三水源新安江模型的函数曲面参数率定方法

目前常规的以误差平方和为目标函数的参数估计方法存在着诸如给非线性函数增加不相关的局部优值等问题,采用函数曲面参数率定方法展开样本函数的泰勒级数,给定参数初值,通过最小二乘估计新的参数向量。以三水源新安江模型为例,将该方法应用于理想模型及邵武、东张两个实际模型的日模参数优选中。理想模型的研究结果表明,该方法随率定参数个数的增多,迭代次数有所增加;实际模型的研究结果表明,由该方法率定的参数在率定期及检验期算出的结果相对误差普遍较小。

Parameter Estimation Method of Xinanjiang Three Sources Model Based on Function Surface

The current parameter estimation methods are mainly based on the error sum of squares and the information provided by the objective function has drawback. For example, linear parameter estimation methods that employ error sum of square as the objective function can increase unrelated local optima for nonlinear models. To solve the problem presented above, this paper adopts the parameter estimation method based on parameter function surface. To apply the method, we unfold the Taylor series of sample functions and get the new parameter values estimated by the least square method. Taking the Xinanjiang three resources model for an example, the method is applied to calibrated one ideal and the actual applications of Shaowu and Dongzhang. The studies of ideal model show that the iterations increase with the number of parameters. The studies of actual models show that the results of calibration period and validation period have low relative errors.