Compared with the traditional geostationary earth orbit (GEO) satellite, the new generation of medium-low-orbit satellite Internet constellation represented by SpaceX, Starlink and O3b has significant advantages such as wide-area coverage, full-time interconnection and multi-star coordination, and has become one of the research focuses in the world today. The traditional satellite resource scheduling method mainly studies the resource scheduling problems with single GEO satellite, which is difficult to meet the resource scheduling requirements of the low-orbit satellite constellation characterized by multi-satellite coordination, joint networking, and mass users. Consequently, an intelligent multi-star collaborative resource scheduling model based on user satisfaction is constructed, and a satellite network resource scheduling mechanism named IRSUP based on reinforcement learning is proposed. IRSUP designs an intelligent user service preference optimization module for the personalized needs of user service customization and an intelligent scheduling module based on reinforcement learning for the problems with joint optimization of multi-star resources. The simulation results show that IRSUP can effectively improve the rationality of resource scheduling, link resource utilization and user satisfaction, among which the business capacity is increased by 30%~60%, and user satisfaction is increased by more than twice.