Optimization methods for the real-time inverse problem posed by modelling of liquefied natural gas storage

If two liquefied natural gases (LNG) obtained from two different sources are inappropriately fed into a storage tank, lighter LNG may lie over heavier LNG forming a stratification, which could eventually lead to a rollover. Few models available in the literature predict time to rollover in LNG storage tanks. These are semi-empirical in nature as they are based upon empirical correlations to estimate heat and mass transfer coefficients across the stratified layers. We present a lumped parameter model in order to predict time to rollover and to investigate its sensitivity to variation of heat and mass transfer coefficients. The novelty of the present work is its ability to estimate heat and mass transfer coefficients from the real time data using an inverse methodology. We assimilate the real time LNG level–temperature–density (LTD) data from LNG storage tank in order to estimate heat and mass transfer coefficients from the densities of the stratified layers. The optimized heat and mass transfer coefficients are then used to predict time to rollover. We present a sequence of LTD profiles obtained from real time LNG terminal and which are leading to rollover in one case study (Section 4.1). The time to rollover predicted using this inverse methodology is compared with the LTD profiles obtained from real LNG tank and also with time to rollover obtained using empirical correlations. Heat transfer coefficients estimated using empirical correlations are found to be over-estimated for some case studies, which under predict time to rollover. For the real time case study, time to rollover predicted using empirical correlations is under predicted by about 84%, where as that using the inverse methodology is under predicted by about 20%.