SPE Reservoir Evaluation & Engineering
Volume 12, Number 6, December 2009, pp. 951-962

Summary

We used an integrated solution by combining "direct" and "inverse" approaches to fracture network characterization in a stochastic numerical model. Static geological data obtained from cores and well logs were used together with dynamic data such as well-test responses to build 3D discrete fracture-network models. We used the data obtained from the fractured carbonate Midale field in Canada.

The fractured-reservoir model was constructed from static and dynamic (drawdown and pulse-interference tests) data. Matrix and several fracture parameters including fracture length, density/spacing, aperture, connectivity, and orientation were evaluated in a quantitative sensitivity study to determine which characteristics have a higher influence on the accurate match to well-test response. We used experimental design to optimize the number of simulations needed for a sensitivity study and history match. The sensitivity analysis revealed a strong influence of matrix quality on the pressure response, suggesting that the history match can be specific to the simulated process and not necessarily unique. The results emphasize the contribution of matrix in the Midale reservoir and the need to simulate a broader range of processes for an accurate description of the fracture/matrix system dynamics. In a general sense, the approach used in this study proved to be useful in integrating fracture data from different sources and assessing its reliability and relative importance.

Introduction

In this study, we apply a widely accepted integrated approach to characterize the matrix/fracture system of the Midale field in southeastern Saskatchewan, Canada. Often serving as an example for classical NFR in literature, Midale is a perfect case study for several reasons. The field produces light oil from a mature carbonate reservoir. Currently Midale is undergoing a full-field CO₂ flooding, which became possible largely thanks to the success of the pilot CO₂ flood project in the late 1980s. Massive amounts of data and experience were accumulated throughout the years. Moreover, considerable research was conducted into matrix and fracture characterization and production mechanisms (Payne 1988; Beliveau 1989; Beliveau et al. 1993; Malik et al. 2006; McKishnie et al. 2005). Nevertheless, the success of enhanced recovery still depends on further research into the matrix heterogeneity, NFN geometry, and fracture/matrix interaction.

Developments in computational and analytical methods have provided us with some useful tools, which facilitate the method described below. This paper describes how integrated fracture and matrix characterization using static and dynamic data helped us to construct and validate a discrete 3D fracture network embedded into reservoir matrix. Furthermore, we present the statistical analysis of designed flow simulations used to clarify the role of NFN properties in the reservoir performance.

© 2009. Society of Petroleum Engineers

View full textPDF ( 992 KB )

History

• Original manuscript received: 30 June 2008
• Meeting paper published: 3 December 2008
• Revised manuscript received: 29 April 2009
• Manuscript approved: 21 June 2009
• Published online: 17 December 2009
• Version of record: 31 December 2009