Storm Water Management Model calibration using multiobjective evolutionary algorithms

ABSTRACT

This paper focused on the calibration of Storm Water Management Model (SWMM) for the Cancela River Basin, located in Santa Maria, Rio Grande do Sul, Brazil, using the multi-objective Evolucionary Reference Point Based Non-Dominated Sorting Genetic Algorithm R-NSGA. Modifications were made in SWMM structure to allow its linking with R-NSGA in MATLAB. The objective functions proposed were the efficiency coefficient of Nash-Sutcliffe (eNS), the Peak Flow Error (EQ) and Runoff Volume error (EVOL), which were applied simultaneously to the model calibration. A method for determining the maximum compatibility to rank the best solutions is proposed. The parameters estimated by SWMM calibration were close to physical values of the basin, except those relating to the Horton equation in the most compatible solutions have a better behavior for the validation event, highlighting the importance of multi-objective optimization.

Keywords:
multiobjective optimization; peak flow; hydrologic models