太阳黑子数平滑月均值的混合预测模型

发布时间：2019-08-05 18:29

【摘要】：提出一种太阳黑子数平滑月均值的混合预测模型.通过最大Lyapunov指数得到太阳黑子数平滑月均值时间序列的最大可预测周期,结果表明太阳黑子数平滑月均值序列的最大可预测周期为42个月.太阳黑子数平滑月均值时间序列中包含着线性与非线性的成分,利用自回归滑动平均模型对线性成分进行预测,将太阳黑子数平滑月均值的实际值与自回归滑动平均模型的预测值作差值得到仅含有非线性成分的残差序列,利用具有良好非线性预测能力的回声状态网络预测残差序列,并通过人工蜂群算法来确定回声状态网络预测模型的最佳参数.将自回归滑动平均模型预测值与回声状态网络预测的残差相加,得到太阳黑子数平滑月均值的最终预测值.通过第23太阳活动周的太阳黑子数平滑月均值的预测表明提出的预测模型具有较高的预测精度.同时,对第24太阳活动周的太阳黑子数平滑月均值进行了预测,结果表明第24太阳活动周将在2020年2月结束.
[Abstract]:A hybrid prediction model for smooth monthly mean of sunspot number is proposed. The maximum predictable period of the monthly mean time series of sunspot number smoothing is obtained by the maximum Lyapunov index. The results show that the maximum predictable period of the sunspot number smoothing monthly mean series is 42 months. The monthly mean time series of sunspot number smoothing contains linear and nonlinear components. The autoregression moving average model is used to predict the linear components. The residual sequence with only nonlinear components is obtained by making the difference between the actual value of the monthly mean value of sunspot number smoothing monthly mean and the predicted value of autoregression moving average model, and the residual sequence is predicted by echo state network with good nonlinear prediction ability. The optimal parameters of echo state network prediction model are determined by artificial bee swarm algorithm. The predicted value of the autoregression moving average model is added to the residual error predicted by the echo state network, and the final prediction value of the monthly mean value of the sunspot number is obtained. The prediction of the monthly mean of sunspot number in the 23rd solar active cycle shows that the proposed prediction model has high prediction accuracy. At the same time, the monthly mean number of sunspots in the 24th solar active cycle is predicted, and the results show that the 24th solar active cycle will end in February 2020.
【作者单位】：沈阳工业大学信息科学与工程学院;东北大学计算机科学与工程学院;
【基金】：国家自然科学基金(编号:11273001) 辽宁省博士启动基金(编号:20141070)资助项目
【分类号】：P182.41

【相似文献】