流出モデル定数の最適化手法

Abstract

Several optimization techniques are examined experimentally for finding what technique canbe utilized as a powerful tool for the parameter identification of three typical runoff models such asthe series tanks model, the storage function model and the kinematic surface runoff model. Thetechniques examined here are as follows: The conjugate direction method proposed by Powellnamed here the Powell method for short. The conjugate gradient method developed by Davidon, Fletcher and Powell called as the DFP method. The QG method combined the techniques ofquasi-linearization and golden section.It is pointed out that the standardization of variables is very usefull in applying the Powelland the DFP methods for the identification of a set of model parameters different in order. Atechnique for standardizing variables by a set of initial values is proposed and named as the SPand SD methods, respectively.It is shown that both SP and SD methods are excellent for the optimal identification of theseries tanks model with sixteen unknown parameters, and that both SP and Powell methods arevery useful for finding optimal parameters of the storage function model with three unknownparameters and also of the kinematic surface runoff model with two unknown parameters.Moreover, a simple method is proposed for estimating the lag time in the stroage function, and the interrelation between the storage function and the kinematic surface runoff models issuggested based on the identified parameters.