Considering that the conventional associations may not be appropriate due to the special characteristics of heterogeneous cellular networks (HCNs), some novel associations that can balance the network loads should be designed. To reflect the resource consumption levels of users and balance loads, a QoS-aware user association is proposed for load balancing in HCNs. In the association, the load is defined as the amount of consumed resources, which can reflect the resource consumption levels of users. Finally, this scheme is formulated as a network-wide weighted utility maximization problem. That is a nonlinear mixed-integer optimization problem, and obtaining its optimal solutions is very challenging (especially for large-scale problems). To solve the problem, a distributed algorithm is designed using dual decomposition. Numerical results show that, compared with the best power association and range expansion association, the proposed scheme has higher load balancing levels and lower call blocking probabilities.