Root Cause Analysis in Power Transformer Failure with Improved Intelligent Methods

Abstract

Root cause investigation, diagnosis and fault classification in power transformers are the essential features to look for when trying to ensure dependability and power quality with minimum interruptions. To achieve greater diagnostic precision, a novel method that combines Stacked Denoising Auto Encoder and Bidirectional Long-Short Term Memory (SDAE-BiLSTM) is indicated in this work. Dissolved gas analysis is the most effective method for determining the cause of electric power transformers(DGA) problems. All forms of faults can be differentiated by separating samples from power transformers. The SDAE-BiLSTM method has much research potential because it uses the dissolved gas in the transformers for analysis and fault diagnosis. A comparative study has been done using various machine learning models, such as Support Vector Machine, Random Forest and Convolutional Neural Network. Compared to the performance of these models, it is clear that the SDAE-BiLSTM model possesses superior accuracy because it has more parameters.