Graph neural network-based lithium-ion battery state of health estimation using partial discharging curve

TL;DR

This work tackles the sensitivity of Li-ion battery SOH estimation to segment selection and distribution shift between offline training and online operation. It introduces a Matrix Profile–guided voltage-segment selection framework coupled with a graph convolutional network that models inter-cycle degradation by connecting cycles through Pearson similarity, enabling online SOH estimation from partial discharge curves. Key contributions include (1) automatic, information-preserving voltage-segment extraction, (2) a GCN-based estimator that fuses intra-cycle features with inter-cycle relationships, and (3) strong empirical results on the MIT Attia dataset showing RMSE around 0.006–0.009 and outperforming LSTM baselines. The approach improves robustness and applicability of SOH estimation in real-world, partial-discharge scenarios, with potential for transfer learning and adaptation to diverse operating conditions.

Abstract

Data-driven methods have gained extensive attention in estimating the state of health (SOH) of lithium-ion batteries. Accurate SOH estimation requires degradation-relevant features and alignment of statistical distributions between training and testing datasets. However, current research often overlooks these needs and relies on arbitrary voltage segment selection. To address these challenges, this paper introduces an innovative approach leveraging spatio-temporal degradation dynamics via graph convolutional networks (GCNs). Our method systematically selects discharge voltage segments using the Matrix Profile anomaly detection algorithm, eliminating the need for manual selection and preventing information loss. These selected segments form a fundamental structure integrated into the GCN-based SOH estimation model, capturing inter-cycle dynamics and mitigating statistical distribution incongruities between offline training and online testing data. Validation with a widely accepted open-source dataset demonstrates that our method achieves precise SOH estimation, with a root mean squared error of less than 1%.

Figures (7)

• Figure 1: (a): Lithium-ion battery SOH degradation of a randomly selected Battery #2 from the benchmark dataset in Attia2020, and (b): Different cycles discharge voltage of the Battery #2.
• Figure 2: Illustration of the GCN Estimation Model.
• Figure 3: GCN-based SOH Estimation framework using partial discharging cure
• Figure 4: (a) Discharge voltage time series and its matrix profile; (b) Voltage segment selection of each cycle
• Figure 5: (a): Discharge voltage time series and its matrix profile of Battery #2 [17], and (b): Discharge voltage segment of 100 time steps from $3.25V$ of Battery #2 [17].