Coarse and fine granularity combined non-standard time sequence spacecraft parameter anomaly detection method

The problems that spacecraft parameter detection precision and calculation cost are difficult to balance, a time sequence processing method based on a long-short-term memory neural network and the like is difficult to apply to a non-standard time sequence, and a large amount of redundant data can be introduced by using an interpolation method to standardize the time sequence are solved. The invention provides a coarse and fine granularity combined non-standard time sequence spacecraft parameter anomaly detection method. Comprising the following steps: reading a historical telemetry data file to obtain a historical telemetry engineering value sequence of a spacecraft; according to the historical telemetering engineering value sequence, generating a coarse-grained normalized time sequence by adopting an adaptive dynamic time window; performing normalization processing on the normalized time sequence to generate a sub-sequence; training a long-short term memory neural network according to the sub-sequence set; g