Research on Life Prediction of Surge Protective Device Based on Machine Learning

This paper proposes a novel method for predicting the life of surge protective devices (SPDs) based on machine learning technology. Firstly, the five key features of SPD and the data acquisition in the article are introduced. Secondly, the main machine learning methods used in the article are introduced. Meanwhile, a method of assigning residual life using the median life feature curve of metal oxide varistors (MOVs) is proposed. Subsequently, a prediction model of SPD remaining life is established using the random forest regression (RFR) algorithm. The performance indexes of linear regression (LR), support vector regression (SVR) and back propagation neural network (BP-NN) models are compared, which shows that the established prediction model has better performance. The importance of features on the model is also analyzed through the variance reduction of the random forest regression model statistics. The whole study provides valuable references for the application of machine learning in SPD life prediction and advances SPD maintenance strategies.