Abstract: Although the Continuous Emissions Monitoring System (CEMS) has been installed for most key sources of air pollutants emission in China, the current assessment of compliance with emission limits of air pollutants is based on manual monitoring data. In order to study and develop emission limit assessment methods based on CEMS data, 68 typical thermal power generation units that meet the validity conditions are selected from 363 units distributed all across China. The 2015 CEMS data of these units are used to analyze the statistical distribution of the hourly, daily average and monthly average SO₂ values. The study establishes five scenarios ranging from strict to lenient, each of which contains three corresponding emission limits, and compares them with those of the EU. The results show that: (1) The determination coefficients of the lognormal distribution of 68 thermal power units all satisfy R²>0.90. (2) In the assessment method of compliance with emission limits based on CEMS data, there is a correlation between the emission limits in different average periods. Finally, the research uses the limit multiple relationship corresponding to the 95% upper prediction as the research result. When 95% of SO₂ hourly values, the daily average values and monthly average values meet the corresponding limits simultaneously in a calendar year. If 90% of the units are required to meet the limits, the hourly limit values can be set to 1.5 times the daily limit values and 1.7 times the monthly limit values. If 95% of the units are required to meet the limit values, the hourly limit values can be set to 1.6 times the daily limit values and 2.0 times the monthly limit values. If 99% of the units are required to meet the limit values, the hourly limit values can be set to 1.6 times the daily limit values and 2.2 times the monthly limit values. (3) The above-mentioned correlation between the emission limits for different average periods is independent of capacity scale of thermal power generation units. The study shows that the CEMS data of 68 units form a log-normal distribution. The assessment method of compliance with emission limits based on CEMS data should be included in the emission standards of the key sources of air pollutants. The method should stipulate hourly values, daily average values and monthly average values at the same time, and clarify their ratio. Thermal power units of various installation scales can be stipulated by the same assessment method of compliance with emission limits.