A bayesian decision-theory-based digital twin for methane flares

Ground flares operate in a high-turndown, standby configuration for a significant portion of their operating life, being fully utilized only under process upset scenarios or emergencies. The low-momentum flow results in poor fuel-air mixing near the flare tip, leading to decreased overall combustion efficiency (CE) and increased emissions of unburnt volatile organic compounds (VOC). In such scenarios, assist medium flow rates that are too low, a state called under-assist, results in excessive visible smoke and particulate pollution. On the other hand, over-assisting degrades CE due to premature quenching of the reaction zones. Prevailing wind conditions directly impact the CE of a flare by promoting mixing to varying degrees. Fluctuating wind gusts also generate turbulence near the flare tip, occasionally inducing fuel stripping from the reaction zones, or creating intermittent puffs of flame due to localized change in the equivalence ratio. Therefore, achieving continuous smokeless flaring while maintaining high overall combustion efficiency (CEoverall) requires active control of the fuel and assist streams based on the local environmental conditions.