摘要
针对太阳黑子的复杂性,利用经验模态分解(EMD)方法,将太阳活动在各时间尺度上的变化分量分解为平稳的固有模态函数(IMF)分量及余项。观察各分量的频谱,根据低频IMF分量和高频IMF分量的特点,分别采用自回归滑动平均模型和神经网络方法进行预测。通过各分量的预测值,重构出原始信号的预测序列。仿真结果表明,该模型具有较高的预测精度。
According to the complexity of sunspots, this paper uses Empirical Mode Decomposition(EMD) method, the solar activity contains all of the time scale changes separated into the inherent weight smooth Intrinsic Mode Function(IMF) and remainders. It observes each component of the spectrum, based on the characteristics of the low frequency IMF component selection Auto-regressive Moving Average(ARMA) model predicted the average directly, and the high frequency IMF using neural network forecast. Through the various components of the primary signal reconstruction predicts a prediction sequence, and increases the prediction accuracy. Simulation results show that the model has higher forecast accuracy.
出处
《计算机工程》
CAS
CSCD
北大核心
2011年第24期176-179,共4页
Computer Engineering
基金
云南省自然科学基金资助项目(2009CD028)
昆明理工大学科学研究基金资助项目(201001)
关键词
太阳黑子数
经验模态分解方法
自回归滑动平均模型
反向传播
sunspot number
Empirical Mode Decomposition(EMD) method
Auto-regressive Moving Average(ARMA) model
BackPropagation(BP)