Practical Flow-Simulation Method for a Naturally Fractured Reservoir: A Field Study

Summary

This paper presents the methodology, implementation, and results of the dynamic modeling of a naturally fractured carbonate reservoir, which consists of a fracture swarm system and high matrix permeability. The focus of this paper is on the description of the dynamic single-media model that was used to history match the production. The challenges in properly quantifying the separate effects of matrix and fracture within the framework of a single-media model become the major objective of the study. A brief description of the static model (which consists of the development of matrix and fracture models, as well as the method to integrate them) is also presented.

Both matrix and fracture systems play an important role in the production mechanism of the reservoir. History matching for 20 years of production was done successfully in a single-media model through an iterative process between static and dynamic models to ensure the consistency between the two models. Different sets of relative permeability curves (for matrix and fracture systems) were generated to properly simulate the fluid movement in the reservoir.

The results indicate an excellent history match that is also followed by the success in the blind tests of newly drilled wells. Significant improvement was obtained compared to the previous model in simulating water production that comes from the aquifer to the wells through a complex fracture network. The results were achieved through various sensitivity analyses of fracture conductivity and by fine-tuning relative permeability curves that are based on the eight rock types defined for the reservoir.

In conclusion, the single-media model used in this study can successfully simulate the behavior of a naturally fractured reservoir. The model has shown that it can produce excellent results in a very practical way.

The methodology described in this paper is suitable for other naturally fractured reservoirs, especially in the Middle East area, where a significant difference between core-derived permeability and well-test-derived permeability exists.

Introduction

This paper presents the methodology, implementation, and results of the dynamic modeling of a naturally fractured carbonate reservoir of an oil field in the Middle East. The study is part of the field's Reservoir Management and Long Term Development Plan (LTDP).

The overall objective of the study is to obtain a representative model that can be used to improve the reliability of the predictions of future performance for the reservoir. The specific objective is to match the 20 years of production history and to predict field future performance with three flow-simulation models that represent the Low (pessimistic), Medium (most-likely), and High (optimistic) cases.

The dynamic model (or flow-simulation model) is based on a static model that was generated through a detailed integrated reservoir characterization study (Al-Deeb et al. 2002; Bahar et al. 2001, 2003a, 2003b; Charfeddine et al. 2002; Ates et al. 2003). The study included the integration of various data/information from the geology, geophysics, and engineering disciplines using a stochastic approach. The static model was developed in a fine-scale grid system of 4.2 million cells that has been upscaled into a coarse-scale grid system of 181,000 cells. The upscaling was performed through a grid optimization process using the streamline simulation to maintain a high level of heterogeneity in the reservoir as much as possible (Ates et al. 2003).

The main characteristic of this reservoir is the presence of a complex fracture network within the highly heterogeneous matrix system. Additionally, the reservoir is characterized by the presence of Sucrosic Dolomite rock (mainly in the south part of the Middle reservoir unit) that significantly enhances matrix permeability further. Matrix permeability ranges from 0.01 md to 12 Darcy, with the average in excess of 100 md. The fractures are described as fracture corridor systems (or fracture swarms) that show two major trends, namely N40E and N70E. Fracture swarms/subseismic faults are structural features extending laterally over several hundred meters (and, sometimes, over several kilometers). This network was assumed to be extended fully in the vertical direction. It is assumed that there are very few fractures outside the fracture swarms.

The flow-simulation model used for the dynamic modeling is the single-media model. Besides having the advantage of being computationally efficient, it is believed that a single-medium model is adequate for this reservoir because it does not consist of a diffuse fracture system but, rather, fracture swarms, as described in the previous paragraph.

Three major stages of dynamic modeling, namely Stage 1 (original model), Stage 2 (calibrated model), and Stage 3 (final model), are presented in this paper (Fig. 1). The progress at each stage is evaluated with a newly developed scoring system. With this system, an objective tool is used to monitor the progress of the history matching. The score achieved at each stage for the Medium case is also shown in Fig. 1.

The history matching of the three realizations (Low, Medium, and High) has been done in the following manner. First, the Medium case is run until it reaches a certain point, at which significant understanding of the model has been obtained. In this study, this point refers to the condition in which approximately 75% of history matching of the medium case has been achieved. At this point, lessons learned from the Medium case are applied to the other two cases.

Ultimately, the overall goal of the study is to have the tool for a better prediction of future performance. The prediction run is currently ongoing. The topic of future performance is beyond the scope of this paper.