对未来两个太阳周太阳活动参数的统计预测

The Statistical Prediction of Sunspot Number for the Next Two Solar Cycles

作者：

丁煌中国电力科学研究院新能源研究中心南京 210009
廖云琛南京信息工程大学南京 210044
肖子牛中国科学院大气物理研究所大气科学和地球流体力学数值模拟国家重点实验室北京 100029

中文摘要：

利用支持向量机（Support Vector Machine，SVM）和后向传播（Back Propagation，BP）神经网络的方法，结合前23个太阳周期（1700—2008年）的周期特征数据，对第24和第25个太阳周的各个周期特征进行了预测，并且通过交叉验证算法得出两种方法都可达到最优。通过分析SVM与BP神经网络方法的预测结果，均表明第25个太阳周将会达到较强的强度，且大于第24个太阳周。此外，两种方法都预测出第25个太阳周的太阳黑子数在谷值年维持异常偏低，周期长度都会维持在10年左右。根据第24个太阳周已经过去的特征验证，BP神经网络的结果与实际情况更为接近，预测太阳活动在2020开始进入第25个太阳周，在2025年达到峰值，峰值年强度比第24个太阳周偏强。

中文关键词：

太阳黑子数，太阳周期，支持向量机，后向传播神经网络

KeyWords：

sun spot number, solar cycle, support vector machine, back propagation neural network

Abstract：

In this work, the 24th and 25th solar cycles’ periodic characteristics were predicted by using support vector machine(SVM) method and back propagation (BP) neural network method respectively, based upon the data of previous 23 solar cycles.The authors found that both of the methods reach the optimal value after applying the cross validation algorithm, it reveals that the intensity of solar activity in the 25th solar cycle will be enhanced comparing with the 24th solar cycle in both the SVM and the BP Neural Network prediction; that the sun spot number (SSN) will maintain low value in the valley phase; and the cycle length will be around 10 years in the 25th solar cycle. According to the periodic characteristics of the 24th solar cycle that was past, there sults from BP Neural Network method prediction are more close to actual situation than that from the SVM. The results indicate that solar activity will enter 25th solar cycle in 2020, and will reach the peak in 2025; the amplitude of peak year in the 25th solar cycle will be stronger than that in the 24th solar cycle.