To deal with issue of high hardware cost and energy consumption, beam selection is used in the lens antenna array based millimeter-Wave (mmWave) massive Multiple-Input Multiple-Output (MIMO) system to reduce the required number of radio frequency chains. However, beam selection requires the base station to acquire the accurate channel state information. To solve this problem, we first use the structural characteristics of beamspace and adopt a Support Detection (SD)-based channel estimation scheme to estimate the large channel with lower pilot overhead and computational complexity. Then, inspired by Ant Colony Optimization (ACO) algorithm in bionics, an ACO beam selection scheme based on SD channel estimation is proposed. The scheme can not only avoid the interference of beams between users, but also maximize the system sum rate. Finally, the simulation results show that the proposed scheme can obtain a near-optimal solution with significantly lower computational complexity, and it is superior to the existing schemes in improving the system sum rate.