Lithium-Ion Battery Equivalent Circuit Modeling Using Gaussian Process-Based Global Optimization

Accurate battery modeling is essential for the reliability of Battery Management Systems (BMS), enabling precise state-of-charge estimation and improved operational efficiency. This paper presents a robust parameter identification framework for the second-order RC equivalent circuit model (2RC-ECM) using Gaussian process (GP)-Based Global Optimization (GPGO). The 2RC-ECM captures both ohmic and polarization effects, making it suitable for predicting voltage behavior. Existing optimization methods, including gradient-based and heuristic approaches, often suffer from local minima or high computational costs. To address these challenges, the proposed framework utilizes a Gaussian Process (GP) surrogate to efficiently approximate the mapping from model parameters to voltage prediction errors, thereby guiding the search and mitigating issues with local minima. The method is validated with experimental full-charge data from 18650 NMC cells under three ambient temperatures (0 °C, 25 °C, and 45 °C). Results demonstrate high-precision voltage tracking with Root Mean Square Errors (RMSE) consistently below 65 mV. Additionally, the algorithm adaptively adjusts impedance parameters to compensate for Open Circuit Voltage (OCV) uncertainties, trading strict local physical fidelity for improved global accuracy. These findings confirm the framework as a reliable and computationally efficient solution for practical BMS applications.