Study On Runoff Forecast Of Jinping ? Hydropower Station And Comprehensive Evaluation Method Of Forecast Level

In the process of continuous development of hydrology,hydrological model is a vital technical means for the study of hydrological laws.It is usually the preferred method for hydrologists in hydrological forecasting.Due to different structural principles in various hydrological models,the forecasting effect is also different.By comparing the forecasting effects of different models,a hydrological forecasting model suitable for the study area can be explored to provide a certain reference for hydrologists.When evaluating the forecast level of the model,the traditional assessment method only assesses the forecast level through a series of correlation analysis of forecast error values without considering the different external scenarios and the corresponding forecast difficulty during forecasting,resulting in the inability to objectively and accurately evaluate the forecast level.Therefore,based on the Jinxi Reservoir as the research area,two hydrological models are selected to conduct runoff forecasting research,a comprehensive forecasting level assessment method considering the difficulty of forecasting is eatablished for forecasters considering the difficulty of forecasting.The main achievements include:(1)SWAT model and Xin'anjiang model are selected to simulate the daily runoff on Jinxi Reservoir,with the measurement of the certainty coefficient,root mean square error and Nash efficiency coefficient.By comparing the process of runoff forecasting and forecasting accuracy,the strengths and weakness of models with different structures in conducting runoff forecasting are analyzed.The results show that the Xin'anjiang model is superior to the SWAT model in terms of certainty coefficients,while the SWAT model shows better forecasting accuracy in terms of root mean square error and Nash efficiency coefficient.(2)Based on the prediction results of the two hydrological models,the Bayesian weighted average method is used for combined forecasting of the research stations.After dividing the flow rate(10% high flow,50% medium flow,40% low flow)according to the measured flow sequence,the forecasting results based on Bayesian model with the runoff forecasting effect of each model are compared.The results show that except in the forecasting effect of the periodic high water portion is not ideal,in most cases the Bayesian model can achieve higher forecast accuracy than SWAT model and Xin'anjiang model.(3)Based on the difference of external working conditions during forecasting,the concept of forecasting difficulty is proposed.The forecasting scenarios are divided in terms of the presence or absence of rainfall and flow level,the discrete error data and error distribution function are used to calculate the forecast difficulty under different forecast scenarios,based on this,a scientific and reasonable comprehensive forecasting level assessment method was established.Through case studies,compared with the traditional method for assessing forecasting level,the comprehensive forecasting level obtained by the research method mentioned in this paper is lower,but due to the distinction between forecast scenarios and the difference in forecast difficulty under different forecast scenarios,the results obtained are more fair and reasonable.