Satellite telemetry time series clustering with improved key points series segmentation

Satellite telemetry data is the only basis for the experts to obtain the working status and the health status of the in-orbit satellite. The pattern mining and extraction of satellite telemetry data are of high significance for automatic judgment and anomaly detection. Clustering, as an important time series data mining method, can achieve automatic and intelligent analysis of satellite telemetry data for pattern discovery. Thus, this paper is devoted to research for time series clustering analysis on satellite telemetry data. Due to the large amount of raw data and pseudo-period characteristic, directly clustering on raw data may be inefficient and susceptible to noise interference. Therefore, a Special Points Series Segmentation method is proposed to extract special point series. This method significantly decreases computational time and reduces the influence of noise. Then, this paper presents a satellite telemetry time series clustering method with Special Points Series Segmentation, which is effective for time series datasets with prominent shape features. Experiments on the open dataset which is similar to satellite telemetry time series prove the superiority and effectiveness of the algorithm.