Integrating Machine Learning, Geophysics, and Rock Physics for Efficient Exploration of Natural Hydrogen Reservoirs

Yashee Mathur
Hydrogen’s role is undebatable in the future energy mix owing to its energy density, use in different industries, long-term energy storage capabilities, and minimal CO2 production. The Earth hosts vast resources of naturally occurring hydrogen in different geological environments but the exploration for natural hydrogen is still in the very early stages. Abiogenic sources such as serpentinization and radiolysis are considered the main secondary sources of natural hydrogen generation. The research focuses on developing efficient methods for natural hydrogen exploration. We start at the regional scale to automate the identification of sub-circular depressions and understand blind hydrogen systems by soil gas sampling. At the well scale, we work on the characterization of hydrogen reservoirs by well log data using machine learning and traditional petrophysics. Moreover, using standard and advanced geophysical methods, we work to tie the well-scale data with the field scale geophysical data and eventually characterize hydrogen reservoirs on a larger scale. The overarching goal is to develop methods to expedite the exploration of natural hydrogen reservoirs using different techniques such as machine learning, geophysics, and rock physics.