State-of-Health Estimation of Lithium-Ion Battery Based on Interval Capacity for Electric Buses
Bao‐Lin Ye, Zhaosheng Zhang, Shuai Wang, Yuhan Ma
Abstract
An accurate and reliable method for state-of-health (SOH) estimation of lithium-ion batteries in electric buses (EBs) is of great significance for learning about the current health status of EBs and the time of decommissioning, as well as facilitating the secondary use of batteries. In this article, the real-world operation data of 12 buses on four bus routes covering a span of two years were collected and preprocessed. A battery SOH estimation method was proposed leveraging the data characteristics of EBs. The method was based on interval capacity, while the state-of-charge (SOC) interval was selected, followed by the definition and calculation of SOH. Then, 29 SOH-related features were extracted from five aspects, voltage, current, SOC, temperature, and others. The features were screened sequentially by Shapley value analysis and the reversed stepwise regression algorithm (RSRA) we proposed. Based on the screened eight features, this article built six machine learning models and compared their performance in terms of SOH estimation. Finally, CatBoost, which showed the best overall performance, was selected as the optimal model for SOH estimation. To overcome the shortcomings of traditional methods, this article proposed a universal SOH estimation method for EBs, which achieved a mean absolute percentage error (MAPE) of 0.326% on real-world testing data.