Medium-long Term Runoff Forecast Model And Its Simulation Analysis

The runoff forecast is a kind of hydrological forecasting, furthermore, it is one of the most important part of that, it is also the scientific basis for flood control and power generation dispatching.When we are going to make research on something scientifically, to seek truth from facts is the ideological line, for a specific hydrological time-series forecasting problem, often need repeated experiments to find out the most suitable modelin the end.Ensemble forecasts, as a kind of forecast products, are often more useful thanindividual forecast. This paper uses the combining of wavelet analysis and timeseries analysis method to solve the long-term runoff forecast problems.This paper introduces several widespread methods of mid-long term runoff forecasting, and then expounds the theoretical knowledge of wavelet analysis, showing the powerful particular time-frequency processing capacity of wavelet analysis, furthermore, give way to the following discrete fast wavelet transform method which widely used in the practical application and it is named Mallat method; this paper also Introduces the wavelet de-noising theory, wavelet threshold method andits implementation steps; introduces the time series model and the method of model order and optimization, finally puts forward the forecast model of this paper: to use the ensemble forecast which have mentioned above, combines the wavelet analysis and time series model together.What’s more, this paper shows the set-up-model steps with specific examples:to do pretreatment on hydrological series, weeds out the data which have problems obviously, fills the missing data. Then, finds out the most suitable wavelet function. And uses the wavelet analysis to reveal the changing characteristics of specific hydrological series, dividing the original sequence into wavelet coefficientsat each scale.Then, this paper selects a appropriate threshold function and nice threshold, use wavelet denoising on the hydrological sequence. The do the wavelet reconstruction, toobtain the hydrological sequence after denoising. Finds the reasonable model order, thenregards the hydrological sequence after denoising as the input terminal of this model, toforecast the future runoff. At the end of this paper，puts the analysis and evaluation on theresults of the model prediction.