Proactive inspection through interpretable online degradation monitoring for heavy-duty EV batteries using pattern recognition

The increasing prominence of electric transportation underscores the critical importance of battery system reliability and longevity. This paper presents an interpretable online degradation monitoring method for heavy-duty electric vehicle (EV) batteries to identify the pre-failure inspection period. The proposed approach uses online battery sensor data to continuously analyze battery performance to predict the pre-failure interval without relying on traditional statistical assumptions regarding battery degradation. By considering the stochastic nature of EV battery degradation, the methodology seeks to identify subtle patterns indicative of performance deterioration through logical reasoning and pattern recognition. This proactive strategy facilitates timely interventions and preventive maintenance, thereby enhancing overall system reliability and safety. The effectiveness of the approach is validated using a 40-foot Electric Bus, demonstrating its capability in predicting pre-failure conditions. The proposed approach achieves an average accuracy of 98.8% with unseen samples across various testing and training ratios, illustrating the robustness of online monitoring.