Fixed Point Neural Acceleration and Inverse Surrogate Model for Battery Parameter Identification

TL;DR

The paper addresses slow and costly parameter identification for lithium-ion battery models under dynamic EV loads. It introduces NeuralSPMe, a physics-embedded neural surrogate of the SPMe, and PUNet, an inverse surrogate that performs fixed-point parameter updates via a transformer-based estimator, enabling rapid convergence. Together they achieve over $2{,}000\times$ acceleration and more than $10\times$ improvement in accuracy over traditional CMA-ES-based approaches, with practical identification times around $1.32$ s under EV-like drive. This framework has strong potential to accelerate battery health assessment and parameter calibration in real-world BMS deployments by leveraging dynamic profiles and physics-informed learning.

Abstract

The rapid expansion of electric vehicles has intensified the need for accurate and efficient diagnosis of lithium-ion batteries. Parameter identification of electrochemical battery models is widely recognized as a powerful method for battery health assessment. However, conventional metaheuristic approaches suffer from high computational cost and slow convergence, and recent machine learning methods are limited by their reliance on constant current data, which may not be available in practice. To overcome these challenges, we propose deep learning-based framework for parameter identification of electrochemical battery models. The proposed framework combines a neural surrogate model of the single particle model with electrolyte (NeuralSPMe) and a deep learning-based fixed-point iteration method. NeuralSPMe is trained on realistic EV load profiles to accurately predict lithium concentration dynamics under dynamic operating conditions while a parameter update network (PUNet) performs fixed-point iterative updates to significantly reduce both the evaluation time per sample and the overall number of iterations required for convergence. Experimental evaluations demonstrate that the proposed framework accelerates the parameter identification by more than 2000 times, achieves superior sample efficiency and more than 10 times higher accuracy compared to conventional metaheuristic algorithms, particularly under dynamic load scenarios encountered in practical applications.

Figures (11)

• Figure 1: The proposed framework.
• Figure 2: The architecture of NeuralSPMe $\bm{\Phi}$.
• Figure 3: The architecture of PUNet $\bm{\Psi}$.
• Figure 4: An example of driving data and the generated current sequence.
• Figure 5: Histogram and CDF of voltage RMSE of NeuralSPMe and VT models: (a) large NeuralSPMe, (b) all models.