Abstract：The temperature of transformer oil is a critical indicator reflecting its internal insulation status. Traditional monitoring methods suffer from issues such as high promotion costs and insufficient real-time capabilities. This paper develops an oil temperature monitoring and early warnign system based on deep learning technology, employing a gated recurrent unit (GRU) network integrated with an attention mechanism to model and dynamically forecast oil temperature data. Tests on a measured dataset from a substation demonstrate that the Attention-GRU model outperforms traditional models across multiple metrics, including mean absolute error and the coefficient of determination. Deployed on a server, the system features a user-friendly human-computer interaction interface, enabling real-time monitoring of oil temperature values, prediction results, and abnormal warnings. This enhances operational efficiency and equipment safety.
张龙, 黄亭, 李建华, 魏小栋, 霍思佳. 基于注意力-门控循环单元的变压器油温监测预警系统研制及应用[J]. 电气技术, 2026, 27(7): 68-73.
ZHANG Long, HUANG Ting, LI Jianhua, WEI Xiaodong, HUO Sijia. Development and application of transformer oil temperature monitoring and early warning system based on attention-gated recurrent unit. Electrical Engineering, 2026, 27(7): 68-73.
2026, Vol. 27 
