Modelling Flood Routing Using Hybrid Heuristic Algorithm

In the analysis of the flood routing problem, it is necessary to determine the change of the flood wave moving through the channel according to time and location, and in this context, nonlinear 3-parameter Muskingum model (NL-MUSK) is widely used. The parameters of the NL-MUSK model should be estimated precisely to hold the appropriate solutions. In general, this important process is solved by the use of heuristic methods, one of which is the Particle Swarm Optimization (PSO).This study aims to improve searching the local solution capability among the possible global solutions revealed by the particle swarm optimization (PSO) in a NL-MUSK flood rooting problem. To this end, the hybrid use of the PSO with the Levenberg-Marquardt (LM) algorithm, based on Jacobian matrix, was used. The developed algorithm (PSO-LM) was applied to the flood data observed in 1960 on the River Wye, U.K. The hybrid PSO-LM algorithm was operated under a wide solution space and the same global solution was obtained from each random experiment. The PSO-LM, which stands out with its stable aspect, has also achieved rapid convergence compared to the PSO, in other words, it has accelerated the access of the model to the optimum result with less iteration. In this respect, it is concluded that PSO-LM can be adapted effectively to different hydrological modeling studies.