基于奇异谱分析的空间环境数据插补方法

空间环境数据具有典型的非线性、非平稳特征,并经常包含有缺失数据,给预报模型的建立、预测以及物理过程的分析带来了一定的困难。为了实现对缺失数据的插补,基于奇异谱分析(SSA)迭代插补的思想,设计了一种能够适用于不同缺失数据分布的插补方案。该方案提取出原始时间序列中缺失数据分布数组,利用缺失数据分布数组生成交叉验证所用的测试数据集,并利用离散粒子群优化算法寻找SSA的2个关键性参数,即嵌入窗口长度和主成分个数。通过不同太阳活动年份实际观测的太阳风参数、地磁指数等实例验证了算法的有效性。

Gap filling method for space environment data based on singular spectrum analysis

The space environment data is known to be nonlinear and non-stationary and often contains missing values, which brings great challenge to the model-building procedures, predictions and posterior analysis. To fill the data gaps, a new gap filling method based on the iterative singular spectrum analysis (SSA) algorithm was put forward. The new method considered the distribution of missing values by extracting a distribution array first and used the array to generate the test data set. The discrete particle swarm optimization algorithm was adapted to obtain the two key parameters of SSA, i.e. the embedded window size and the number of principal components. Taking the solar wind parameters and geomagnetic indices of different solar activity years as examples, the test results demonstrate that the filling method is effective.