Modeling and simulation of thermodynamic behavior of an LNG fuel tank by using artificial neural network based on operational data

As marine environmental regulations have been strengthened, the use of LNG (Liquefied Natural Gas) as a marine fuel is increasing. In an LNG tank under cryogenic condition, boil-off gas is generated due to heat ingress, and the pressure and temperature of the tank continuously change. For safe and efficient operation of LNG fueled ships, it is necessary to accurately predict the thermodynamic state of an LNG fuel tank. In this paper, a difference estimating artificial neural network (D-ANN) model is proposed to evaluate the thermodynamic behavior in an IMO type C fuel tank. The D-ANN model predicts the pressure, temperature and filling ratio of the LNG fuel tank under marine operating conditions through the initial tank state and operating variables. Input factors that are difficult to be included in conventional physics-based thermodynamic models, such as the effects of ship motions and equipment operations, are considered in the D-ANN model. Onboard measured data from an LNG fueled ship are used to train, optimize and verify the D-ANN model, and model structures, input features and hyperparameters are optimized through parametric studies. According to the verification results, the pressure, temperature and filling ratio are estimated with stable error levels within 10 %. Comparing the simulation results of the D-ANN model with conventional models, the D-ANN model shows better prediction accuracy for all self-pressurization, depressurization and long-term operation conditions. • A data-driven model is proposed for thermodynamic behaviors in an LNG fuel tank. • Actual marine operating data of an LNG fueled ship are used for the model. • Effects of ship motions and equipment operations are considered. • The simulation results are verified using onboard measured data during voyages.