基于EMD和混沌神经网络的月平均温度预测方法研究

o improve the accuracy of the mean monthly temperature prediction, and in view of the mean monthly temperature time series has nonlinear and non-stationary characteristics, the forecasting models for the mean monthly temperature were brought forward integrating empirical mode decomposition(EMD), the phase space reconstruction theory and neural network. Firstly, using the EMD theory,the mean monthly temperature time series is decomposed into several intrinsic mode function (IMF) components and a trend component, reducing the non-stationary in the signals. Then, using chaos theory, select the appropriate time delay and embedding dimension for each component, and reconstructs the phase space. By using the BP neural network, the component of decomposition is predicted. Finally the predicted results are added. Using the method to predict mean monthly temperature. The example indicates that the model can accurately reflect the change trend of mean monthly temperature time series,has higher prediction precision.