Abstract: Aiming at the increasingly serious problem of interference caused by the complex topology of ultra-dense networks, the paper proposes an improved interference graph-based clustering interference management algorithm. The cluster weight value was defined based on the interference relationship and the interference graph was constructed. The pico base station was divided into clusters according to the clustering algorithm, and the clustering threshold was set to avoid the vertices being added into a cluster indefinitely. After the clustering was completed, channels were assigned to users according to their data rate values. Combined with the power allocation algorithm, interference in the cluster can be reduced. The simulation results show that the proposed algorithm can reduce the interference between pico base stations compared with the non-clustering algorithms, and the power allocation iterative algorithm assisted by the energy efficiency function can obtain higher system throughput.