76 / 2025-04-01 16:52:57

Enhanced SOH Estimation for Electric Vehicle Power Batteries Considering Climate Variations and Driving Habits

State of health,Electric vehicle,Battery,Machine learning

Accurate and efficient state of health (SOH) estimation for electric vehicle power batteries is crucial for timely maintenance and facilitating battery reuse. This study proposes a precise and computationally efficient SOH estimation method based on four years of operational data from nine electric vehicles. First, battery interval capacity is estimated from charging segments in the raw dataset to generate SOH labels. Then, 15 candidate features encompassing battery characteristics, driving habits, and climate conditions are extracted, with 12 selected for SOH estimation. Finally, seven machine learning models are evaluated, with LightGBM chosen as the optimal model for mapping features to SOH. Cross-validation results demonstrate that the proposed method achieves an MAE of 2% in SOH estimation.