Abstract: To address the challenges of inefficient access point (AP) resource allocation and high backhaul load in cell-free massive multi-input multiple-output (MIMO) non-orthogonal multiple access (NOMA) systems, a joint optimization algorithm of user clustering and AP selection based on binary particle swarm optimization (BPSO) is proposed. The algorithm aims at maximizing the average downlink rate of users by jointly modeling user clustering and AP selection as a binary combinatorial constraint optimization model. An improved BPSO method is employed to achieve efficient solution and effectively coordinate the coupling relationship between clustering and AP selection. Through theoretical analysis and simulation verification, the average rate of system users and AP occupancy rate under different AP selection algorithms are compared. The results show that the proposed algorithm is superior to the existing schemes in terms of the average downlink rate of users, edge user performance and convergence speed. Under imperfect successive interference cancellation with 60 APs, the average downlink rate of users reaches 11.2 Mbit/s, which is significantly higher than the comparison algorithms such as quantum bacterial foraging optimization and deep reinforcement learning. The average AP occupancy rate is reduced by approximately 52.1% compared to full connection schemes, effectively alleviating backhaul link pressure. The low-speed mobile scene experiment further verifies that the algorithm has good dynamic robustness and provides a feasible solution for the actual deployment of the system.