PRACTICAL DYNAMIC UPDATING OF RESERVOIR MODELS USING FREQUENTLY ACQUIRED 4D SEISMIC DATA: Final report for Phase II of the Seismic History Matching Project

Karl Dunbar Stephen, Colin MacBeth, Asghar Shams, Nureddin Ramadan Edris, Farzaneh Sedighi Dehkordi, Alireza Kazemi

Research output: Book/ReportCommissioned report

Abstract

Repeat surveys of 3D seismic are now gathered on a routine basis in a number of North Sea and Gulf of Mexico fields as companies such as BP, ConocoPhillips, Shell, Statoil and Total are moving towards Life of Field Seismic. Changes in hydrocarbon pressures and saturations may be inferred from differences between surveys leading to greatly improved reservoir monitoring and hence better management decisions. In many of those fields the seismic data is used to update the geological or flow simulation model, mostly in a qualitative fashion. While this process leads to a better understanding of the reservoir, it is labour intensive and the results are somewhat subjective.
At Heriot-Watt University, we have developed an assisted history matching method where simulations are quantitatively compared to observed seismic and production data and then updated in an objective manner. Our approach has been tested on two UKCS fields (Nelson and Schiehallion) in Phases I and II of the project. We have successfully integrated the seismic into the history matching workflow and developed methods to automatically choose new parameters (such as permeability, net:gross, fault transmissibility, stress sensitivity of the petro-elastic model) and developed methods for quantifying model errors that arise from using coarsely defined geo-models or simplified flow physics such as streamlines.
Several major topics have been addressed during the project. We investigated the value of 4D seismic in history matching and found that it is extremely useful for estimating reservoir properties between wells and around injectors. These properties have much less effect on the production data that are used in conventional history matching. The curse of dimensionality has been a common challenge in our studies. In the Schiehallion field, we found that incorporating history data into the updating process in stages helps to reduce the number of parameters and thus leads to a quicker and better history match. Similarly,
in the Nelson field study, we divided the history matching problem into more manageable elements spatially and gained significant improvements in the search for the best model or models. Updating the Nelson field by region reduces the number of parameters considerably. We have developed an approach for doing this in a more quantified fashion that includes a parallel search of parameter values. We are thus able to significantly reduce the number of models required by a factor of 3 or more. In synthetic studies, simulation model errors have been shown to cause biased results and optimistic uncertainty measures. These can be reduced with suitable calibration. This has been particularly useful when using streamlines or upscaling to speed up simulation. We also compared simulation errors to residual errors in the geological model after updating and found the former to be more important.
History matching using time-lapse seismic is a complex process and with the above developments, we are able to improve the models upon which business decisions are based.
Original languageEnglish
PublisherInstitute of Petroleum Engineering, Heriot-Watt University
Number of pages19
Publication statusPublished - May 2008

    Fingerprint

Cite this