Author(s): S. Y. Liong; N. Muttil; Dr. Liong Shie-Yui
Linked Author(s):
Keywords: Shuffled complex evolution; Factorial design and central composite design; Global optimization; Conceptual rainfall runoff model calibration
Abstract: The successful application of a conceptual rainfall runoff (CRR) model depends on how well the model is calibrated. The estimation of CRR model parameters has proved to be a difficult task for several reasons, which include the existence of regions of attraction and multiple local optima. The Shuffled Complex Evolution (SCE) has been used extensively for the calibration of CRR models. Despite the SCE's having a high probability of finding the global optimum, the method was found to occasionally terminate prematurely. To further reduce the chance of such premature termination, a method of systematically generating the initial population of points for SCE, using the experimental design techniques of Factorial Design (FD) and Central Composite Design (CCD), is proposed. This study shows that SCE with initial population of points generated by FD-CCD, instead of a randomly generated one, reduces the algorithm's chances of getting trapped in local optima
Year: 2001