As the core component of the flexible direct current transmission system, the safe and reliable operation of the MMC power sub-module is of vital importance. However, due to differences in the put-in frequency, the reliability of the MMC power sub-module is decentralized. This is compounded by the adoption of an offline random fixed-proportion spot-check operation and maintenance method with a fixed annual cycle. This small and random sampling method with incomplete coverage puts the system at risk of shutdown. To address this issue, this paper proposes a non-intrusive method for abnormal monitoring and online status evaluation of MMC power sub-modules, based on the statistical laws of the output switch timing signals of the inherent control algorithm of the MMC system. The proposed method involves extracting the put-in frequencies of each bridge arm sub-module in the MMC system and analyzing the optimal statistical distribution law. This is followed by establishing the correlation between the input frequency of the MMC power sub-module and the capacitor capacitance, allowing for online estimation of the capacitance value of abnormal MMC power sub-modules. Finally, the distribution law of sub-modules in the MMC system is extracted and the effectiveness of the proposed method is verified through simulation of aging states of power sub-modules with different serial numbers. The results of the simulation show that, based on the statistical distribution law of the input frequency of the power sub-module, the proposed method can accurately screen abnormal sub-modules and estimate the capacitance value of the film capacitor causing its aging. This is beneficial in improving the operational reliability of the MMC system and reducing operation and maintenance costs.