Physical Property Prediction and Simulation Analysis of Hydrogen-Doped Natural Gas Pipeline

Abstract

Hydrogen-doped natural gas, a blend of hydrogen and natural gas, has emerged as a promising candidate due to its potential to reduce greenhouse gas emissions and enhance energy efficiency. However, the physical properties and operational dynamics of hydrogen-doped natural gas restrict the efficient operation of natural gas pipelines. Physical properties and operational dynamics of hydrogen-doped natural gas pipelines are investigated to combine machine learning techniques and simulation models, which promote the development of zero carbon emission energy, hydrogen energy, thereby contributing to the reduction of global carbon emissions. The REFPROP software is utilized to construct databases for predicting the physical properties of hydrogen-doped natural gas. Among various machine learning models, the Wide Neural Network emerges as optimal, exhibiting an exceptional R² value exceeding 0.9999 and the ability to predict over 264,000 data points per second for density and viscosity. Additionally, a simulation model is developed and rigorously validated against COMSOL 5.0 commercial software, demonstrating its capability to accurately simulate pipeline operations. Matching the calorific value of hydrogen-doped natural gas is very important to ensure downstream energy supply and production operations' efficiency. Thus, various operation cases such as constant pressure and constant calorific value were studied. Overall, this study provides valuable insights into optimizing hydrogen-doped natural gas pipeline operations and advancing green energy initiatives.