Modular multilevel converter (MMC) with reduced number of voltage sensors is conducive to the reduction of equipment cost and hardware complexity, however, usually resulting in the performance degradation. For example, when only one voltage sensor is equipped for the measurement of arm voltage, estimation of capacitor voltages may be not accurate enough, moreover, condition monitoring and fault diagnosis of submodules (SMs) are hard to implement. Thus, capacitor voltage estimation with capacitance self-updating based on grouping measurement (GM) is proposed in this article. By grouping and updating SM capacitance periodically, the accuracy of voltage estimation is greatly increased, meanwhile, severely aging or faulty capacitors can be detected, convenient for prompt replacement. Furthermore, GM-based fault diagnosis and fault disposal are proposed. Utilizing the measured and estimated group voltages, single or multiple SM open-circuit failures can be located in several sampling cycles and the isolating time of faulty SMs can be easily obtained even considering the time delay of bypass switch. Therefore, capacitor overvoltage and current distortion during SM failures are restrained effectively, which enhances the reliability and achieves seamless operation of MMC. Finally, the effectiveness of the proposed scheme is confirmed by simulation and experimental results.