基于RBF人工神经网络的X级以上太阳耀斑预报研究  

A study of the prediction for X-class and above flares based on RBF artificial neural network

作　　者：裴世鑫[1,2] 李仲怡 崔芬萍[2] 朱漪婷 

机构地区：[1]南京信息工程大学数学与统计学院,江苏南京210044 [2]南京信息工程大学物理与光电工程学院,江苏南京210044

出　　处：《西北师范大学学报（自然科学版）》2014年第6期26-32,共7页Journal of Northwest Normal University(Natural Science)

基　　金：江苏省自然科学基金资助项目(BK2011829)

摘　　要：采用第23太阳活动周X级以上耀斑的数据,通过回归分析、Gauss拟合和RBF人工神经网络等方法对X级以上耀斑进行预报研究.结果表明,将黑子群的位置、卡灵顿经度、耀斑爆发时间与黑子群达到最大面积的时间关系、每7d黑子群的最大面积、太阳耀斑流量的积分值、CME速度和F10.7射电流量7个预报因子作为参量对RBF人工神经网络预报模型进行训练,训练后建立的RBF模型的输出结果和训练数据的相关系数高达98%,对耀斑强度的预报结果与观测结果的误差在0.5以内,预报模型符合耀斑短期预报的要求.This paper presents a statistical study on the flares based on the data of X-class and above flares in 23 solar cycles.The methods of regression analysis,Gauss fitting and RBF artificial neural networks are used to study a series of predictors.The statistical results show the correlation coefficient between output results of RBF model,which is established to be trained,and training data can reach 98%,while the error between the predicted results and the observations results is less than 0.5.The RBF artificial neural networks model contains seven parameters, including sunspot group of position, absolute longitude,the most widespread spots in each 7 days,the relationship between the start time of flares and the time of sunspot groups that achieve the most widespread,integral values of solar flares,F10.7 radio flux and the speed of CMEs.These seven parameters are used to train the RBF model.Based on this model,the flares can be forecasted for a short period.

关 键 词：空间天气 太阳耀斑 RBF 人工神经网络 预报 

分 类 号：P353[天文地球—空间物理学]