In order to meet the requirements of multi-satellite parallel telemetry data processing and high-concurrency analysis for massive data, facilitateing data mining and intelligent warning, this paper proposes and implements a real-time compression algorithm for satellite telemetry data. Based on the characteristics of telemetry data, a telemetry adaptive classification method is proposed, using an improved RLE compression and incremental compression algorithms combined with database technology to achieve compression of telemetry data. In the development project of a certain type of satellite, this algorithm was used for data compression, and statistical analysis shows that this algorithm has achieved good compression results.