Efficient Large-Scale Reservoir Simulation on Modern Many-Core

In this project, we will research and test modern and advanced computing techniques relevant for massively parallel next-generation engineering applications with special emphasis on simulation of subsurface flow in the context of Carbon Capture Storage (CCS).

Design, optimization and risk management of CO2 storage schemes are based, in part, on predictions from reservoir simulation tools that take as input a model of the subsurface in addition to a set of physical models that describe the behavior of fluids that reside in or are injected into the subsurface. Many oil and gas reservoirs have large amounts of seismic, geologic and dynamic reservoir data available. This vast amount of data provides high resolution geological models. However, for conventional reservoir simulators to run in practical times, upscaling of these high-resolution geological models from data points at a density of e.g. 25-50 meters to 250 meters is required. Simulations based on upscaled reservoir properties often fail to predict oil recovery accurately and are associated a great deal of uncertainty. To fully utilize the seismic data and capture the flow of components in the reservoirs more accurately, simulators must be able to accommodate multimillion-cell models. The overarching goal of the proposed research activities is to further develop, implement and test advanced numerical methods in subsurface flow simulation that promise to work well in combination with modern programming paradigms and hardware. Most existing codes in industry today are legacy codes, which cannot immediately and optimally utilize new technology and programming paradigms. Therefore, this project will continue recent work on exploiting the potential of such methods in reservoir simulation tools with direct application in design and risk assessment of CCS operations.



Supervisor: Assoc.Prof. Allan P.Ensig-Karup, apek@dtu.dk

Co-supervisor: Assoc. Prof. Kristian Jessen, Tie-Line Technology

Consultant; Stefan Lemvig Glimberg, Lyod's Register


Allan Peter Engsig-Karup
Associate professor
DTU Compute
+45 45 25 30 73