基于BP神经网络的太阳黑子年均值预测  

Prediction of Annual Sunspot Mean Based on BP Neural Network

作　　者：王敏 杨鲲 孙彦菲 WANG Min;YANG Kun;SUN YanFei(Shandong University of Science and Technology, Qingdao Shandong 266590, China;Tianjin Key Lab of Survey Technology for Water Transport Engineering, Tianjin 300456, China)

机构地区：[1]山东科技大学测绘科学与工程学院,山东青岛266590 [2]天津市水运工程测绘技术企业重点实验室,天津300221

出　　处：《北京测绘》2020年第2期204-207,共4页Beijing Surveying and Mapping

摘　　要：以1700年至2008年间太阳黑子数年均值为研究对象,在数字信号处理及神经网络的基础上,分析归一化自相关函数以及离散傅里叶变换(DFT)后的图像特征,得到可靠的太阳黑子活动周期,以此周期为依据设置参数建立BP(back propagation)神经网络,将已有的随机时间样本进行分组训练和验证预测,拟合神经网络输出预测值与实际值并计算其拟合优度,验证了将BP神经网络应用在太阳黑子活动方面有较高的科学性和可行性。Based on the sunspot number average 1700 to 2008 as the research object,the digital signal processing and neural network,on the basis of analysis of normalized autocorrelation function and discrete Fourier transform(DFT)after image characteristics,get reliable sunspot activity cycle,cycle is according to set parameters to establish the BP(back propagation)neural network,the existing random time grouping sample training and validation prediction,fitting of the neural network output and the actual and estimated values and calculate its goodness of fit,It is proved that BP neural network is scientific and feasible to apply in sunspot activity.

关 键 词：天文学 太阳黑子数 自相关 傅里叶变换 BP神经网络 预测 

分 类 号：P122[天文地球—天体测量] P207[天文地球—天文学]