[Objective]Most parametric models are model-driven.They make certain assumptions about the probability distribu-tion and correlation relationships of hydrological sequences.However,the true distribution of hydrological series is complex and diverse.The correlation cannot be described solely by linear correlation.Therefore,there are differences between traditional model-driven stochastic models and real probability distributions.It is difficult to simulate the strong asymmetry or multimodality in the probability distribution of actual monthly runoff.[Methods]The Nonparametric Disaggregation Model(NPDM)can effec-tively simulate the randomness and correlation changes between runoff from a data-driven perspective.Considering the compre-hensive impact of annual runoff and previous monthly runoff,the Improved Nonparametric Disaggregation Model(INPDM)estab-lishes a conditional probability distribution function.With the Least Squares Cross Validation(LSCV)as the objective function to seek the optimal bandwidth,the model uses an optimization algorithm based on the golden section search and parabolic interpola-tion method,and combines variable kernel method to correct the boundary.To deeply explore the applicability of the two models in the Huai River Basin,NPDM and INPDM were established for stochastic simulation of annual and monthly runoff at Wujiadu Station from 1950 to 2007 and Lutaizi Station from 1951 to 2007 in the Huai River Basin.A series of statistical characteristic val-ues and related characteristics were used to compare and analyze the simulation effect.[Results]The result show that in terms of mean,both NPDM and INPDM have ideal simulation effects,achieving 100％control under one standard of mean square devia-tion;The standard deviation and variation coefficient of simulated sequences at Wujiadu Station and Lutaizi Station are controlled within the two mean square deviation standards;The INPDM outperforms traditional models in reproducing the maximum values of the original sequence,the correlation between monthly runoff,and the statistical characteristics of nonlinear state correlation,The proportion of months in which the second-order autocorrelation coefficient of NPDM at Wujiadu Station cannot be controlled under a mean square error standard is 50％higher than that of the INPDM model,and 16.6％higher than that of the Lutaizi Station.The first-order autocorrelation coefficient also showed similar result,indicating that the statistical characteristic values of the auto-correlation coefficient of the INPDM simulated sequence are closer to the statistical characteristics of the original sequence.But the NPDM has a stronger ability to describe the skewness coefficient of monthly runoff series.[Conclusion]In summary,it indi-cates that the improved nonparametric disaggregation model based on the optimization algorithm can fully reproduce the statistical characteristics of the runoff series,and ultimately provide a reference for conducting random simulation research on annual and monthly runoff in the Huaihe River Basin.
关键词
径流模拟/非参数解集模型/改进模型/可变核/淮河流域/适用性分析/影响因素
Key words
runoff simulation/nonparametric disaggregation model/improved model/variable kernel/Huaihe River basin/applicability analysis/influence factor
维普
万方数据