Reservoir modelling conditioned to seismic and well data

 

PhD Candidate: Deirdre Ann Walsh

Supervisor: Associate Professor Tom Manzocchi

Funded by: Science Foundation Ireland and co-funded under the European Regional Development Fund with further funding from the Petroleum Infrastructure Programme and its member companies.

 

Abstract:

The reservoir modelling workflow generally applied in practical oil-field case studies is to assign different portions of the model volumes to different sedimentary facies (the rock or facies model) and then to assign poro-perm properties to each cell on a facies-by-facies basis (the property model) (e.g. Ringrose and Bentley 2015). Modelling is guided by a conceptual model of the reservoir architecture, by constraints of variable confidence from the seismic data and by the interpretation of the wells. The relative importance of the 2 parts of the workflow, as well as the lateral and vertical scale at which rock modelling gives way to property modelling, depends fundamentally on the sedimentary systems of interest.

This project is concerned with improving the geometrical flexibility available in facies modelling. The three principal classes of algorithm used in facies modelling are pixel-based, object-based and texture-based methods. Texture (or multiple-point statistics) based methods are the most recently-developed class (Strebelle, 2002) and were developed to overcome inadequacies of the other two methods. Like pixel-based methods, texture-based models assign regions of the model to particular facies on a cell-by-cell basis, but this classification is guided not by a variogram, but by a training image. The training image is often an object-based model in which the precise geometrical characteristics of the objects (e.g. channel lengths, widths, wavelengths, depths, orientations) and spatial associations between different object types (e.g. channel levees bounding channels) are defined, but their locations are assigned stochastically by the algorithm. The principle advantage of texture based models over the object-based models is that while it is virtually impossible to honour hard reservoir data, such as the precise distribution of facies observed in the wells, in object based models, it is straightforward to do so with texture-based models (e.g. Caers 2005, Sumner et al. 2005). A constraint of the method, however, is that a texture-based model can only be as good as the training image upon which it is based.

In a pair of industry funded projects (Fault Analysis Group 2005, 2009; Manzocchi et al. 2007), an entirely new class of object-based modelling for deep marine turbidite reservoirs was developed. This so-called compression method is able to reproduce the poorly amalgamated but high net:gross ratio sequence characteristic of many lobe reservoirs. Models produced with the method contrast with conventional object-based models of such sequences in which the amalgamation ratio is inevitably equal to the net:gross ratio. These models therefore become highly connected at low net:gross ratios (Manzocchi et al. 2007). Because of this bias, the industry-standard solution is to use pixel-based, rather than object based methods for facies modelling in lobate reservoirs, as conventional object-based models of moderate or high net:gross systems are simply too connected. The compression-based method has been modified to deal with hierarchical sedimentary elements, and has successfully been applied, in association with a Gulf of Mexico operator, on a field study assessing the implication on production of bed-scale heterogeneity at a full-field scale (Zhang 2015).

A problem common to both conventional and compression-based object-based models is the difficulty of conditioning output realisations to well data. We propose in this project to work around this problem by combining the compression algorithm with the texture modelling. This will require a new workflow that generates a texture model on the basis of a “decompressed” object-based model as a training image and “decompressed” wells as conditioning data. The compression method will then be applied to the texture model to provide the final model which will have appropriate levels of amalgamation yet will honour the well data.

The basic workflow needed will be defined and tested, with the existing in-house code modified to provide the necessary pre-and post- processors to Petrel in which the texture modelling will be performed. The success of the output models will be tested by comparing static connectivity and dynamic flow characteristics of the output texture models against the input training models. The method will be tested using more sophisticated conceptual reservoir models, for example including hierarchically constructed training image. The method will be refined to include “soft” seismic as well as “hard” well conditioning data. The project aims to test the modelling in a real sub-surface reservoir case-study. Outputs from the project will be a set of new modelling algorithms and workflows, as well as descriptions and demonstrations of their applicability.

 

References:

Caers, J. 2005. Petroleum geostatistics. Society of Petroleum Engineers, 88p.

Fault Analysis Group. 2005. Final Report of the FIFT Project: Quantitative characteristics of faults and faults zones and their impact on flow within deep water turbidites, onshore New Zealand. Unpublished Report, University College Dublin.

Fault Analysis Group. 2009. Final Report of the FIFT2 Project: Modelling of the combined effects of faults and sedimentology on fluid flow within turbidites. Unpublished Report, Unviersity College Dublin.

Manzocchi, T., Walsh, J.J., Tomasso, M., Strand, J., Childs, C. & Haughton, P.D.W. 2007. Static and dynamic connectivity in bed-scale models of faulted and unfaulted turbidites. In: Jolley, SJ., Barr., D., Walsh JJ and Knipe Rj. Structurally Complex Reservoirs. Geological Society, London Special Publications 292, 309-336, doi: 10.1144/SP292.18.

Ringrose, P. & Bentley, M. 2015. Reservoir Model Design: A practitioner’s guide. Springer, 249pp.

Strebelle, S. 2002. Conditional simulation of complex geological structures using multiple-point statistics. Mathematical Geology, 34, 1-21, doi: Doi 10.1023/A:1014009426274.

Sumner, W.R., Loutit, S., Tohill, E., Way, D., Criddle, D., Palfrey, A., Hakes, W.G., Archer, S., Colleran, J., Scott, A., Kaiser, K., Harding, A., Castellini, A., Clark, J. & Lianshuang, Q. 2005. Managing reservoir uncertainty after five years of field life: Britannia Field. IN DORE ‘, A.G. & VINING, B.A. (eds) Petroleum Geology: North-West Europe and Global Perspectives- Proceedings of the 6th Petroleum Geology Confrence, 511-525. Petroleum Geology Conferences Ltd. Published by the Geological Society, London.

Zhang, L. 2015. Quantitative characterization and hierarchical modelling of deepwater lobes, Unpublished Phd Thesis, University College Dublin.for low-enthalpy resources, Journal of the Geological Society, London, 164, 371-382.