Applied Research of the Rainfall Forecast Model Based on NNBR and Monte-Carlo Algorithm
Hexi Zhang, Daocai Chi, Yongtao Wang, Fuchen Wang · DOAJ (DOAJ: Directory of Open Access Journals) · 2014
From the statistical point, the complex historical process of rainfall was understood and analyzed, and the inherent law of its existence was found for the purpose of providing the theoretical basis for predicting the future rainfall. This paper had analyzed the Monte-Carlo algorithm and its distribution function, we had used the P-type Ⅲ distribution function to simulate rainfall and put forward the rainfall forecast model based on NNBR algorithm and the Monte-Carlo algorithm. Based on comparing Monte-Carlo predicted value to obserbed value, the result of error is less than 10%, the more cycling times, the less error, so the prediction model established in this paper can meet the production requirements basically. At the same time, the model based on NNBR Monte-Carlo algorithm is better than that of only using the Monte-Carlo algorithm of the ratio of 89.1%, indicating that the forecast accuracy presented in the paper is the best. Finally, with Beipaniiang drainage 50 years data，we had validated the result from the forecast model, and rainfall was predicted in the next five years. The results not only embody the randomness and statistical regularity of rainfall, but also reflect the rainfall sequence of timeliness, and the application effect is good.