Capacity Estimation of Lithium‐Ion Batteries: A New Method Based on Anchor Points

ABSTRACT

The capacity of lithium‐ion batteries gradually declines with repeated cycling or prolonged storage, posing significant challenges for battery management systems (BMS). Accurate capacity estimation is therefore essential for ensuring the reliable and efficient use of batteries. This study proposes a novel estimation approach based on the concept of anchor points extracted from differential voltage (DV) curves during the charging phase. Unlike conventional methods that rely on multiple features from the incremental capacity (IC) and/or DV curves, the proposed approach only requires the extraction of state of charge (SOC) distributions at anchor points, avoiding complex feature engineering. A linear relationship between the capacity and SOC at the anchor points is established, simplifying the estimation process while retaining high accuracy. An in‐house dataset for second‐life batteries (SLBs) is developed through experimental studies to validate the method. In addition, three open‐source datasets are employed for further verification and performance evaluation. A comparative analysis against convolutional neural networks (CNN) and long short‐term memory (LSTM)‐based models, using mean absolute error (MAE) and root mean square error (RMSE), demonstrates that the proposed approach achieves comparable or superior accuracy. Estimation results based on real‐world electric vehicle data further demonstrate the practical applicability of the proposed method in BMS implementations.This study proposes a novel lithium‐ion battery capacity estimation method using anchor points extracted from DV curves during charging. The approach eliminates complex peak tracking and denoising procedures by establishing a simple linear relationship between capacity and SOC. Validation across in‐house and public datasets demonstrates high accuracy and low computational cost, making it well‐suited for large‐scale second‐life battery applications.