Based on the minute resolution rainfall data of Beijing basic station from 1987 to 2016, Kruskal-Wallis test was used to statistically analyze extreme precipitation event samples obtained from multi-resolution data such as 1 min, 1 h, 3 h, and 6 h. The differences in rainfall characteristic indicators of three types of extreme precipitation events, including short extreme precipitation(SEP), long extreme precipitation (LEP), and persistent extreme precipitation (PEP) were compared, and the temporal variation patterns of extreme precipitation events using methods such as Sen’s slope, MannKendall test, and ordered clustering were explored. The results show that extreme precipitation in Beijing is mainly caused by LEP events, and the resolution of rainfall data significantly affects the statistical results of precipitation amount and duration. The rainfall data of 1 h resolution can basically reflect the main time characteristics of extreme precipitation events. Differences in data resolution can cause deviations in the identification of rainfall concentration and rainfall concentration periods for extreme precipitation events. Extreme precipitation in Beijing is mostly distributed from June to October, and mainly concentrated in July and August, but there is still a possibility of PEP events occurring in September and October. SEP and LEP events often occur at night, while PEP events are more likely to form rainfall peaks from 7:00 to 12:00, posing higher potential risks to residents’ emergency shelter and waterlogging emergency response. Within the statistical period, the difference in data resolution does not have a significant impact on the prediction results of precipitation amount and frequency trends, but it will increase the uncertainty of rainfall duration prediction and cause a “jump” in the year of sudden changes in rainfall characteristic indicators. The precipitation amount, frequency, duration, and concentration of PEP events are less affected by data resolution.