The pros and cons of load balancing among multi cells has a big impact on the network performance. Due to the trade-off of the existing methods, they are difficult to ensure the key performance indicators (e.g. the call blocking rate) to obtain best performance in the network. In order to address the issue, Load balancing is formulated as a multi-objective optimization, of which the objective function for Quality of Service (QoS) requirement is a joint optimization function of the load balancing index and average load function of the network, and the objective function for Best Effort (BE) user is the total utility function of all the BE usersthroughput, taking the available resources and users QoS requests as the constraints. Additionally, a distributed load balancing algorithm is proposed in view of the computational complexity in a practically system operation, which includes the resource scheduling policy, user switching conditions and call access control. The simulation results show that the proposed method has achieved a better load balancing index, thus effectively reducing the new call blocking rate by the QoS users and improving the network resources utilization.