A Novel Global-Local Multistage Algorithm for Fast Noninvasive Parameterization of P2D Lithium-Ion Battery Models

The pseudo-two-dimensional (P2D) model is a full-order physics-based model for lithium-ion batteries (LiBs), capable of providing accurate predictions of battery behavior. However, parameterizing the P2D model is complex due to its extensive parameter space, making parameter estimation (PE) particularly challenging. Thus, this research presents a novel, multistage, noninvasive algorithm to efficiently and accurately estimate these parameters. The algorithm operates in three stages. First, it defines essential inputs such as battery chemistry, parameter boundaries, and reference data. The second stage focuses on PE using a unique global-local search strategy. Here, particle swarm optimization (PSO) explores the global parameter space, followed by a novel parallel implementation of simulated annealing (SA) for refined local optimization. Finally, the third stage is a validation process that selects the output parameter set with the lowest prediction error across both estimation and validation stages. The proposed algorithm achieves a root-mean-square error under 17 mV, estimating 18 high-sensitivity parameters with an average error of 30%, and a rapid processing time of 52 min on average—among the fastest on record. This PE algorithm offers researchers and engineers a highly efficient, precise open-source tool for P2D model parameterization, potentially enhancing battery management and performance in practical applications.