Prediction of gas concentration in transformer oil based on improved gated recurrent unit
WEI Yongpeng1, SU Yihui2, WANG Shengli1, ZHANG Xin1, WANG Heng2
1. Maintenance Company, State Grid Gansu Electric Power Company, Lanzhou 730050;
2. College of Electrical and Information Engineering, Lanzhou University of Technology, Lanzhou 730050
Predicting the dissolved gas concentration in oil-immersed power transformer oil can provide important reference for transformer condition assessment and fault diagnosis. In order to further improve the prediction accuracy, an improved gated recurrent unit (GRU) prediction model for gas concentration in transformer oil is proposed by combining RAdam optimizer and GRU. The improved model takes seven typical gas concentrations as input, uses RAdam optimizer to train the model, and predicts the corresponding gas concentration as output. The results show that the improved GRU model can better predict the change trend of dissolved gas concentration in transformer oil and expand the selection range of initial learning rate compared with the ordinary GRU model and long short-term memory (LSTM) model.
卫永鹏, 苏益辉, 王胜利, 张鑫, 王衡. 基于改进门控循环单元的变压器油中气体浓度预测[J]. 电气技术, 2022, 23(2): 55-60.
WEI Yongpeng, SU Yihui, WANG Shengli, ZHANG Xin, WANG Heng. Prediction of gas concentration in transformer oil based on improved gated recurrent unit. Electrical Engineering, 2022, 23(2): 55-60.