Development the reinforcement learning model for sources identification of H2S industrial emissions

P. 987–991.

Kychkin A., Chernitsin I., Vikentyeva O.

Industry 4.0 concept focuses on sustainability problem that requires to control air emissions, especially for harmful substances like H2S, and reduction their impact on nature by using environmental monitoring and sources identification systems. This task requires solving inverse problem of dispersion models, which should establish complex mathematical dependences between the sensor data, the location and emission rate on a chimney in dynamics. In our research we propose the reinforcement learning (RL) model for H2S sources identification on given example of real-life data. The search algorithm is based on the Q-Learning that uses dispersion simulations on industrial emissions within different rates, meteorological data and landscape specifics for training. Python library and visualization tool as the industrial software application has been developed, with the help of which it is possible to analyze the possible H2S sources on large industries.

Keywords: forecasting, system architecture, Internet of Things, Platform, Machine Learning, industrial emissions

In book

2025 International Conference on Industrial Engineering, Applications and Manufacturing (ICIEAM)

IEEE, 2025.