一种城市百米级PM2.5预测方法、系统、介质及产品
By using a multi-level autoencoder residual neural network and multi-source satellite AOD fusion technology, the problems of spatiotemporal resolution and accuracy in urban PM2.5 prediction have been solved, achieving high spatiotemporal resolution mapping of urban PM2.5 concentration distribution at the 100-meter level and accurately capturing atmospheric pollution characteristics.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CENT SOUTH UNIV
- Filing Date
- 2026-05-13
- Publication Date
- 2026-07-17
AI Technical Summary
Existing technologies cannot achieve PM2.5 prediction with a spatiotemporal resolution of hundreds of meters at the urban scale, cannot accurately capture the spatial differentiation characteristics of air pollution at the urban microscale, and have problems with the difference in accuracy weights in multi-source data fusion, leading to systematic bias.
By employing a multi-level autoencoded residual neural network, and through multi-source satellite AOD fusion, seamless downscaling reconstruction of national AOD, macroscopic inversion and microscopic correction of urban PM2.5, combined with ground-based real observation data, a high spatiotemporal resolution map of urban PM2.5 concentration distribution at the 100-meter level is achieved.
It achieves high spatiotemporal resolution dynamic inversion of PM2.5 at the street level within the city, accurately characterizes the spatial heterogeneity and sudden dynamic features of air pollution, eliminates systematic bias in multi-source data fusion, and provides high-quality PM2.5 concentration distribution products.
Smart Images
Figure CN122174183B_ABST