论文部分内容阅读
为了更加准确地预报磁暴(Kp>5)的发生,充分利用ACE卫星积累的上游行星际条件的数据,以开磁通生成速率函数dΦMP/dt和太阳风磁层粘滞作用项n1/2v2为主要输入参数,应用神经网络方法,构建了三个模型,预报三小时时段的Kp值。根据实际需要,这三个模型采用了不同的训练集构造方法和提前时间量。模型1输入当前的开磁通生成率,粘滞作用项,太阳风速度、密度,和行星际磁场总强度、By分量、Bz分量,提前1~3.5h预报Kp;模型2在模型1的基础上加入Kp现报,提前1~3.5h预报Kp;模型3输入9小时延迟的开磁通生成率和粘滞作用项,当前的太阳风速度、密度,行星际磁场总强度、By分量、Bz分量,提前3小时预报Kp。对1998年、2002年和2006年的测试结果表明:三个模型的预测值与实测值之间的相关系数分别为0.88、0.90、0.85,预测的均方根误差分别为0.65、0.62、0.72。
In order to forecast the occurrence of magnetic storms (Kp> 5) more accurately and make full use of the data of upstream interplanetary conditions accumulated by ACE satellite, the open flux-generating rate function dΦMP / dt and the solar wind viscous action term n1 / 2v2 are the main Input parameters, the application of neural network method, the construction of the three models, three hours forecast Kp value. According to the actual needs, these three models adopt different training set construction methods and the amount of advance time. Model 1 inputs current open flux generation rate, viscous action term, solar wind velocity, density, and total interplanetary magnetic field strength, By component and Bz component, and predicts Kp from 1 to 3.5 h in advance. Model 2 is based on Model 1 Kp is added and Kp is predicted 1 ~ 3.5h ahead of time. In Model 3, 9-hour delayed open magnetic flux generation rate and viscous action term, current solar wind velocity, density, total interplanetary magnetic field strength, By component, Bz component, 3 hours in advance forecast Kp. The test results of 1998, 2002 and 2006 show that the correlation coefficients between predicted and measured values of the three models are 0.88, 0.90 and 0.85 respectively, and the root mean square errors of prediction are 0.65, 0.62 and 0.72 respectively.