The self-organizing mapping neural network clustering (SOM) was utilized to conduct an objective classification study of large-scale samples of atmospheric circulation in Kunming from 2017 to 2019. Meteorological conditions and circulation characteristics of different weather types were assessed by using ground-based ozone concentration and daily ozone pollution data. Findings indicate that there are 9 atmospheric circulation field types in Kunming, with significant differences in occurrence probability and seasonal variation. Different weather types significantly affect the frequency of ozone pollution days (O₃-8h≥160 µg/m³) in Kunming. Among all ozone pollution days, T2 (high pressure ridge in northwest Yunnan) and T6 (westward-moving typhoon) have the highest frequency, accounting for 30.8% and 38.5% of total ozone pollution days respectively, defined as “pollution weather types”. T4, T7, and T8 were classified as “relatively clean weather types” due to absence of ozone pollution, while T1, T3, T5, and T9 are “relatively pollution weather types”. There are significant differences in circulation patterns between T2 and T6. Under the control of T2 weather, Kunming and even the entire Yunnan region are dominated by the westerly wind in front of the high-pressure ridge. Easterly winds surrounding the typhoon affect T6 weather. However, under both circulation conditions, the Kunming and Yunnan region are under the control of high pressure circulation, easily forming high temperature, low humidity, strong radiation, and other environmental conditions conducive to ozone exceeding standards.