Life Prediction for Molded Case Circuit Breakers Based on BiGRU

Molded case circuit breakers (MCCBs) are critical components in power systems. Prediction of their lifespan can enhance equipment reliability, safety, and production efficiency while reducing maintenance costs and associated risks. This study proposes a novel lifespan prediction method for MCCBs based on a bidirectional gated recurrent unit (BiGRU). First, degradation data spanning the entire lifecycle of MCCBs were collected using an electrical life testing platform, and key features characterizing their health status and remaining lifespan were extracted. Subsequently, the exponential weighted moving average (EWMA) method was applied to enhance the trend of these features, improving the accuracy of the performance degradation description. A stacked autoencoder (SAE) was then employed to reduce and fuse the features, compressing the original 18-D feature space into six dimensions, thereby significantly reducing the data complexity. Spearman’s rank correlation coefficient (SRCC) was utilized to conduct the correlation analysis on the reduced-dimensional features, eliminating redundancy and constructing an optimal feature subset as input for the prediction model. Finally, a lifespan prediction model based on BiGRU was developed and trained. Experimental results demonstrate that the EWMA method effectively reduces prediction volatility, while the SAE approach better characterizes the performance status of circuit breakers, reduces model training time, and mitigates overfitting. Compared with traditional machine learning methods, such as support vector regression (SVR) and shallow backpropagation (BP) neural networks, the BiGRU-based predictions exhibit high consistency with the remaining lifespan trends of experimental testing samples and achieve higher prediction accuracy. This study provides an efficient and accurate method for predicting the lifespan of MCCBs, offering significant theoretical insights and practical engineering value.