Analysis of Transformer Operation State Based on Multi-dimensional Data Joint Analysis

Abstract

It is of great significance to accurate grasp the operating status of the power transformer and to discover the latent faults existing in the transformer in a timely and effective manner. Aiming at the problems of fewer parameters and lower diagnosis accuracy in power transformer evaluation model, this paper proposes a transformer state prediction model based on multi-dimensional data joint analysis. Comprehensively considering the oil dissolved gas data, partial discharge data, infrared data and environmental information data, we first use the sample case database to train the wavelet neural network, and then use the deep long short-term memory network that considers the time series attention mechanism to predict the parameters. The results are fed into the classifier, and we can obtain the operation result of the transformer. Tested on the actual transformer data, the experimental results show that the method has high prediction accuracy, which provides a reliable basis for maintenance personnel to predict the operating state of the transformer, and has certain application value.