The invention discloses a self-optimization method for interference management based on non-orthogonal multiple access, which belongs to the field of wireless networks. Specifically, firstly, in an ultra-dense scene, each small base station calculates and requests each small base station separately for a fixed sub-bandwidth. The signal-to-noise ratio between users, according to the perceived interference signal strength of adjacent base stations, deletes the adjacent interfering base stations of the base station, and independently constructs a dynamic local interference map; then, considering the user's communication needs and local interference map information comprehensively, Each small base station completes resource allocation independently. In order to reduce inter-network interference, the interference optimization problem is modeled as a game problem based on satisfaction, and a distributed algorithm is used to obtain the resource allocation result. Each small base station reasonably allocates different resources for each requesting user. The power and sub-bandwidth can optimize the system interference. The invention effectively reduces the network interference, ensures the user's communication requirements, and has good convergence.