In the context of efficient integration of high-penetration renewable energy into distribution network，due to the relatively high ratio of resistance to reliance in the distribution network，the stochastic and fluctuating power output of distributed energy sources can easily lead to voltage violations in distribution network. Distributed modular energy storage can realize the functions of peak shaving and voltage regulation through the charging and discharging process，thus mitigating the voltage violations. However，the current cost of energy storage system is high. In order to give full play to the performance advantages of energy storage system and maximize the allocation benefits，a novel cardinality-constrained robust planning model is proposed to determine the optimal placement and number of distributed modular energy storage system for voltage regulation. By employing mixed-integer reconfiguration and dual linear programming，the proposed nonconvex nonlinear optimization problem can be relaxed to a mixed-integer linear programming problem，which is capable of being solved efficiently with optimization software. Finally，the effectiveness and feasibility of the proposed robust optimal planning method is verified using the IEEE-33node model as the test system. Through reasonable distributed energy storage configuration，the problem of voltage violations under extreme conditions can be solved.