Improved General Nonlinear Model of Batteries Using a Physics-Informed Genetic Algorithm

The state of charge (SOC) is a critical metric for effective operation and assessment of electric vehicle batteries. Various models are reported for estimating the SOC, among which the general nonlinear (GNL) model is notable for its ability to capture essential batteries' electrical characteristics, particularly the self-discharge effect. Nevertheless, identifying the parameters in GNL model presents significant challenges. Genetic algorithm (GA) methods exhibit good performance in this parameterization due to their global searching capability. Unfortunately, the undefined initial values and unconstrained searching space in existing methods for GNL models limit the derivation speed and necessitate a large resource. To address this issue, this paper proposes a physics-informed GA method for estimating the SOC. Specifically, the initial parameter set is obtained by averaging the measured values from multiple on-site battery tests, while the corresponding searching spaces are determined based on their upper and lower limits. Using the LiFePO4 C102F Blade battery as a case study, the results demonstrate the effectiveness of the proposed method, which can reduce the iterations by more than 10 times. Additionally, the influences of with/without initialization as well as different searching space sizes are discussed for clarity.