中文摘要：

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SR的确定，尽管FMI（Finnish Meteorological Institute,芬兰气象学院）方法能够比较准确地识别规则日变化SR，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SR；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%.

英文摘要：

The difficulty to scale K-indices is how to identify the solar regular variation （SR） and FMI method is a verified effective method for appropriate elimination of SR. However FMI method is not able to give K-indices in real time because there is always one day′s delay to acquire SR. To solve this problem we propose a new method based on radial basis neural network which is able to give real time K-indices. Firstly the solar regular variations of H element is obtained by the modified FMI method and then radial basis neural network is used to model this time series, and finally according to the model output and the current mean minute value of H element K-indices are scaled in real time. Experiments show that this method can give real-time solar regular variations with a standard error of 3.8598 nT. The comparison between the K-indices scaled by FMI-H method with one day′s delay and the real-time K-indices confirm this method： 69.8% are in agreement,0.77% differ more than one unit.