摘要
本文采用了1891年至1988年期间的太阳黑子相对数的月均值资料序列,进行了用非线性时间序列分析的跳步门限自回归(LSTAR)模型作预测的研究。三年的预测结果表明,预测值与实际月均值之间的相对误差约为17％,预测值的平滑值与实际月均值的平滑值之间的相对误差接近8％左右。该方法预测的初步尝试结果表明,1992年每月的太阳黑子相对数仍将可能稳定在100以上,第22周太阳黑子相对数的谷值期将可能发生在1996年。
Adopting the monthly mean data of relative sunspot numbers during 1891 through 1988,the prediction study is made by Leap-Step Threshold Auto Regressive (LSTAR) models in non-linear time series analysis. It is shown from predicted results for 3 years that the relative error between the predicted and the practical monthly mean values is about 17%, but that between the smoothed monthly values obtained by moving average of both the predicted and the observed monthly mean values is about 8%. According to the predicted results of preliminary attempts by using this method, the monthly values of relative sunspot numbers in 1992 will basically he more than 100 and the valley values in the 22nd solar cycle might occur in 1996.
出处
《中国科学院上海天文台年刊》
1993年第14期164-169,共6页
Annals Shanghai Astronomical Observatory Chinese Academy of Sciences
基金
国家基础性研究重大关键项目
国家自然科学基金资助研究
