Research on air quality prediction based on ResNet-LSTM

Based on the air quality monitoring data and meteorological data in Hanzhong from 2015 to 2024, the temporal variation characteristics of air quality index (AQI) in Hanzhong were analyzed, and the air quality prediction model based on RESNET LSTM algorithm was constructed. The evaluation indexes were compared with CNN, LSTM, CNN-LSTM, RandomForest and XGBoost algorithm models. The results show that the overall air quality in Hanzhong is improving. The monthly variation of AQI showed a "U" shape. September and October were the best months of the year, and January and February were the months with poor air quality; O3 was the main pollutant in summer, and PM2.5 and PM10 were the main pollutants in the rest of the time; AQI was positively correlated with temperature and negatively correlated with precipitation; AQI was negatively correlated with air pressure and wind speed in winter and spring, and positively correlated with wind speed in summer; The relationship between AQI and humidity, water vapor pressure is complex, and there are seasonal differences; The fitting effect of the established RESNET LSTM model for AQI prediction is better than that of the five comparison models, and its root mean square error, average absolute percentage error and average absolute error are reduced by 4.11%, 5.23% and 5.62% compared with the suboptimal model, and the determination coefficient is increased to 0.78. The accuracy rate of air quality grade prediction is 70.82%, which is higher than CNN, LSTM, CNN-LSTM, RandomForest and XGBoost algorithm models, and has certain short-term AQI mutation trend capture ability.