Development of correlation equations on hydrogen properties for hydrogen refueling process by machine learning approach. (19th January 2022)

• Record Type: Journal Article
• Authors: Park, Byung Heung; Chae, Chung Keun
• Is Part Of: International journal of hydrogen energy. Volume 47:Number 6(2022)
• Page Start: 4185
• Page End: 4195
• Publication Date: 2022-01-19
• Subjects: Hydrogen refueling -- Hydrogen property -- Machine learning -- Correlation equation -- Refueling simulation; Hydrogen as fuel -- Periodicals; Hydrogène (Combustible) -- Périodiques; Hydrogen as fuel; Periodicals; 665.81
• DOI: 10.1016/j.ijhydene.2021.11.053
• Languages: English
• ISSNs: 0360-3199

Abstract

Abstract: The energy transition which refers to shift of the energy system from fossil-based resources to renewable and sustainable energy sources becomes a global issue to mitigate the progression of climate change. Hydrogen can play an important role in long-term decarbonization of energy system and achievement of carbon neutrality. Currently, the utilization of hydrogen in the energy system is focused on a road transportation sector as a fuel in a vehicle fleet. Compressing gaseous hydrogen is the most well-established technology for storage in hydrogen-fueled vehicles. The refueling hydrogen requires short filling time while ensuring the safety of storage tanks in a vehicle. However, a fast filling of hydrogen in high pressure leads to a rapid temperature rise of hydrogen stored in tank. Therefore, many numerical and experimental studies have been carried out to analyze the filling process. Various thermo-physical properties of gaseous hydrogen such as density, viscosity, and thermal conductivity are required for the numerical studies and the accurate hydrogen properties are essential to obtain reliable results. In this work, a polynomial equation is proposed with respect to temperature and pressure in ranges of 223.15 K < T < 373.15 K and 0.1 MPa < P < 100.1 MPa to present various hydrogen thermo-physical properties by adopting different coefficients. The coefficients are determined by a machine learning method to regress the equation using a great number of reference data. The equation is trained, tested, and validated using different datasets for each property. The order of the equation has been changed from 2 to 5. Then, the accuracies are estimated and compared with respect to the order. The average relative errors (REs) of the 5th order equation are assessed to lower than 0.3% except for molar volume and entropy. The accuracy of the equation is also examined with experimental data and other correlation equations for density, viscosity, and thermal conductivity which are required for numerical simulations of hydrogen refueling. The proposed equation presents better accuracy for viscosity and thermal conductivity than literature equations. In density calculation, a literature equation shows better performance than the proposed equation, but the difference between their accuracies is not so significant. In calculation time comparison, it is revealed that the proposed equation rapidly responses adequate to computational fluid dynamics (CFD) simulations. Results of the study can provide accurate and reliable hydrogen property values in a fast and robust means specifically for simulation of hydrogen refueling process, but not restricted only to the process. Correlation equations proposed in the present work can aid in optimizing a hydrogen value chain including production, storage, and utilization by providing accurate hydrogen property. Graphical abstract: Image 1 Highlights: Correlation equations are proposed for various thermo-physical properties of gaseous hydrogen. Temperature and pressure ranges are 223.15–373.15 K and 0.1–100.1 MPa, respectively. The equations are polynomial forms and coefficients are determined by a machine learning approach. Total 7626 data points for each property are used to train, test, and validate the equations.