Integrated machine learning and physics-based workflows for rapid qualitative and quantitative insights on monitoring carbon capture and sequestration

Abstract

With the onset of extensive development and focus on clean and sustainable sources of energy, our subsurface geoscience tools and technologies can be effectively repurposed and evolved to support the energy transition. These sustainable energy sources are easily replenishable and low carbon. Some of the emerging sectors are geothermal energy, carbon capture and sequestration (CCS), energy storage, offshore wind energy, sustainable battery-grade lithium extraction, and hydrogen production. These sectors can benefit immensely from economically viable geophysical tools that are available at our disposal to deliver better and faster business outcomes. In this paper we will discuss the integrated machine learning (ML) and physics-based workflows that can be used for enabling effective and rapid energy transition solutions. In our study we will focus on CCS subsurface monitoring workflows, but the same methods have been applied to many other energy transition areas (for example, offshore wind and geothermal energy). The integrated workflows mentioned above were applied on the Sleipner gas field in the North Sea, where CCS operations have been ongoing since 1996. Rapid qualitative workflows like amplitude analysis and seismic blending were used to build an initial understanding of the carbon dioxide (CO2) plume and its spread. Geoscientist-driven ML-assisted horizon mapping facilitated rapid interpretation of structure across different time vintages. Then, advanced techniques like amplitude vs. offset (AVO) analysis and time-lapse seismic inversion were used for quantitative analysis. These analyses gave comprehensive information on CO2 spread in each subformation, as well as an estimate of the changing CO2 volume through time. These workflows can be applied in CCS operations to successfully monitor CO2 in the subsurface and quickly detect and assess the risk of leaks.