State of Health Estimation for Lithium-Ion Batteries Based on Features from Dynamic Time Warping of Expansion Curves

Abstract

Lithium-ion batteries (LIBs) have become the core energy storage components in electric vehicle powertrains due to their high energy density and long cycle life. The state of health (SOH) of LIBs serves as a critical indicator for evaluating their remaining lifespan and reliability. Conventional expansion-based SOH estimation methods typically extract only peak and valley points from expansion curves as features, failing to analyze the complete evolution patterns of entire expansion profiles. This limitation inadequately characterizes the expansion dynamics and compromises SOH estimation accuracy. To address these challenges, this study proposes a novel SOH estimation method based on features from dynamic time warping of expansion curves. Building upon analysis of charge/discharge expansion characteristics, correlations between battery aging states and surface expansion variations during cycling were established. The methodology employs dynamic time warping (DTW) algorithms to evaluate similarities between charge/discharge expansion profiles across different aging states, incorporating four key parameters as long short-term memory (LSTM) network inputs: reversible expansion, irreversible expansion, DTW-processed expansion curves, and single-cycle duration, with battery SOH as the network output. Under a$6: 1: 3$ratio for training, validation, and test sets, the proposed method achieves remarkable estimation performance, with MAE and RMSE values of 0.329 % and 0.390 % respectively. These results demonstrate the effectiveness of framework in accurate SOH estimation for LIB systems.