This paper considers fairness designs under the user-centric framework with heterogeneous receivers comprising access points (APs) and a satellite. We exploit the closed-form ergodic throughput per user to formulate a generic optimization class that addresses the network fairness subject to the association patterns of all the users. Exhibiting the combinatorial structure, the global optimal solution to the association patterns can be obtained by an exhaustive search for small-scale networks with a small number of APs and users. For large-scale networks, we design two low computational complexity algorithms based on evolutionary computation to obtain a good solution in polynomial time. Specifically, we adapt the genetic algorithm (GA) to handle the discrete feasible region and the fairness metrics. Numerical results demonstrate that the optimized association patterns significantly improve the per-user throughput. The proposed GA-based algorithms yield the global optimum for small-scale networks coincided with an exhaustive search. Besides, the GA-based algorithms unveil practical association patterns for large-scale networks.