Satellite orbit forecasting method and device based on machine learning, medium and product

The invention discloses a satellite orbit forecasting method and device based on machine learning, a medium and a product, and relates to the technical field of satellite orbit forecasting. Orbit errors are learned and trained in advance to obtain a trained orbit error forecasting model; the orbit error, the speed prediction value and the acceleration prediction value corresponding to each historical moment in the historical time period are directly used as input, the orbit error corresponding to the prediction moment is obtained by using the trained orbit error prediction model, and the orbit prediction value corresponding to the prediction moment is obtained by prediction based on the SGP4 analytic method orbit prediction model. The orbit prediction value corresponding to the prediction moment is corrected by using the orbit error corresponding to the prediction moment, so that the advantage of high calculation efficiency of the SGP4 analytic orbit prediction model is reserved, the calculation precision of