Comparing Simulated Annealing (SA) and Particle Swarm Algorithms (PSO) for Optimizing Hydrological Parameters in Mahabadchay Watershed

Rainfall and runoff process is one of the main phases of the hydrological cycle and is simulated using hydrological models to examine the relationship between rainfall and runoff. The accuracy of the runoff estimation depends on the input parameters of the hydrological model including the sub-basins Curve Number (CN), Initial abstraction (Ia) and Lag time (Lt).

In this study, to more accurately estimate the simulated discharges computing through the hydrological model HEC-HMS in Mahabadchay watershed, the input parameters were calibrated. For this purpose, two evolutionary algorithms including Particle Swarm Optimization (PSO) and Simulated Annealing (SA) were used. In each iteration, the basin hydrological parameters are estimated using optimizer and given to the simulator. Afterwards, simulations are made. The Root Mean Square Error (RMSE) was used as objective function. The AutoIt software was used to automatically couple the optimization algorithms with HEC-HMS model. The rainfall-runoff data related to the 5 events of year 1387 were used to calibrate and validate the optimal parameters.

The results showed PSO convergence speed is more than SA algorithm in finding optimal objective function value and calibrating the hydrological parameters including CN, Ia, and Lt. Also, the optimal computed hydrograph found by SA algorithm had good agreement with that of by PSO one. Considering the results, it is concluded that the hydrologic parameters of watersheds without data can accurately be estimated by linking optimization and hydrologic models.