This project aims to efficiently integrate intelligent data mining algorithms with meteorological data using the LSTM model. Focusing on Wuxi, it combines PM2.5 observation data with ECMWF meteorological forecasts to perform extrapolation forecasting of PM2.5 at 6-hour intervals within 72 h. By addressing the limitations of conventional methods in handling nonlinear dynamics and multi-source data, this research seeks to provide a robust framework for enhancing air quality management, supporting policy decisions, and mitigating the impacts of PM2.5 pollution. The findings contribute to the application of deep learning in environmental forecasting, offering a scalable solution for urban areas with similar air quality challenges. The ultimate goal is to enable more effective prevention of air pollution, improve air quality, and enhance Wuxi’s management and control capabilities for PM2.5.