Region-aware cross-condition electrochemical parameter identification of high-capacity lithium-ion battery through sequential global–local optimization

Abstract

The widespread deployment of large-format lithium-ion batteries for grid-scale energy storage demands accurate and computationally tractable electrochemical models tailored to their unique characteristics. Existing parameter identification methods either lack robustness across varying operating conditions or struggle with the high-dimensional, correlated parameter spaces inherent to high-capacity cells. In this paper, we develop a region-aware sequential global–local optimization framework that integrates cross-condition sensitivity analysis, Bayesian exploration, and granular-adaptive local refinement to accurately and efficiently estimate parameters of physics-based battery models. We validate the proposed method using discharge data from a commercial 280 Ah lithium-ion cell across multiple C-rates, demonstrating consistent accuracy and fidelity, particularly in critical state-of-charge regions. Our results reveal the dominant roles of geometric and transport parameters across discharge regimes, offering physically interpretable insights while addressing longstanding challenges in model generalization. This work establishes a rigorous foundation for next-generation battery modeling and management, particularly under the demanding conditions of grid-integrated energy systems.