Heterogeneous cellular networks improve the spectrum efficiency and coverage of wireless communication networks by deploying low power base station(BS)overlapping the conventional macro cell.But due to the disparity between the transmit powers of the macro BS and the low power BS,cell association strategy developed for the conventional homogeneous networks may lead to a highly unbalanced traffic loading with most of the traffic concentrated on the macro BS.In this paper,we propose a load-balance cell association scheme for heterogeneous cellular network aiming to maximize the network capacity.By relaxing the association constraints,we can get the upper bound of optimal solution and convert the primal problem into a convex optimization problem.Furthermore we propose a Lagrange multipliers based distributed algorithm by using Lagrange dual theory to solve the convex optimization,which converges to an optimal solution with a theoretical performance guarantee.With the proposed algorithm,mobile terminals(MTs)need to jointly consider their traffic type,received signal-to-interference-noise-ratios(SINRs)from BSs,and the load of BSs when they choose server BS.Simulation results show that the load balance between macro and pico BS is achieved and network capacity is improved significantly by our proposed cell association algorithm. Heterogeneous cellular networks improve the spectrum efficiency and coverage of wireless communication networks by deploying low power base station (BS) overlapping the conventional macro cell. Due due to the disparity between the transmit powers of the macro BS and the low power BS, cell association strategy developed for the conventional homogeneous networks may lead to a highly unbalanced traffic loading with most of the traffic concentrated on the macro BS. In this paper, we propose a load-balance cell association scheme for heterogeneous cellular network aiming to maximize the network capacity. By relaxing the association constraints, we can get the upper bound of optimal solution and convert the primal problem into a convex optimization problem.Furthermore we propose a Lagrange multipliers based distributed algorithm by using Lagrange dual theory to solve the convex optimization, which converges to an optimal solution with a theoretical performance guarantee.With the proposed algorithm, mo bile terminals (MTs) need to jointly consider their traffic type, received signal-to-interference-noise-ratios (SINRs) from BSs, and the load of BSs when they choose server BS. Simulation results show that the load balance between macro and pico BS is achieved and network capacity is improved significantly by our proposed cell association algorithm.