GEOPARD - Geostatistical Event-based Object-model Predicted from Analogue Reservoir Deposits

The goal of the GEOPARD project is to create a more geologically realistic basis for decision-making regarding subsurface resource utilization. The commercially available models that exist today are based on technology that is more than 20 years old and have several known weaknesses when it comes to representing realistic geology. Additionally, it can be difficult for geologists without statistical expertise to use them effectively. The GEOPARD project makes a step-change in this realm of technology by creating digital 3D models of rocks types where geological knowledge forms the core of the modeling framework. We aim to make the geological description more accurate and easier to use. This is done by calibrating the input to analogue field data and the reservoir. The solution lies in creating geological rules that fit into a statistical toolbox. With these rules, we describe how the different rocks are deposited, without having to perform complex physical calculations to describe the same. A central contribution from statistics is a robust methodology for conditioning theoretical models to observations, while providing a measure of uncertainty at every stage. An important result of this new way of creating geological models is a more realistic description of fluid flow in pore spaces. This is of great importance in flow simulation of oil, gas, and CO2 in reservoirs on the Norwegian continental shelf. Additionally, it is a very useful tool for geologists who wish to test different hypotheses to improve their understanding of both field analogs and subsurface reservoirs. For example, it is possible to determine which depositional direction best fits the given observations, or what influence the geological properties will have on the further development of the reservoir. GEOPARD is a thoroughly interdisciplinary project and has benefited from knowledge in geoscience, statistics, geomodeling, and software development, together with valuable insights from representatives in the energy industry.

The main project outcome of GEOPARD is a new facies modelling algorithm consistent with both geological rules and reservoir data. The technology developed within the GEOPARD project has the potential to greatly improve the consistency and quality of industry workflows to characterize the subsurface. It can be used to improve understanding of reservoirs on the Norwegian Continental Shelf, with potential for enhanced value creation for Norway and stakeholders alike. The GEOPARD results can be utilized in several applications where understanding of shallow marine depositional processes is important, such as reservoir modelling and carbon sequestration or field studies in academia. Scientific outcomes are advancing Bayesian theory for geological application, more efficient model setup through parameter estimation, and a framework for utilizing geological knowledge in statistical modelling. The new facies model enables the use of analogue data in the modelling process. This will unlock the potential that lies in the large amount of data stored in analogue databases such as SAFARI. It will shift the mind-set of the user when setting up a model from manual parameter tuning to focusing on the underlying geological scenario and how well it fits with the reservoir. The model has potential to work with ensemble history matching workflows, which makes it a candidate for a unique solution to a long-standing problem. Once the model and anticipate that the new facies model will be quickly adopted by the industry.

This proposed project, referred to as GEOPARD, is about bringing more geological realism into the 3D subsurface models used by the Norwegian petroleum industry. The more than twenty year old technology commonly used today is long overdue for an upgrade, and the industry calls for a modern algorithm that can handle increasingly complex well patterns and ensure a realistic representation of the geology. Our solution is to integrate geological rules into the core of the proven Bayesian statistical framework. A geological rule can be for example the stacking pattern of facies objects as a result of the depositional process. A rule-based approach will produce geologically meaningful predictions, allow for efficient testing of geological scenarios and increase the value of reservoir and analogue data. The dominant task is to develop and implement a new facies modelling algorithm that can be used by the petroleum industry in their reservoir management workflows, and research tasks will be focused on supporting this development. The key challenge is to define a set of rules that balances geological realism with statistical consistency. We will utilize geological analogues to define and develop representative rules and objects. It is important that we manage to preserve geological realism in the presence of reservoir data. To support ease of use, we will implement an algorithm to estimate input parameters from interpreted analogue data stored in the SAFARI database. To succeed in this interdisciplinary project, we have joined the forces of geoscience at University of Bergen and statistics at NTNU in collaboration with the modelling community at Norwegian Computing Center. The project involves science recruiting through two research fellowships. International collaboration is established through John Howell at University of Aberdeen, the project leader for SAFARI.