<p>The remaining service life of switchgear in MV distribution networks is influenced by numerous diverse and multifaceted factors. The real-time fluctuations of these factors increase the complexity of predicting the remaining service life of switchgear. To address this challenge, a method for predicting the remaining service life of switchgear in MV distribution networks is proposed, based on the attention-improved Informer deep learning approach. Analysis of the switching process and contact wear characteristics of switchgear reveals that the service life of switchgear is reflected in the contact wear state. This wear state is affected by the number of interruptions, the magnitude of the interruption current, and the duration of arc ignition. These three factors are used as parameters for estimating the remaining service life of switchgear. The estimation parameters are collected using an online monitoring device for switchgear contacts and input into a method for predicting remaining service life of switchgear based on the attention-improved Informer. The deep learning model for estimating the remaining service life of switchgear, based on the attention-improved Informer, employs a strategy gradient learning method. After training and setting the model weights and bias parameters, the model encoder extracts real-time changes in the influencing factors’ features, transmits them to the decoder, and outputs the remaining service life estimation results that correspond to the changes in the influencing factors’ features. The experimental results demonstrate that this method can accurately predict the remaining service life of switchgear (circuit breaker), with minimal prediction deviation.</p>