A spaceflight abnormal event detection method based on a generative adversarial network

The invention relates to a spaceflight abnormal event detection method based on a generative adversarial network. The method comprises the steps of utilizing the positive generative adversarial network simulation to generate a normal event sample, and utilizing the reverse generative adversarial network simulation to generate an abnormal event sample; calculating an input event, a simulated normalevent generated by the positive generator and a reconstruction error of a simulated abnormal event generated by the reverse generator by using a detector through a reasonable design algorithm, solving an abnormal score, and judging that the abnormal event is detected when the abnormal score is greater than a threshold value according to the size of the abnormal score; realizing the unsupervised abnormal event detection for spaceflight abnormal events which are rare, complex and changeable in sample and cannot be accurately observed, and carrying out the spaceflight normalized accident prevention and abnormal event ea