Use of a Genetic Algorithm and Multi-objective Programming for Calibration of a Hydrologic Model


Abstract

A genetic algorithm was used to calibrate the RUNOFF component of the EPA storm-water management model, SWMM. A multi-objective function was developed which attached user-specified weights to error terms for estimates of peak flow rate, runoff volume, and time of peak. The genetic algorithm proved to be a valuable tool for isolating the neighborhood of the optimal parameter set. A conventional calibration scheme was used whereby the model was first fitted to a low intensity storm which produced runoff from impervious al eas only After parameters for the impervious cover were found, a larger storm was used to determine the variables for pervious land use. The calibrated model was used to simulate Two additional storms with good accuracy.