This study reports a new process of constructing a discrete fracture network (DFN) model, which reflects various statistical properties of fractures extracted from static data such as borehole images and dynamic data such as pressure derivative curves.
A DFN model is constructed by generating a number of disc-shaped fractures based on their statistical properties. The process of constructing DFN model proposed consists of two streams. One stream is for the fractures with wide apertures, which work as main flow path of fluid. Another stream is for the fractures with narrow apertures, which work as main storage of fluid. In the each stream, a DFN model is constructed by applying fractal theory and geostatistics that define the size and spatial distribution of fractures, respectively. The overall DFN model is obtained by merging the two types of DFN models. Then, the DFN model is converted to a continuum model with equivalent permeable blocks based on cubic law and is evaluated by the fluid- flow behavior.
Japex has been operating the Yufutsu fractured basement gas reservoir since 1996. A total of approximately 8,000meters of electrical borehole images from eleven wells is measured to understand the fracture system. According to the proposed process, the DFN models are constructed. To evaluate the DFN models by comparing the simulated pressure derivative curves with the observed curve, flow simulations are carried out as a simulation of a single porosity type. The observed pressure derivative curve shows a concave shaped curve like a dual porosity behavior, but the curve never reaches the second flat state. Through the parameter studies of the statistical properties, which relate to the size distribution and the spatial distribution of fractures, the observed curve was successfully reproduced and the quantitative relationship between the fractures with wide apertures and the fractures with narrow apertures was found.
JAPEX has been operating the Yufutsu fractured basement gas reservoir since 1996, which is located in central Hokkaido, northern Japan. The reservoir extends over a depth range of 4,000 to 5,000 meters. Totally 14 wells have been drilled in this reservoir up to now. The measured pore pressure at each well follows the same pressure trend so that each well is inferred to be drilled in an identical pressure system. However, some of wells are almost none productive in spite of the existence of gas showing and fractures. To study the behavior of the reservoir and to optimize the operation and development of the Yufutsu field, a reliable fluid flow model is anticipated.
Heterogeneity of fractured rock masses makes it difficult to construct a mathematical model to represent fluid flow. The recent studies focus on discrete fracture network (DFN) models constructed by generating a number of disc-shaped fractures based on their statistical properties.2,3,6,8,11 The advantage of the DFN model is to accommodate more realistic fracture network geometries and characteristics than conventional dual porosity models. It is available to put various kind of information of fractures observed in fields into the DFN model with respect to their statistical properties.
One of the important issues in constructing a reliable DFN model is what kinds of data should be used for the construction of DFN models. In the case of the Yufutsu field, seismic data is not suitable for the quantitative analysis of the fractures due to its low quality influenced by the overlaid thick volcanic rocks. The static data such as borehole images are used for detecting each fracture in the reservoir to extract statistical properties of fractures. The dynamic data such as transient single well tests are used for evaluating the fluid flow behavior of DFN models constructed by the extracted statistical properties. Another issue is what process for constructing DFN models should be adapted to reflect the features of their statistical properties.
The purposes of this study are to propose a process for constructing DFN models using electrical borehole images of multiple wells, to incorporate the variation of productivity of each well into the DFN models, and to study parameters of statistical properties of DFN models by comparing with the observed fluid flow behavior.