A significant component in winning the energy transition challenge comes down to managing and reducing the emissions from manufacturing activities of the energy-intensive industries. In most of the facilities in oil & gas production, refining and petrochemical industries, flare is used to burn off waste gases, and typically a manual work process is adopted to ensure complete and smokeless combustion. The work process requires the facility operators to monitor camera feed that shows real-time status of flare operations and make control changes on the DCS to optimize combustion ensuring smokeless flaring.
To automate this manual work process and further enhance operator efficiency and environmental performance in flare operations, we present in this work an end-to-end flare smoke detection, alerting and DCS control solution that leverages existing flare CCTV cameras used at many facilities. The core of this solution is a deep-learning computer vision model that we have developed leveraging an extensive and diverse data set and the model has shown fit-for-use performance in wide-ranging conditions.
We will present results of this application implemented in the plant to further demonstrate the end-to-end solution and its positive impact on sustainable operations.

Facilitator: Adi Punuru, ExxonMobil Technology and Engineering Company

Speaker:
Ye Hu, ExxonMobil Technology and Engineering Company