新安江模型参数的线性化率定

摘要

为了解决一般概念性水文模型参数率定结果不稳定的问题，以新安江模型为例，提出了新安江模型日模参数的线性化率定方法。首先通过理想模型将该方法与SCE UA方法及单纯形方法进行对比研究。率定结果中3种算法所获得的平均目标函数值分别为0．02、0．10、8．39 m3/s ，平均循环次数分别为8、637、327，而且由线性化率定方法所获得的各参数值方差也要比其他2种方法小得多；说明线性化率定方法能够找到参数真值，计算精度和效率更高，率定结果更稳定。然后采用建阳和长滩河2个流域的实测资料对该方法进行应用检验，结果表明，同样可以较快地率定出稳定的模型参数优值。建阳流域的10组率定结果的目标函数值皆为100．35 m3/s ，循环次数也皆在8次以内，而且2个流域检验期的径流深相对误差皆在9．68％以内，确定性系数皆在0．819以上。因此，线性化参数率定方法确实能够解决非线性模型参数率定结果不稳定的问题，不会产生不相关的局部参数优值，并且不受参数初值影响，计算精度高，循环次数少，是一种可行有效的全局优值参数优选方法。%In order to solve the problem of unstable parameter calibration results of conceptual hydrological model ,the linearized calibration method of Xin’anjiang daily model parameters was put forward .Firstly ,this method was compared with the SCE-UA method and the Simplex method through an ideal model .The average values of objective function obtained by the 3 methods were 0 .02 ,0 .10 and 8 .39 m3/s , respectively and the average iteration numbers were 8 , 637 and 327 , respectively . Moreover ,the parameter variances obtained by the linearized calibration method were much smaller than those of other two methods .These results demonstrated that the linearized calibration method can find the true parameter values .It also has higher accuracy and convergence speed with more stable calibration results .Then the performance of the new method in model parameter calibration was examined using the measured data of Jianyang and Changtan River basin . The results showed that the stable optimal parameter values could also be quickly got .All of the objective function values of the 10 runs in Jianyang River basin were 100 .35 m3/s and the iteration numbers were all within 8 .In the validation phase of the two river basins , the relative errors of runoff depth were all within 9 .68% and the values of determination coefficient were all above 0 .819 .It can be concluded that the linearized calibration method can solve the problem of unstable calibration results about nonlinear model parameters without producing unrelated local optima .Furthermore ,it is not influenced by the different initial parameter values and has high computation accuracy and needs small iteration numbers . Hence , the linearized calibration method is an effective global optimization method .