论文部分内容阅读
K指数是一种重要的地磁活动指数,标定K指数的难点在于规则日变化S_R的确定,尽管FMI(FinnishMeteorological Institute,芬兰气象学院)方法能够比较准确地识别规则日变化S_R,给出合理的K指数,但是该方法存在一天的延迟,无法实现实时标定.为了解决这一问题,本文提出了一种基于径向基神经网络的K指数实时标定方法:首先用修正后的FMI方法提取H分量的时均值序列,接着以径向基神经网络对该序列进行建模,最后基于神经网络模型实时获取规则日变化,并结合H分量分均值观测数据标定K指数.实验结果表明:该方法能够以3.8598 nT的标准误差实时获取规则日变化S_R;实时标定的K指数与直接用FMI-H方法延迟一天标定的K指数相比,完全吻合的占69.8%,差别大于一个标度的仅占0.77%.
The K index is an important geomagnetic activity index. The difficulty of calibration K index lies in the determination of the daily variation S_R of the rules. Although the FinnishMeteorological Institute (FMI) method can accurately identify the daily variation S_R of the rules, a reasonable K However, this method has a one-day delay and can not achieve real-time calibration.In order to solve this problem, a K-index real-time calibration method based on RBF neural network is proposed in this paper. Firstly, the modified FMI method is used to extract the H component Time series, then the radial basis function neural network is used to model the sequence, and finally the diurnal variation of the rule is obtained based on the neural network model in real time, and the K index is calibrated by means of the H component mean-value observation data.The experimental results show that the method can take 3.8598 The standard deviation of nT was used to obtain the daily diurnal variation S_R. The real-time K-index was 69.8% compared with the K-index delaying the one-day calibration using the FMI-H method, accounting for only 0.77% of the difference.