To accurately estimate the State of Health (SOH) of substation direct current (DC) batteries and assist in DC system operation decisions, this paper proposes a substation battery SOH estimation method based on improved Gaussian process regression. First, based on actual substation operational experience, battery operating conditions are accurately classified, and corresponding battery state criteria are established for each condition. Based on the charge–discharge curve data of substation batteries under various operating conditions, Health Features (HF) of battery indicators are established for each condition, which then can form a battery health matrix. Furthermore, the Gaussian process regression algorithm is adaptively improved. A mapping between the Health Feature (HF) and the State of Health (SOH) is established by integrating actual historical operational data with offline test data of substation batteries proportionally. Experimental results indicate that this method demonstrates good estimation performance for batteries in the specific scenario of substations. Compared with other existing methods, it achieves more accurate fitting results and lower estimation error, thus providing a theoretical basis for the operation and maintenance of substation DC systems.
Ding et al. (2026) studied this question.