The Study of System Response Parameter Calibration Method for Xinanjiang Model Parameters

Abstract:

Non-linear function parameter calibration is usually based on the objective function of minimum error square sum, which including the construction of objective function and the first-order derivation of objective function. Since the traditional non-linear parameter calibration will introduce unrelated local optimums, a new non-linear parameter calibration method- the system response parameter calibration method was proposed in the paper. The nonlinear parameter can be linearized by this method which can reduce the un-related local optimums. In this study, the system response parameter calibration method was applied in the Xinanjiang model parameter calibration. The Xinanjiang (XAJ) model was generalized into a system. The changes of model parameters were considered as the input of the system. According to the system response curve caused by the changes of parameters need to be calibrated, the optimal parameters would be calculated using least square estimation based on the observed and calculated discharge. Firstly, the method was tested by an ideal scenario. The results showed that the method can find the given optimums very efficiently. Secondly, the performance of the linearized calibration method in model parameter calibration was studied using real data. The results showed that the system response parameter calibration method was an effective calibration method.