Robust State of Health Estimation for Lithium-Ion Batteries Considering Random Charging Behaviors

Abstract

Accurate and reliable state of health (SOH) estimation is crucial for the safe management of lithium-ion batteries in electric transport tools. However, the diverse charging behaviors of users pose significant challenges for SOH estimation, especially when applying machine learning methods. To address this issue, a robust SOH estimation method is developed that ensures accurate estimation under varying charging behaviors. The real-world operation data from 20 electric scooters are analyzed to characterize charging behaviors and determine the occurrence frequency of different charging voltages. Based on the voltage range of the characteristic point for the incremental capacity curve and voltage frequency, the charging behaviors are categorized into 15 classes. For each category, specific health feature extraction strategies are designed. Subsequently, a hybrid model combining a temporal convolutional network and a gated recurrent unit (GRU) neural network is proposed for SOH estimation. The proposed method is compared with different machine learning algorithms, and the influence of voltage noise and different charging behaviors are investigated. The results indicate that the proposed method can lead to an accurate SOH estimation with an error of less than 2%, even when faced with measure noise and battery cell inconsistency.