Research on Temperature Prediction and Fault Early Warning of Transformer Bushings

As a major component of power transformers, high-voltage bushings are susceptible to overheating failures due to factors such as the working environment, load changes, and equipment aging. High temperature not only damages the bushings themselves but also reduces insulation performance, increases the risk of electric shock and electrical breakdown, and can even trigger serious safety accidents such as fire and explosion. In order to detect and address bushings temperature abnormalities at an early stage, This paper presents A temperature prediction algorithm combining temporal convolutional neural network (TCN) and long short-term memory (LSTM) based on Attention mechanism is proposed. Additionally, an early warning model is established based on the prediction results. Statistical methods are utilized to analyze the distribution of prediction residuals and set the fault threshold interval. Finally, the actual bushing temperature data are input into the early warning model to evaluate the operating condition of the transformer bushing. The results demonstrate that the Attention-TCN-LSTM model exhibits faster convergence speed and higher prediction accuracy compared to other models. The established early warning model shows high sensitivity in assessing the operating condition of transformer bushings and can issue early warning information at the early stage of faults, providing sufficient time for maintenance personnel to prevent safety accidents resulting from deteriorating bushing faults and mitigating economic losses.