Numerical model for dynamic identification of industrial emission sources and prediction of spatial distribution of harmful substances in the atmospheric air

Relevance of the project

The development of megacities, climatic changes, increasing requirements to industrial safety require better control and risk management of atmospheric air pollution. Despite the fact that physical processes in the atmosphere are well enough studied, many industrial enterprises do not have tools for prompt assessment of interrelationships between their pollution sources and the actual environmental situation. It becomes urgent to create a model for dynamic identification of industrial emission sources and prediction of spatial distribution of harmful substances in the atmospheric air.

Project goal

is to create a digital model that is able to integrate into environmental monitoring platforms.

Advantages of the created model:

Use of dynamic parameter optimisation mechanisms

Optimisation is based on reinforcement learning (RL - Reinforcement Learning). The applied machine learning technologies allow to implement autonomous learning, high accuracy of emission sources identification and dynamic forecasting of harmful (polluting) substances spreading.
Availability of an optimisation tool in the dynamics of emission source identification and prediction of its spread

The tool is promising from the point of view of taking into account the factors of weather environment, changes in wind speed and direction, humidity, presence of precipitation.
New model for predicting the spread of emissions of low computational complexity

Practical significance of the project:

It will be possible to automatically analyse the accumulated data on pollutant concentrations at different observation points, meteorological data, geo-information on the location of gas analysers, topology of the enterprise.

The solutions of the digital model will satisfy the requests of enterprise management, environmental protection services of enterprises, specialists of state supervisory authorities, and certain categories of eco-monitoring equipment manufacturers in terms of efficiency and accuracy.

The developed digital model can be integrated into typical environmental monitoring platforms using API software interfaces, which will significantly expand their application in practice.

The operation of the model will allow to significantly reduce the industrial load on the environment, improve the level of environmental safety of production, collect real-time statistical information on the state of the environment, and, by constantly monitoring the environmental condition at the monitoring object, predict environmental risks and prevent adverse events at enterprises.