The use of energy storage systems (ESS) can guard against many hazards caused by distributed power sources joining the distribution network, and the reasonable configuration of ESS is a prerequisite for their effective application. In this paper, considering the coupling between planning and operation, a multi-objective site selection and capacity model for ESS in a distribution network with distributed generation (DG) is established from three aspects: grid vulnerability measurement indicators, active power loss, and ESS-rated capacity. A reformative multi-objective particle swarm arithmetic is devised. The arithmetic introduces a quasi-adversarial learning strategy in the population update process to enhance the coverage and convergence rate of the solution, and adopts an adaptive split strategy to separate prematurely clustered particles according to the number of iterations, thereby enhancing the diversity of particles. This has the ability to escape the local optimum while ensuring astringency. Through analysis on the IEEE-33 node power distribution system, the rationality of the proposed model and algorithm in optimizing the location and capacity and operational strategy of distributed energy storage is verified, and it can effectively improve the operating economy and vulnerability of a power grid with a stronger global search capability. This work is supported by the National Key Research and Development Program of China (No. 2016YFB0900600).