Pressure and saturation estimation from PRM time-lapse seismic data for a compacting reservoir

Abstract

Observed 4D effects are influenced by a combination of changes in both pressure and saturation in the reservoir. Decomposition of pressure and saturation changes is crucial to explain the different physical variables that have contributed to the 4D seismic responses. This thesis addresses the challenges of pressure and saturation decomposition from such time-lapse seismic data in a compacting chalk reservoir. The technique employed integrates reservoir engineering concepts and geophysical knowledge. The innovation in this methodology is the ability to capture the complicated water weakening behaviour of the chalk as a non-linear proxy model controlled by only three constants. Thus, changes in pressure and saturation are estimated via a Bayesian inversion by employing compaction curves derived from the laboratory, constraints from the simulation model predictions, time strain information and the observed fractional change in 𝑉𝑃 and 𝑉𝑆. The approach is tested on both synthetic and field data from the Ekofisk field in the North Sea. The results are in good agreement with well production data, and help explain strong localized anomalies in both the Ekofisk and Tor formations. These results also suggest updates to the reservoir simulation model. The second part of the thesis focuses on the geomechanics of the overburden, and the opportunity to use time-lapse time-shifts to estimate pore pressure changes in the reservoir. To achieve this, a semi-analytical approach by Geertsma is used, which numerically integrates the displacements from a nucleus of strain. This model relates the overburden time-lapse time-shifts to reservoir pressure. The existing method by Hodgson (2009) is modified to estimate reservoir pressure change and also the average dilation factor or R-factor for both the reservoir and overburden. The R-factors can be quantified when prior constraints are available from a well history matched simulation model, and their uncertainty defined. The results indicate that the magnitude of R is a function of strain change polarity, and that this asymmetry is required to match the observed timeshifts. The recovered average R-factor is 16, using the permanent reservoir monitoring (PRM) data. The streamer data has recovered average R-factors in the range of 7.2 to 18.4. Despite the limiting assumptions of a homogeneous medium, the method is beneficial, as it treats arbitrary subsurface geometries, and, in contrast to the complex numerical approaches, it is simple to parameterise and computationally fast.

Finally, the aim and objective of this research have been met predominantly by the use of PRM data. These applications could not have been achieved without such highly repeatable and short repeat period acquisitions. This points to the value in using these data in reservoir characterisation, inversion and history matching.