In order to reduce the heavy calculation of the existing Model Predictive Control (MPC) for Modular Multilevel Converters (MMC), an optimized control set based model predictive control is proposed. In this paper, the number of inserted sub-module for each arms is seen as the control variables, cost function is established for AC-side current and circulating current. The optimized control sets are designed in consideration of the fluctuations of capacitor voltage. Simultaneously the adjustment factor is settled according to the fluctuation of capacitor’s voltage to adjust the optimal control set boundary, as a result, the accurate number of each arm is ensured in a limited optimal calculation. Compared with the existing predictive control strategy, control computation is reduced to (N+1+Ntotal) from (N+1)². Then circulating current is suppressed and grid current is tracked effectively, heavy calculation of global optimization is avoided at the same time. The optimization efficiency and accuracy are improved effectively. Simulation and experimental results validate the characteristics of low computational complexity, fast dynamic response and high steady accuracy of the proposed optimized control set based MPC strategy. This work is supported by National Natural Science Foundation of China (No.51407114) and Natural Science Foundation of Shanghai (No. 15ZR1418200 and No. 15ZR1418000) and Shanghai Science and Technology Innovation Action Plan (No. 16DZ0503300).