This paper presents a new indirect finite control set model predictive control (IFCS-MPC) method for modular multilevel converters (MMC) with reduced number of sensors approach. The proposed method employs a submodule (SM) capacitor voltage observer algorithm to estimate the SM voltages of the converter. This observer is based on the discrete-time model of the SM capacitor. Compared with sensor-based methods, this scheme requires only half the number of voltage sensors to achieve the voltage-balancing of the converter. The proposed algorithm also requires the switching patterns of each SM switch, which are provided by the controller used without the need for extra sensors. The substantial reduction in the number of voltage sensors improves the system reliability and decreases its cost and complexity. The simulation results are carried out to verify the effectiveness of proposed control system and the feasibility of reducing the number of voltage sensors.