Virtual Booth

This project introduces a novel approach for predicting PM2.5 dispersion in Bangkok by combining the one dimensional Advection Diffusion Equation with Physics  Informed Neural Networks (PINNs). By leveraging 366557 real world data points including PM2.5 levels, geographic coordinates, and time, the model integrates physical laws into the learning process to improve accuracy. The equation models both advection and diffusion of particles in the atmosphere. Evaluation results demonstrate strong performance, with a Data Loss of 0.0063045 and a Physics Loss of 0.0000107. While initial outcomes are promising, further development is necessary to support practical applications in real-time air quality monitoring systems.