The rainfall-runoff analysis model in urban watersheds should be constructed to establish flood damage countermeasures. The SWMM (Storm Water Management Model) is a representative model for rainfall-runoff analysis of urban watersheds. While this model is based on many parameters and provides relatively reliable results, it contains many ambiguous parameters. Therefore, parameter estimation is essential for rainfall-runoff analysis model and can be done using optimization algorithms. Harmony search algorithm is used to automatically estimate the parameters of the SWMM. Unlike the previous studies, the parameters are estimated by considering not only the inflow data but also the sewer level data. Parameter estimation is applied to the flood simulation on the catchment of Yongdap pump station basin, Seongdong-gu, Seoul, South Korea. The results estimated by supposed model are reliable in terms of both inflow and sewer level. The verification results of the calibrated model show the error within 5%, which are within the allowable error range.