基于符号化表示相似度度量的卫星多元工程参数异常检测

随着卫星系统复杂程度的日益增加，综合分析卫星多元参数之间的相关性异常对于卫星安全运行和空间任务的正确执行具有重要意义。利用某卫星工程参数数据，基于符号聚合近似算法（SAX），研究卫星多元工程参数的异常检测问题，解决了当前异常检测方法中多元参数融合时不考虑上下文信息造成信息丢失的问题，实现多元参数有效融合，形成一种优化的基于快速动态时间规整算法（Fast-DTW）的异常检测算法。研究结果表明，模型在某卫星电源子系统的异常检测过程中，recall,precision和F1 score分别为0.947,0.9和0.923，能够实际应用于卫星异常检测，提高卫星在轨运行的安全性。

Multivariate Engineering Parameter Anomaly Detection of the Satellite Based on Similarity Metric of the Symbolic Representation

The complexity of the satellite system is increasing, and the comprehensive analysis of the abnormal correlation between the satellite with multiple parameters is important for the safe operation of the satellite and the correct execution of space tasks. Aiming at the characteristics of large amount of engineering parameter data, high parameter correlation and pseudo-period, a satellite multivariate engineering parameter anomaly detection method based on symbolized representation similarity measurement is proposed. Using real satellite engineering parameter data, based on Symbol Aggregation Algorithm (SAX), this paper studies the anomaly detection problem of satellite multiple engineering parameters, solves the problem of the context information, realizes effective fusion, and forms an optimized anomaly detection algorithm based on similarity measurement Fast-DTW algorithm. Results show that during the abnormality detection process, the recall, precision and F1 score are 0.947, 0.9 and 0.923 respectively in the real satellite power subsystem, and the algorithm can be actually used in satellite anomaly detection to improve the safety of satellite in-orbit operation.