In response to the challenges posed by mega-constellations in satellite-to-ground mission planning, a method for remote control task scheduling is investigated. Traditional approaches struggle to meet the requirements of large satellite numbers, highly dynamic tasks, and enormous task volumes. To achieve efficient and reliable satellite command scheduling, an improved approach integrating deep feedforward network and greedy algorithm is proposed, which processes remote control tasks derived from tracking plans while considering constraints such as command types and satellite status. Simulation experiments compared with other algorithms demonstrate that the proposed method achieves higher task computation success rate and greater computational efficiency, thus fulfilling the practical engineering needs of large-scale mission planning in mega-constellations.