Transient Temperature Prediction of Three-Phase Gas Insulated Busbar Contacts Based on CNN-LSTM

In the temperature detection of a three-phase Community Gas Insulated Bus Bar (GIB), the spatial distribution relationship between the measuring point of the housing temperature and the internal three-phase contact is not clear, and the time lag between the housing temperature and the internal contact temperature is strong. This paper constructs CNN based on convolutional neural network (CNN) and long short-term neural network (LSTM) structures. Prediction of the transient temperature of the internal contact by CNN-LSTM. First, the physical simulation experiment of the temperature rise of the three-phase common body GIB shell was carried out, and the temperature data of the three-phase common body GIB shell and contacts under different test currents were obtained; then, the convolutional neural network was used to extract the time of each temperature measurement point Features; After that, the long and short-term memory neural network is used to capture the spatial features between different measuring points; finally, the transient temperature prediction value of the GIB internal contact of the three-phase common body is output through the fully connected layer (FC). The results show that the multi-step prediction accuracy of the CNN- LSTM model on the test set can reach more than 94%, which is better than prediction algorithms such as MLP, LSTM, and CNN. The model can be provided as a new method and reference for GIB overheat monitoring.