Relay Lifespan Prediction Based on Service Performance Degradation Parameters

This paper introduces a novel relay life prediction technique that leverages degradation parameters of service performance to predict the relay lifespan. The process begins with employing the soft threshold method for wavelet denoising to refine the relay performance data, thereby eliminating noise interference. Subsequently, principal component analysis (PCA) is deployed to extract the key health factors that influence relay lifespan from the dataset. The first principal component score serves as the relay's health factor. These extracted health factors are integrated into an LSTM-based model designed for predicting relay lifespan. The predictive prowess of the Long Short-Time Memory (LSTM) model is compared with that of the Autoregressive Integrated Moving Average (ARIMA) model in the context of relay life prediction. The comparison indicates that the LSTM model's superior predicting capabilities. Moreover, this approach surpasses traditional life prediction methods in terms of scientific rigor and reliability, thereby enhancing the precision of relay life predictions and curtailing operational and maintenance expenses.