The prediction method based on time series has broad application prospects in the field of spacecraft telemetry data prediction. One obvious characteristic of time series is memory, and memory is the performance of any observed value in the time series. The basic idea of this model is to establish the statistical model of the data according to the characteristics of the observation data, and to use the statistical characteristics of the model to explain the statistics of the data, so as to achieve the purpose of forecasting. The algorithm using pattern recognition and parameter estimation method, and combined with the spacecraft telemetry dynamic data, the establishment of spacecraft telemetry data time series prediction model, to detect and forecast the trend of the spacecraft telemetry data.