
Russian scientists trained a neural network to determine the quality of natural gas
Scientists from the Russian State University of Oil and Gas named after I.M. Gubkin and the Institute of Computer Science of the Russian Academy of Sciences created a neural network for quickly analyzing the quality of natural gas. The technology reduces verification time from 20–40 minutes to several seconds, ensuring accuracy that meets regulatory requirements, the university press service reported.
The neural network was trained on data from more than a million gas mixtures. To simplify the analysis, scientists developed a model of pseudogas - a virtual twin of natural gas, which mathematically describes possible mixtures. This made it possible to reduce the number of measured parameters to three key ones: the speed of sound in the gas, its thermal conductivity and the concentration of carbon dioxide.
At the second stage, the researchers created a multi-stage method for assessing the accuracy of neural network analysis. A compact measuring module was also tested, which takes the necessary indicators and transmits them to the computing unit. The module can be installed at metering stations and other gas transportation system facilities.
The technology has been successfully tested at production facilities. Its implementation will make it possible to quickly respond to emergency situations and increase the reliability of gas supply. In the future, the method can be adapted for the analysis of biogas, LNG and gas after hydraulic fracturing in fields.