Research on Traceability Method of Gas Leak for Emergency Auxiliary Support

In the field of environmental emergency, sudden gas leaks cause great harm to the economy and society. In recent years, gas leaks have occurred frequently, and the problem of locating the source of air pollution is mainly divided into two steps: dispersion and traceability. However, the application of mechanism modeling or data-driven modeling methods for dispersion prediction is difficult to meet computational efficiency and prediction accuracy. Conventional source estimation methods also have the disadvantage of local optima in tracing pollution sources. In this paper, we propose a gaussian-machine learning model for gas dispersion prediction using gaussian features as inputs, and the improved particle swarm optimization algorithm that takes into account globally search and optimization. Predicted concentration is utilized as the input of the traceability model. Results demonstrate that this method have better performance on estimating leak location and concentration.