Improved Hybrid Gaussian Process Regression Model for State-of-Health Estimation of Li-ion Batteries

Accurate State-of-health (SOH) estimation con-tributes to the healthy and stable operation of the Li-ion batteries. This paper proposes an improved hybrid Gaussian process regression (GPR) method incorporating physical information for the high-accuracy estimation of battery SOH. The estimation process considers the effects of temperature, discharge rate, and depth of discharge (DOD) on the battery health, and the relevant features of the electrode degradation process are extracted for battery SOH estimation. The kernel function of GPR comprehensively considers the mechanical characteristics of battery degradation and can effectively track the battery degra-dation process. The validation experiments were conducted on the A123 LFP/Graphite battery dataset. Sufficient comparative experiments have validated the accuracy and robustness of the proposed method.