GPINND: A deep-learning-based state of health estimation for Lithium-ion battery

TL;DR

GPINND tackles real-time, mechanistically interpretable SOH estimation by integrating deep learning with electrochemical physics through a sequential training pipeline. A hybrid-driven surrogate provides a differentiable physical kernel learned from high-fidelity SPMe data, enabling a self-supervised, non-iterative aging-parameter identification from external measurements, followed by an SOH estimator that corrects residuals via data-driven mapping. The framework achieves a voltage reconstruction RMSE of $0.0198$ V and an SOH RMSE of $0.0014$, outperforming state-of-the-art baselines while delivering millisecond-level inference. This approach offers robust interpretability of aging mechanisms while maintaining high accuracy and computational efficiency, with potential extensions to multi-physics coupling and diverse battery chemistries.

Abstract

Electrochemical models offer superior interpretability and reliability for battery degradation diagnosis. However, the high computational cost of iterative parameter identification severely hinders the practical implementation of electrochemically informed state of health (SOH) estimation in real-time systems. To address this challenge, this paper proposes an SOH estimation method that integrates deep learning with electrochemical mechanisms and adopts a sequential training strategy. First, we construct a hybrid-driven surrogate model to learn internal electrochemical dynamics by fusing high-fidelity simulation data with physical constraints. This model subsequently serves as an accurate and differentiable physical kernel for voltage reconstruction. Then, we develop a self-supervised framework to train a parameter identification network by minimizing the voltage reconstruction error. The resulting model enables the non-iterative identification of aging parameters from external measurements. Finally, utilizing the identified parameters as physicochemical health indicators, we establish a high-precision SOH estimation network that leverages data-driven residual correction to compensate for identification deviations. Crucially, a sequential training strategy is applied across these modules to effectively mitigate convergence issues and improve the accuracy of each module. Experimental results demonstrate that the proposed method achieves an average voltage reconstruction root mean square error (RMSE) of 0.0198 V and an SOH estimation RMSE of 0.0014.

Figures (8)

• Figure 1: Schematic diagram of the SPMe model.
• Figure 2: Schematic diagram of the proposed GPINND framework.
• Figure 3: Schematic diagram of the sequential training strategy.
• Figure 4: The impact of different degradation mechanisms on the dQ/dV curves and internal concentration dynamics.
• Figure 5: Comparison of reconstructed voltage and the corresponding concentration dynamics for different methods on the simulation dataset.