Research on temperature prediction of subway transformer based on LSTM

Abstract

The traditional subway condition monitoring system can only reflect the current temperature state and historical temperature trend of the transformer. The current operation state of the equipment can be judged by relying on the temperature threshold. But the system can not predict the future temperature change of the transformer. Transformer winding temperature is an important index to judge the operation state of the equipment. Accurate prediction of winding temperature can not only assist in judging the operation state of equipment, but also provide a reliable basis for the formulation of operation and maintenance plan for operation and maintenance personnel. The winding temperature is affected by multiple factors such as operating power and ambient temperature that often presents nonlinear changes. It was difficult to improve the accuracy of traditional prediction methods. With the development of deep learning technology, the transformer winding temperature was predicted based on long short-term memory network, and a large number of samples were used for off-line training. The relationship between multiple influencing factors and winding temperature was reflected by the transformer temperature prediction model. Finally, the transformer operation and maintenance data were input into the trained model to analyze and verify the feasibility and accuracy of the algorithm by comparing the real value and predicted value of winding temperature.