Study On The Combination Forecast Model Of Annual Rainfall Based On The Grey-neural Network

Precipitation is an important link of hydrological cycle system and also one of theimportant weather phenomenon. It is an important parameter in terms of forecastingfloods, management of water resources and hydrological forecast analysis flooding.Therefore, the degree of water resources in a region depends on its annual precipitation.That accurate prediction of annual precipitation is able to give water conservancy,agriculture and other departments to provide effective help. The precipitation has greatrandomness because has the characteristics such as complex and diverse weatherconditions. Generally, Prediction accuracy by using the traditional method to predictrainfall is really low. In order to improve the forecasting accuracy of annualprecipitation, this paper proposes a combination forecast model based on graywaveform prediction algorithm and improved wavelet neural network.Combination forecast model is prediction method that can an retain the usefulinformation in single forecasting model effectively. It can generalize prediction model,avoid the excessive fitting of the model and improve the prediction accuracy offorecasting model. In this paper, there are three parts of the research content has beendone：First of all, optimize the wavelet neural network.Research and analyze the principle of wavelet neural network prediction modeldeeply，and then found that the determination of parameters and thresholds depend on away to try. All of these not only spend more time, but also the effect of prediction is notvery good.Secondly, improve the grey waveform prediction algorithm.Through the analysis of grey waveform algorithm, there are two aspects of thedefects was found. They are the selection of the initial value and the study of thenonlinear problem. Based on the above problems, this paper improved the new greymodel NGM (1,1, k) of the initial value firstly，and then established the grey waveforecasting model based on the improved NGM(1,1,k) which is as the group of forecastmodel.Finally, improved the combination forecast model and analyzed the instance of theannual rainfall.In order to further improve the prediction precision, combination forecast model was set up in this paper. It is based on the parallel combination of the optimized waveletneural network prediction model and the improved grey waveform prediction algorithm.In order to verify the prediction effect of the prediction model, the instance of annualprecipitation has been analyzed through the Matlab-R2010a simulation tools. At first,the improved grey wave forecasting algorithm was applied to prediction of annualprecipitation in Chongqing, which proves that the effect of the forecast model is verygood. Then the improved combination forecast model was applied to the prediction ofannual precipitation in Chongqing. The forecast effect of the optimized combinationmodel was compared to the optimized wavelet neural network and the improved greywave prediction algorithm forecasting model. The results of the comparison not onlyverify the effectiveness of the combination one but also shows that it’s effect is betterthan the other two optimized single forecasting model.