Remaining Useful Life Prediction for AC Contactor Based on MMPE and LSTM With Dual Attention Mechanism

Abstract

In order to realize the remaining useful life (RUL) prediction of AC contactor and improve the operation reliability of the low-voltage power distribution system, a RUL prediction method based on modified multi-scale permutation entropy (MMPE) and long short-term memory (LSTM) with dual attention (DA) mechanism is proposed. First of all, MMPE is used to analyze the performance degradation of AC contactor, mine the change law of feature parameters, and effectively detect the performance inflection point in the degradation process. Secondly, the LSTM with DA mechanism realizes the quantitative RUL prediction. In order to improve the RUL prediction performance, the feature attention mechanism and the temporal attention mechanism respectively assign weights to input features and time steps. Finally, a case analysis is carried out. The results show that the proposed method can effectively realize the quantitative RUL prediction, and the prediction error is smaller compared with the existing methods.