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

Oil companies such as BP, Shell and Statoil now repeat 3D surveys in the North Sea almost routinely. These surveys can convey important information regarding the movement of fluids during hydrocarbon production and they are becoming an essential tool in reservoir management. The transference of time-lapse information into changes in the reservoir description is often qualitative, however, and oil companies recognise that the full value of the data is not yet being attained.
One of the key difficulties in applying time-lapse data lies in integration with the Petroleum Engineer’s workflow. This is partly due to cross-disciplinary barriers but mainly due to the non-qualitative use and the non-uniqueness of the time-lapse data. The Seismic History Matching Project at Heriot-Watt University has worked from May 2002 to August 2004 to address this. We have developed a quantitative approach where repeated surveys can be integrated into reservoir modelling and simulation studies to provide a more accurate understanding of the reservoir and its uncertainties. Our methodology bypasses many of the data handling problems of current approaches by converting reservoir simulation properties to the seismic grid in a systematic manner.
In this report, we present our automated multiple model history matching method, which integrates time-lapse seismic with production data, and determines parameter uncertainty. For each simulation model, we compare observed seismic attributes to synthetic impedance obtained via a petroelastic modelling step and grid transformation. For seismic and production data, we then obtain a misfit which is used to update our model parameters in a Bayesian framework and accounts for model errors and data covariance.
We have applied the method to a UKCS field. Based on the operator’s original model, we demonstrate the full potential for incorporating time-lapse seismic by history matching to synthetic production and seismic data obtained from a simulation model. We also show a good match to observed field data after first matching the static model to the RMS amplitudes and subsequently matched time lapse impedance to changes in an attribute obtained from the sum of negative values of the Coloured Inversion Stack around the top of the reservoir.