Lithium-ion Battery State of Health Estimation based on Cycle Synchronization using Dynamic Time Warping

TL;DR

This paper introduces a cycle synchronization method using dynamic time warping to align battery degradation data, enhancing the accuracy of state of health estimation with LSTM models by over 30%.

Contribution

It proposes a novel cycle synchronization technique that preserves information and enables equal-length inputs for SOH estimation models.

Findings

Improved SOH prediction accuracy by over 30%.

Effective handling of uneven cycle lengths in battery data.

Enhanced data utilization through cycle synchronization.

Abstract

The state of health (SOH) estimation plays an essential role in battery-powered applications to avoid unexpected breakdowns due to battery capacity fading. However, few studies have paid attention to the problem of uneven length of degrading cycles, simply employing manual operation or leaving to the automatic processing mechanism of advanced machine learning models, like long short-term memory (LSTM). As a result, this causes information loss and caps the full capability of the data-driven SOH estimation models. To address this challenge, this paper proposes an innovative cycle synchronization way to change the existing coordinate system using dynamic time warping, not only enabling the equal length inputs of the estimation model but also preserving all information. By exploiting the time information of the time series, the proposed method embeds the time index and the original measurements into a novel indicator to reflect the battery degradation status, which could have the same length over cycles. Adopting the LSTM as the basic estimation model, the cycle synchronization-based SOH model could significantly improve the prediction accuracy by more than 30% compared to the traditional LSTM.