Pore-scale modelling of methane-driven helium extraction in deep tight sandstones.

Helium is a strategic resource for cryogenic, aerospace, semiconductor, and other advanced technologies, but in deep tight sandstone reservoirs it may occur partly as dissolved helium in formation water, and its transfer into a producible gas phase remains poorly quantified. Here we examine methane-driven helium extraction under representative deep tight-reservoir conditions by coupling sub-voxel pore reconstruction, gas-water partitioning calculations, and pore-scale transport simulations. The analysis uses geological boundary constraints from the Ordos Basin, weakly supervised convolutional neural-network reconstruction of sub-voxel porosity, thermodynamic gas-water partitioning calculations, and lattice Boltzmann pore-scale transport simulations. The reconstructed digital rock better preserves micro-nano pore connectivity than simple binary segmentation and is used to estimate Knudsen/slip-controlled transport contrasts between helium and methane. The simulations indicate that helium can maintain a slightly higher apparent permeability than methane in nano-confined pore throats because of its smaller molecular diameter. This relative mobility advantage increases at low pressure but becomes small under the representative deep-reservoir pressure used in this study. Thermodynamic calculations indicate that high temperature, high pressure, and high salinity favour helium partitioning from the aqueous phase into the gas phase, whereas methane flow increases interfacial contact and convective stripping. Multi-slice extraction simulations further suggest that helium release is concentrated near gas-water contact zones and is later limited by diffusion-driven replenishment from adjacent water-bearing pore regions. The results support an analogue pore-scale interpretation of methane-assisted helium mobilization in a tight-sandstone pore network, but their reservoir-scale generality remains limited by the use of an analogue digital-rock dataset and slice-based pore-scale simulations.