Fast parameter identification of lithium-ion batteries via classification model-assisted Bayesian optimization

Lithium-ion batteries encompass a comprehensive set of parameters crucial for constructing an efficient battery management system. Utilizing parameter identification assisted by the pseudo-two-dimensional (P2D) model is far more cost-effective than employing direct measurement methods. Nonetheless, the time-consuming simulations associated with the P2D model can significantly hamper the efficiency of a parameter identification algorithm. This situation would be even worse when encountering inappropriate parameter vectors, which can cause the P2D model to fail to converge, consequently leading to further computational time consumption. To address these two issues, this paper proposes a classification model-assisted Bayesian optimization (CMABO) framework for parameter identification of lithium-ion batteries. In CMABO, Bayesian optimization is employed to search for optimal parameters. Its inherent capability to leverage the complete information conveyed by historical data renders Bayesian optimization sample-efficient, thereby enhancing the efficiency of the identification process. Additionally, a classification model is established to discern parameter vectors that could lead to unsuccessful simulations of the P2D model. This additional step of classification enhances the efficiency even further. CMABO is the first attempt to consider the failed simulations of an electrochemical model when identifying parameters. Simulations and experiments show that it is more accurate and efficient than some electrochemical model-based methods including genetic algorithm (GA), particle swarm optimization (PSO), and SA-TLBO. Besides, among different acquisition functions for Bayesian optimization, the lower confidence bound reveals the best performance.