用BP神经网络预报太阳活动第23周的黑子数

本文设计、训练和利用BP神经网络，对1750年以来的各太阳活动周上升段和下降段太阳黑子数的变化数据进行了分类和模式识别，得到各太阳活动周上升周期及其上升期间太阳黑子数平滑月均值相当好的模拟结果；在此基础上获得较好的太阳活动第22周上升周期及太阳黑子数的最大平滑月均值预报结果；还作出太阳活动第23周的上升周期及太阳黑子数的最大平滑月均值的预报结果．

APPLICATIONS OF BP NEURAL NETWORKS IN FORECASTING SUNSPOT NUMBERS FOR SOLAR CYCLE 23

BP neuraI networks were designed, thened and used t0 caregorize the data set andrecognize pattems in sunspot number vallation during the ascent and descent of solarcycles- A hist0rical data set of smo0thed monthly mean sunsPOt numbers for l75O ADtO the first half of l995 AD and the predicted values f0r the latter half of l995 tol996 were used. We 0btained good simulation results of the ascending Periods and smoothed monthly mean sunspot numbers for each ascent Period of Solar Cycle l-22and fairly good forecasting results of the ascending Period and the maximum ofsmoothed monthiy mean sunspot number for Solar Cycle 22- Flnally, the paper givesthe following eshmared values for Sol4f Cycle 23: The maximum of smoothedmonthly mean sunspot number is l96 and the ascending Period is 38 months-