The electricity consumption of urban rail transit increases year by year with its rapid development. The regenerative braking energy generated by the train can be absorbed and reused by the ground energy storage systems, which can effectively reduce the traction energy consumption, so as to achieve the goal of low carbon and energy saving. It is necessary to consider how to configure its capacity. At present, ground energy storage systems mostly use intelligent algorithms for capacity configuration. Although this method improves the speed and accuracy of capacity configuration, it is essentially a fast traversal of the precise model. Although the number of traversals is reduced, due to the large scale of the precise model of the traction power supply system, the single solution speed is very slow, and its configuration efficiency is still not high. In this paper, on the premise that the substation energy can flow in both directions, the obtained substation rectification and inverter power curve is divided according to the power and capacity of the energy storage systems to be configured, so as to estimate the loss value of the braking resistor of the precise model, which improves the single solution speed, saving capacity configuration time. Taking a domestic subway line as an example, using the Elitist Non-dominated Sorting Genetic Algorithm (NSGA-II) to solve a simplified energy storage model based on the division of rectifier and inverter power curves to determine the capacity configuration scheme can improve the efficiency of energy storage capacity configuration.