In this paper, a relay life prediction method based on service performance degradation parameters is proposed for the performance degradation of relays during long-term operation. Firstly, the soft threshold method is selected for wavelet noise reduction to process the relay performance data and remove the noise interference. Then, principal component analysis is used to extract the health factors affecting relay life from the original data, and finally the first principal component score is taken as the relay health factor and smoothed with the sliding average method. Finally, the extracted health factors are utilized in the LSTM-based relay life prediction model to predict the relay life, and the performance of the LSTM model is compared with that of the ARIMA model in relay life prediction, and the experimental results show that the LSTM model has better prediction performance. Meanwhile, compared with the traditional life prediction method, this method is more scientific and reliable, which helps to improve the accuracy of relay life prediction and reduce the operation and maintenance cost.