Rainfall Forecast Of Wendeng City Based On Artificial Neural Network Model

Neural network unique non-linear adaptive information processing capability Served the traditional methods of artificial intelligence for the intuitive defects,so that it includes in the field of forecasting has been successfully applied in many areas. In recent years, neural Networks in a simulated human cognitive on the road have more in-depth development of a computational intelligence, and become an important direction of artificial intelligence.By the Matlab language typical of the activation of neural network transfer function, network designers can call as needed in the toolbox of neural network design training procedures, Use it to quickly achieve the modeling of practical problems to solve. Because of its simple programming, he can from the cumbersome programming extricate themselves, and into the practical issues to the study.In the long-term rainfall forecasts, use of all health function, statistical analysis, interpolation methods such as fitting, stressed that historical climate data and the future of the non-linear relationship between precipitation forecast is more complicated. These methods in different climate prediction reflects their respective strengths, but also because both are given a specific function, which means that precipitation will change the law, formulated However, the evolution of regional precipitation trends are often complex, dynamic And therefore such limitations of the method of prediction accuracy. Regional precipitation changes in the drivers are extremely complex, yet it is difficult to find appropriate characterization of the precipitation changes in climatic factors or a combination, the study still further explored. The artificial neural network forecasting methods based on the input and output variables on the nonlinear mapping, and it is only relevant to the objectives of training and samples. This method not only overcome the limitations of the specific function expression, but also through learning, training network to choose the optimal target value relative to predict.With use of the ANN toolbox in Matlab, based on the principal of forecast of Artificial Neural Network and BP and Elman model was applied by the authors in the field of rainfall forecasting. With the rainfall data of Wendeng City from 1953-2003, two rainfall forecast models with BP network and Elman network were set up separately. The change trend in the ten years of rainfall in Wendeng City was forecasted, and the difference between the two networks was discussed. The computation results showed that the BP model and Elman model had good quality on forecasting precision and generalization ability. Besides, it provided a new method for rainfall forecasting.From the forecast results can be seen BP and Elman network on historical data is very good approximation, the network in line with the forecast of historical development curve. The results showed that the predictive value of Elman network and Bp network have a number of different forecasts, but little difference in the network convergence of circumstances, Bp and Elman network have a similar output, but Elman network to further verify the reliability of the forecasts . This case is that of artificial neural network technology to forecast long-term rainfall data not only to enter a better fitting, more output can also be used to detect network performance forecasts. More important is the artificial neural network projections do not know predictors and forecasting of the function. As BP network and Elman network have stronger non-linear mapping capabilities, to avoid the forecast and prediction of factors determine the function. At the same time, because the black-box network effect, no forecast factor and the forecast of non-linear relationship, such as restrictions in polynomial simple function, enhance the reliability of the forecasts.