<p>Under the background of global climate change and accelerated urbanization, low impact development (LID) facilities have limited ability to control ' heavy rain ', and rainwater storage tanks have become the key measures for flood control and drainage in built areas. This study takes Kongtong District, Pingliang City, Gansu Province, China as an example. A three-stage multi-objective optimization model based on NSGA-II algorithm is constructed. By aiming at the minimum number of storage tanks, the minimum node overflow flow and the highest land score, the initial position is determined by using NSGA-II algorithm coupled with SWMM model. Gradient increment optimizes inlet pipe and outlet node length. Finally, the inlet pipe diameter, inlet offset and outlet flow were dynamically optimized with the cost of storage tank, overflow reduction rate and peak flow reduction rate as the objectives. The results show that when the cost of the optimization scheme offline storage tank is close to that of the traditional online storage scheme, the overflow reduction rate is increased by 10.07% at most. It has important theoretical and practical significance for improving the efficiency of waterlogging prevention and control in sponge cities and achieving disaster mitigation and economic benefit balance.</p>