基于EMD和组合模型的太阳黑子时间序列预测

Sunspots Time-series Prediction Based on EMD and Combination Model

太阳黑子是非线性、非平稳、多尺度变化的时间序列,且观测结果大多存在噪声的干扰.针对太阳黑子时间序列预测的复杂性,首先将原始数据序列通过小波去噪进行预处理,然后将去噪后的信号通过EMD分解产生若干个从高频到低频的IMF分量和余项.针对低频分量变化缓慢和高频分量波动性较大的特点,分别采用RBF神经网络模型和SVM模型进行预测,最后将各个分量的预测结果相叠加得到最终预测值.仿真结果表明,该模型具有较高的预测精度.

Sunspots are non-linear, non-stationary, multi-scale changes time-series, and the observations were often interfered by noise. According to the complexity of sunspots time-series prediction, first of all, this paper preprocessed the original data through wavelet de-noising method, then the denoised signal was decomposed into several IMF components and remainder by EMD. In view of the characteristics of the low frequency and high frequency components, RBF neural network model and SVM model were used to predict them respectively, the final predicted value would be got by adding each component＇ s result at last. The simulation results show that this model has higher prediction accuracy.