基于随机森林模型的长沙市O3浓度逐小时预测OA

The average O3 mass concentration data and meteorological data of 10 state-con-trolled urban ambient air quality monitoring sites in Changsha City can build a random for-est machine learning model of ozone pollution(spring,summer and autumn),and predict the hourly O3 mass concentration.The results show that:1)The monthly average mass concentration of O3 between 2020 and 2023 is characterized by"M"type distribution with"spring,summer and autumn high and winter low".The high value is mainly concentrated in April to October,and the peak value appears in May and September.Hourly mass con-centration changes of O3 present a single peak distribution,and the daily peak value ap-pears at 15-16.2)For the importance of spring,summer and autumn,the top five are hour>RH>PM2.5>T>NO2,hour factor has the greatest impact on O3,RH,PM2.5,T and NO2 variables are ranked lower with secondary influence.In the analysis of correlation,the positive correlation between O3 and temperature and the negative correlation with rela-tive humidity are the strongest.The positive correlation between O3 and PM2.5 and PM10 in summer is also strong.3)The results of goodness-of-fit analysis of the O3 prediction model based on random forest show that the goodness-of-fit R2 is high,and shows excellent pre-diction performance and can better capture the hourly variation of the O3 concentration.4)By predicting the hourly O3 mass concentration of the spring,summer and autumn in Changsha City in 2024,the results show that the random forest model has a good generali-zation ability.Indicating that the random forest model can more accurately predict the trend of the hourly O3 mass concentration.