Abstract
In order to provide global coverage, high-speed, low-cost networks, a large number of LEO satellites and a small number of GEO/MEO satellites are used to build a multi-layer satellite network. It is gradually accepted by people. However, this causes the structure of the satellite network to be complicated. Therefore, people propose to simplify the network with clustering LEO satellites, but existing researches mostly adopts simplified methods for how to clustering, which decreases the satellite network's overall performance. In this paper, we have studied the way of clustering LEO satellites based on services, introduced service forecasts, and developed an efficient algorithm that directly calculates the cluster solution of the LEO satellite by the forecast services. Finally, we conducted a wide range of simulation experiments, and the results showed that the cluster solution we proposed was better than the baseline solution in terms of service throughput at least.