A Study On The Method Of Medium And Long-term Rainfall Forecasting

With the development of the economy, the disaster of storm and flood happens increasingly. If the disaster happens, the lives and properties of people will be suffered large damage, and the stability of society will be influenced. So, as for decreasing the damage of the disaster, it is very important significance that studying the rainfall forecasting.With studying a great deal of data on the medium and long-term rainfall forecasting, auto-regressive model, artificial neural network and Shepard interpolation model are used on the annual rainfall forecasting, in order to test if these models fit into the rainfall forecasting. And the result of the rainfall forecasting proves that these models for rainfall forecasting are practical and efficient. The weight changeable combination forecasting model and the weight unchangeable combination forecasting model are attempted to forecast the annual rainfall. Because combination forecasting model can synthetically use the information that is provided by every single model, the shortage of every single model can be avoided. Generally speaking, the result of combination forecasting model is balanced, and the precision of it is better than that of every single model.In the thesis, the parameters of all models are optimized by RAGA (Real coding based Accelerating Genetic Algorithm). RAGA has the ability, which is to deal with the complicated problem. And RAGA can steadily adjust and compress the variable changed space. The times of accelerating circle is not many, commonly, the times is under 10 times. Even though the problem is very complicated, the times of accelerating circle is less than 50 times. So the workload of RAGA is comparatively small.