Abstract

This study reports a procedure for validating a reliability of a location distribution of fractures in highly populated portions (clusters) of a constructed discrete fracture network (DFN) model. In the procedure, we focus on a relationship between a fracture density distribution along boreholes and spatial extents of clusters of fractures in a DFN model. The numerical experiments show that a shape of a fracture density distribution function evaluated using fractures crossing boreholes mainly depends on a location distribution of fractures, and is less affected by a size distribution and an orientation distribution of fractures.

The DFN model is constructed by the procedure we had proposed previously. At the procedure, geostatistics method is used for predicting extents of clusters of fractures. However, it is pointed out that the extents of clusters have uncertainty due to the prediction by using only the observed locations of fractures along boreholes. This uncertainty reduces a reliability of a constructed DFN model.

To meet the issue, the fracture density distributions along boreholes are calculated for constructed DFN models with different extents of clusters and compared with the observations. Through the studies, it is confirmed that the fracture density distributions along boreholes are useful to constrain the uncertainty of the extents of the clusters and that the optimum extent of the cluster can be extracted by focusing the fracture density distribution along boreholes. Additionally, fluid flow behaviors of the constructed DFN model with the optimum extents of clusters is verified by flow simulations.

Introduction

Japex has drilled totally 14 wells in the Yufutsu oil/gas field in central Hokkaido, northern Japan, where a fractured gas reservoir is developed in a basement rock. From the evaluation of the well tests, it is found that productivity of each well varies widely and some of wells are almost none productive in spite of the existence of fractures and gas showing. One of the factors affecting the wide variation of the productivity is inferred to be related to a strong heterogeneity of fracture distributions. To optimize the development and the operation for the Yufutsu reservoir, a reservoir model, which includes reliable fracture distributions, is anticipated.

We have been taking a Discrete Fracture Network (DFN) approach to represent a fracture network in the Yufutsu reservoir, and had proposed the procedure to construct a DFN model from micro-resistivity borehole image data4. In the procedure, a DFN model is constructed by using three kinds of fracture parameters, i.e. the size, the orientation and the location distributions of fractures. These fracture parameters are estimated from the information obtained from the boreholes.

Generally speaking, it is difficult to expand the statistics extracted from one-dimensional boreholes to a three-dimensional space directly, in the case that fracture distributions show heterogeneous features. If one wants to construct DFN model by expanding 1-D statistics, it requires an inverse process to verify the suitability of the fracture distribution after constructing DFN model.

To validate a 3-D fracture distribution in a DFN model created by using well data, Dershowitz et al.1 and LaPointe et al.2 had proposed a forward DFN modeling procedure to calculate 1-D statistics along simulated boreholes placed in a DFN model. In the procedure, the 3-D fracture distribution is modified and validated by comparing the 1-D statistics along the simulated boreholes with the observations. They apply the procedure to validate a size distribution of fractures by using a trace length distribution of fractures along boreholes.

In the case of the Yufutsu DFN modeling, the location distribution of fractures includes the largest uncertainty compared with the other two parameters, because it is estimated from borehole on the assumption that fracture centers are located just along boreholes. The location distribution of fractures should be validated.

This content is only available via PDF.
You can access this article if you purchase or spend a download.