Power Transformer Fault Diagnosis Based on Improved BP Neural Network

• DOI: 10.3390/electronics12163526
• Publication Date: 2023-08-21T12:53:31Z
• Authors: Yongshuang Jin; Hang Wu; Jianfeng Zheng; Ji Zhang; Zhi Liu
• Keywords: Dissolved Gas Analysis

Power transformers are complex and extremely important piece of electrical equipment in a power system, playing an role changing voltage transmitting electricity. Its operational status directly affects the stability safety grids, once fault occurs, it may lead to significant economic losses social impacts. The traditional detection methods rely on technical level system operation maintenance personnel, based Dissolved Gas Analysis (DGA) technology, which analyzes components dissolved gases transformer oil for preliminary diagnosis. However, with increasing accuracy intelligence requirements diagnosis DGA analysis method is no longer able meet requirements. Therefore, this article proposes improved residual BP neural network. This deepens network by stacking multiple modules, fuses expands gas feature information through In network, SVM introduced judge extracted vectors at each layer, screen out high accuracy, increase their weights. vector highest cumulative weight selected as input utilizes multi-layer mapping extract more differences after fusion expansion, thereby effectively improving diagnostic accuracy. experimental results show that, compared methods, proposed algorithm has higher diagnosis, rate 92%, can ensure sustainable, normal, safe grids.