一种基于角度偏离的卫星分系统异常检测方法

摘要

In order to ensure the stable operation of a satellite and prolong its life,an anomaly detection method based on angle deviation(ADMAD) is proposed.In the high-dimensional data space of the satellite telemetry data,the method applies the shared nearest neighbors (SNN) algorithm to construct the reference point sets.Then the method selects the feature attributes associated with the anomaly by applying a method based on angle deviation using angle replacing distance.Finally,the normalized Mahalanobis distance is used to calculate the anomaly scores of points.Combining with the statistical knowledge,the threshold based on the anomaly scores is obtained,and the data sets are classified.We verified the proposed method using the telemetry data in control and power subsystem of a satellite in July to September,2014 and July to September,2015 respectively.The experimental results indicate that the accuracy of the proposed algorithm could reach more than 95％ under the condition of lack of the field knowledge.The robustness of the proposed algorithm is higher.Simultaneously,it can detect the anomaly of satellite subsystem timely and effectively.%为保证卫星稳定运行,延长卫星寿命,提出一种基于角度偏离的卫星异常检测算法(ADMAD).针对卫星遥测数据构成的高维数据空间,利用共享近邻(SNN)算法构建相关数据集空间,用角度代替距离,采用基于角度偏离的属性选择算法筛选与异常相关的属性,使用归一化的马氏距离计算异常值,结合统计学知识计算得到异常阈值并对数据集进行分类.采用某卫星2014年7-9月、2015年7-9月控制和电源分系统的遥测数据分别进行验证,试验结果表明,在领域知识匮乏的情况下,该算法的准确率可以达到95％以上,算法鲁棒性较高,能够有效地实时检测卫星分系统异常.