Experimental studies of the combustion of pulverized coal to form the basis for training a neural network

Abstract

Abstract In this paper, we considered the solution to the problem of efficient and environmentally friendly combustion of hydrocarbon-containing fuels and the prevention of emergencies in combustion chambers and boiler plants. On the basis of the existing scientific background and the existing instrumental and experimental base, modern technologies for automating the combustion process using machine learning methods have been developed. Automation will reduce energy costs and prevent possible emergencies at thermal power plants and combustion chambers. The approbation of the technology and the formation of a data corpus for training the network was carried out using semi-industrial fire installations with a thermal power of about 5 MW, which make it possible to vary the parameters of fuel combustion in a wide range and, accordingly, to implement various combustion modes.