Satellite telemetry, tracking, and command (TT &C) operations are critical to ensuring the normal operation of mega satellite networks. However, the distribution and number of ground stations are limited, making that the existing TT &C mission scheduling methods are difficult to satisfy the TT &C requirements in mega satellite networks, resulting in low TT &C mission completion rates. In this paper, we first construct a space-ground integrated distributed TT &C mission scheduling frame-work utilizing the broad coverage characteristics of geostationary earth orbit (GEO) satellites. Then, we explore the similarity in the TT &C mission scheduling process among adjacent ground stations or GEO satellites, that the TT &C missions execute within the visible time window between satellites and TT &C antennas. Building on this, we share similar features of mission scheduling between the stations through federated learning (FL) and capture the temporal features of TT &C mission scheduling using deep reinforcement learning (DRL) at each station. Therefore, we propose a federated deep reinforcement learning (FDRL) assisting TT &C mission scheduling algorithm in mega satellite networks to enhance TT &C mission completion rates. Finally, the effectiveness of the FDRL algorithm is verified through simulation experiments. Compare to the traditional space-ground integrated algorithm, the FDRL algorithm improves the TT&C mission completion rate by about 26.5 %.