Peak-flow forecasting with genetic algorithm and SWMM

Abstract

The success of a catchment model is known to depend a great deal on the catchment-model calibration scheme applied to it. This paper presents the application of a genetic algorithm (GA) in the search for the optimal values of catchment calibration parameters. GA is linked to a widely used catchment model, the storm water management model (SWMM), and applied to a catchment in Singapore of about 6.11 km2 in size. Six storms were considered: three for calibration and three for verification. The study shows that GA requires only a small number of catchment-model simulations and yet yields relatively high peak-flow prediction accuracy. The prediction error ranges from 0.045% to 7.265%.