Instruments and Systems: Monitoring, Control, and Diagnostics Annotation << Back Intelligent Methods for Automated Control and Predictive Management of Flare Emissions at a Primary Oil Refining unit based on a Digital Twin and a Machine Vision System K.D. SKOBELEV, V.G. BLAGOVESHCHENSKY This article examines the problem of intelligent automation of flare emission monitoring at a primary oil refining unit based on the combined use of a machine vision system and a digital twin of the process facility. The relevance of the study stems from the need to improve the environmental and process safety of oil refi ning facilities, as well as the limitations of traditional control methods based primarily on operator observation and analysis of individual process parameters. In this paper, flare emissions are considered as a complex, multiparameter automated control object whose condition is determined by both the internal characteristics of the process mode and external visual indicators of flare combustion. A methodological approach is proposed that includes a machine vision subsystem, a module for integrating visual and process data, a digital twin of the primary oil refining unit, and a predictive condition assessment system. The structure of input parameters, visual indicators, and diagnostic assessments used for intelligent analysis of the fl are system's condition is formalized. It has been demonstrated that the developed convolutional neural network model enables the identification of flare zones in images of a real industrial facility and the use of the recognition results within an automated control system. The practical significance of this study lies in creating a foundation for earlier detection of unfavorable conditions, improving the efficiency of control, and laying the foundation for predictive flare management in oil refining operations. Keywords: intelligent control, flare emissions, machine vision, digital twin, predictive control. DOI: 10.25791/pribor.6.2026.1679 Pp. 07-14.