Abstract
Liquified Natural Gas (LNG) Boil-off Gas (BOG) is the vaporized or displaced gas from LNG
liquid due to increasing temperature or decreasing pressure in LNG containment conditions. The
volume or energy of BOG is typically estimated by calculating the difference in LNG volume or
energy before and after custody transfer during loading or unloading operations. Current
practices involve reporting BOG quantities at the end of operations, which often leads to
inefficiencies and delays. This study proposes a predictive machine learning model trained on key
parameters influencing LNG vaporization. The model enables operators to predict BOG quantities
in real-time, facilitating the preparation of necessary systems to collect BOG without halting
operations or causing environmental contamination. The model’s performance was benchmarked
against a ridge regressor, demonstrating significant improvements in prediction accuracy. This
advancement supports the goal of zero-emission operations and enhances the efficiency of LNG
handling processes, thereby reducing the environmental impact of LNG operations. The findings
of this study are highly relevant to the petroleum industry, as they provide a practical solution to
a common operational challenge, promoting sustainability and operational efficiency in LNG
terminals.