ANN-based correlation for predicting self-pressurization rate in quasi-spherical liquid hydrogen storage tanks

Abstract

Hydrogen has been considered a promising clean energy carrier and its storage in the liquid state is the most economical choice. One of the main problems related with non-venting storage of liquid hydrogen (LH₂) is the accurate prediction of self-pressurization rate (SPR). For such a prediction task, many thermal models and empirical correlations were developed; however, significant deviations arose. In this study, a correlation has been developed for predicting SPR in quasi-spherical LH₂ tanks, based on a well-trained artificial neural network (ANN) model using 333 data points collected from different literature sources. Normal evaporation rate (NER) and non-venting hold time (NVHT) were considered as the input parameters for the ANN model. Self-pressurization rate was analytically correlated with both normal evaporation rate and non-venting hold time. The applicability range of the developed ANN model for predicting the SPR falls in the value-range between 0.00996 and 0.52026 MPa/day. The developed correlation was validated experimentally and compared to the available literature empirical correlations and numerical models. Compared to experimental work, the ANN-based correlation has been proven to predict SPR with an average error of 1.4% and a favorable correlation coefficient of 0.999897 between the predicted SPRs and corresponding targets, better than the available literature correlations and numerical models. Also, the feasibility of applying this correlation to different case studies, was illustrated. Moreover, the feasibility of potential incorporating the developed ANN model into real-time monitoring systems with some highlights for uncertainty sources, has been considered. The findings of this research offer a valuable correlation for designers and practitioners without the burden of high computational costs and time consumption.