基于EMD与Elman神经网络的太阳黑子活动预测

摘要

A combined forecasting model based on empirical mode decomposition (EMD) and Elman neural network is established to tackle the problems of the nonstationary,chaos and unpredictable time series of sunspots.The time series of sunspots are decomposed into intrinsic modal functions and residual components of different time scales by empirical mode decomposition method.Respectively,to set up samples,to input them to the Elman neural network for training and prediction,the prediction value of each model is obtained,and then the predicted values are summed to get the final forecast results.Simulation results show that the model has high prediction accuracy.%针对太阳黑子时间序列非平稳、混沌及难以预测等特性,建立一种基于经验模态分解(empirical mode decomposition,EMD)与Elman神经网络的组合预测模型.将太阳黑子时间序列通过经验模态分解方法分解为一些不同时间尺度的本征模函数分量和剩余分量,分别对其构建样本,输入到Elman神经网络进行训练和预测,得到各个分量的预测值,并对这些预测值进行求和,得到最终预测结果.对比仿真结果表明,该模型预测精度较高.