Due to the insufficient consideration of population distribution differences in existing site selection research, as well as the difficulty in meeting the needs of population refuge under the condition of dam-break floods, the site selection plan is insufficient. The population distribution was simulated by using land-use and land-cover change (LUCC) and nighttime light data, and the number and distribution of the affected population were determined based on the simulation results of dam-break floods. On the basis of screening qualified candidate shelters, the effective capacity and accessibility information under normal weather and extreme rainfall conditions were analyzed and quantified. A multi-objective site selection model was built by comprehensively considering factors such as population transfer efficiency, construction cost and capacity limitations of shelters. The NSGA-II algorithm was used to solve the final site selection plan. The method was applied to the Dafangying Reservoir in Hefei City, Anhui Province. The results show that the overall average relative error (MRE) of population simulation. for 35 downstream streets is 11.16%, which has high fitting precision and can truly reflect the distribution of population. The number of qualified refuge candidate shelters screened under normal weather scenarios is 352, and the number of refuge sites retained under extreme rainfall weather is 266. Population distribution and weather factors have significant effects on the choice of refuge. The population distribution and weather factors have a significant impact on the selection of shelters and the allocation of evacuees. [ABSTRACT FROM AUTHOR]