A Satellite Telemetry Data Modeling Method Based On Pattern Evolution

Recently,the main issue of satellite telemetry modeling applies to the problem of short-term evolution description and state estimation for some typical telemetry parameters.Due to big size and complex variation of the long-term satellite telemetry data,it is difficult to directly apply the current methods on long-term evolution characteristics modeling.Moreover,since the downstream speed of satellite telemetry data is quite high,collecting large amounts of data in a few days is challenging.Thus,before modeling the telemetry data,the sampling technique should be used firstly to compress the amount of data.In this paper,a satellite telemetry data modeling method based on pattern evolution is proposed,aiming at overcome the shortcomings of the traditional methods.Through analyzing the large amounts of long-term satellite telemetry data,we establish the models to describe the relationship between telemetry parameters and in-orbit running time.The synthetic application of long-term telemetry data segmentation,time series clustering and classification,multiple-sequence based pattern modeling,and pattern evolution modeling based on the stochastic processes solves the problem of long-term satellite telemetry data modeling.Some useful methods are utilized in this paper to conduct the data processing,such as the outlier removal method based on median filtering,the telemetry data clustering method based on Pearson coefficient and system clustering,the pattern classification method based on series’ shape-features and SVM,the formal modeling method of typical pattern based on piecewise approximation,and the telemetry data pattern evolution method based on Markov process.Based on the models established in this paper,we also develop a virtual telemetry data generating method,which is valuable in satellite ground test and simulation.A case study demonstrates the effectiveness and validation of the method in modeling the long-term satellite telemetry data.