Research on online diagnosis method of transformer winding deformation based on correlation mining of operating voltage and current

Transformer is one of the main equipment of power system, winding deformation will lay a hidden danger for the operation of transformer. Most of the existing winding deformation diagnosis methods belong to off-line diagnosis, which has the disadvantages of power cut and long interval period. In this paper, an on-line diagnosis method of transformer winding deformation based on the correlation mining of operating voltage and current is proposed. Through collecting practical cases, the monitoring data before and after transformer winding deformation are compared longitudinally by using the logic regression method, and it is found that only the monitoring data of voltage and current are significantly related to the deformation among the four monitoring indexes of current, voltage, power and oil temperature. Combined with the algorithm of permutation entropy, the data before and after transformer deformation are compared crosswise, it is found that the permutation entropy of many indexes of transformer with winding deformation before and after short circuit is obviously different, while the permutation entropy of each index of transformer without winding deformation is basically the same before and after short circuit. And the on-line diagnosis of winding deformation can be realized by analyzing the actual state of permutation entropy of key indexes of transformer. In this study, a total of 29 transformer winding deformation diagnosis was completed, 27 transformers were correctly identified, and the diagnostic accuracy rate was 93.10%, which verified the effectiveness of the method and its popularization among different types of transformers.