached 88.89%. This paper verifies the effectiveness of the online diagnostic model and the scalability between different types of transformers.