Prediction Method of Remaining Useful Life of Expressway Power Supply Guarantee Equipment

To ensure the reliable and stable operation of expressway energy system under the condition of power failure, this paper proposes a method for predicting the remaining useful life of expressway power supply guarantee equipment based on time mode attention mechanism (TPA). Firstly, by analyzing the working scene of lithium battery uninterruptible power supply (UPS) and the difficulty of obtaining its monitoring data, a four-dimensional state space affecting UPS health indicators is established. Secondly, a network model combining TPA and BiLSTM is constructed, and the training parameters are determined by random search and artificial parameter optimization. Finally, the model is tested using the data of NASA lithium battery dataset. The results show that the mean absolute error (MAE) and root mean square error (RMSE) of the prediction method proposed in this paper are 55% and 47% lower than those of support vector machine regression (SVR), and 59% and 45.7% lower than those of convolutional neural network (CNN), respectively. The results obtained by the prediction model proposed in this paper are more accurate.