Data-driven model for predicting cargo tank pressure of an LNG carrier considering environmental effects

The utilization of natural gas is expanding as part of efforts to reduce greenhouse gas (GHG) emissions. Natural gas is typically liquefied at cryogenic temperatures in order to enhance the efficiency of maritime transport. When these cryogenic cargoes are shipped, BOG (Boil-Off Gas) is generated by the external heat and wave-induced ship motion. Proper management of BOG is critical to maintaining the cargo tank pressure within a safe operational range. In the case of LNG (Liquefied Natural Gas) carriers, BOG is used as fuel for main engines and generator engines, with any surplus being burned in the GCU (Gas Combustion Unit) or reliquefied by a reliquefaction system. Accurate prediction of BOG generation and cargo tank pressure is therefore essential for optimizing reliquefaction system operations and voyage planning. Although various experimental and CFD-based studies have been conducted, it remains challenging to capture the complex, irregular characteristics of real marine environments, particularly the effects of ship motion and sloshing. This study presents a framework for developing a data-driven model that predicts cargo tank pressure in LNG carriers. The data-driven model is based on long-term operation data from a 174K-class LNG carrier, enabling consideration of the combined effects of BOG consumption, reliquefaction performance, and marine environmental conditions on cargo tank pressure. The variables related to cargo tank pressure are derived from ship operation, BOG consumption, and marine environmental conditions. Several regression and machine learning algorithms were compared to identify the most effective predictive model. The model's accuracy was verified by comparing predicted values with actual measurements from an LNG carrier that had been in operation for 2 years, and the results confirmed high predictive accuracy. This approach provides a practical framework for data-driven cargo tank pressure prediction and contributes to improving energy efficiency and reducing GHG emissions in LNG carrier operations. • This paper proposes a data-driven approach to predicting the cargo tank pressure of ocean-going LNG carriers. • Variables are found to be relevant input variables for developing cargo tank pressure prediction models. • Various models for predicting cargo tank pressure are developed using MLR, ElasticNet, RFR, XGBoost, DNN and LSTM. • The proposed method is verified using long-term in-service data from an LNG carrier.