Study On Monthly Runoff Stochastic Model And Evaluation Method Of Simulation Effect

Hydrological stochastic model is a mathematical model to analyze the stochastic variation characteristics of hydrological process based on the hydrological observation data.A number of hydrological sequences simulated by the stochastic model have important application values for water resources evaluation,reservoir scheduling scheme formulation,determination of various water conservancy parameters,etc.However,it is difficult to reflect the dynamic and complex characteristics of the hydrological system due to the lack of evaluation indexes of stochastic models.Meanwhile,the applicability of stochastic models of hydrological system is worthy of further study.In order to solve the above problems,taking six typical hydrological stations in Yellow River Basin and Huaihe River Basin of Henan Province as examples,seasonal autoregressive(SAR)model,typical disaggregation(TD)model,complementary ensemble empirical mode decomposition-typical disaggregation(CEEMD-TD)model and non-parametric kernel density estimation(NP)model of monthly runoff stochastic models are established in this paper.The simulation effects of different stochastic models are evaluated by multi-level evaluation index system,and stochastic models of each typical hydrological station are selected.The main research contents and results are as follows:(1)A variety of monthly runoff stochastic models are established.According to the classification of hydrological stochastic models,SAR model,TD model and NP model are selected from regression,disaggreregation and non-parametric models as typical stochastic models in this paper.At the same time,a new hydrological stochastic model,CEEMD-TD model,is proposed in this paper in order to overcome the limitation that TD model is only applicable to stationary random series.(2)The evaluation index system based on runoff intersection and runoff process is established.The traditional evaluation index system is constructed based on the runoff intersectional statistical parameters,including six indexes reflecting the intersectional statistical characteristics of runoff series and three indexes representing the average runoff of multi-year flood season from July to September in the main flood season,with a total of 9 evaluation indexes.On this basis of the traditional evaluation index system,considering the non-linearity and dissipative characteristics of runoff as a time series,this paper puts forward two indexes to reflect the annual distribution process of runoff,i.e.,the maximum runoff sum of four consecutive months and the concentration degree,and two indexes to reflect the complexity of runoff change process,i.e.,approximate entropy and sample entropy,with a total of 13 indexes.(3)The stochastic model of each typical hydrological station is optimized.In view of the traditional evaluation index system and the evaluation index system constructed in this paper,grey correlation analysis method,set pair analysis method and fuzzy evaluation method are used to evaluate the simulation effect of each random model in six typical hydrological stations.The results indicate that the simulation effects of the optimal models selected by the 13 evaluation indexes system constructed in this paper are all superior to the optimal models selected by the 9 traditional evaluation indexes system,among which the simulation effect of CEEMD-TD model is better than that of SAR model,TD model and NP model.