Date of Award

Fall 2017

Degree Type

Thesis

Degree Name

Master of Science (MS)

Committee Chair

David Reichhardt

First Advisor

B. Todd Hoffman

Second Advisor

Hilary Risser

Abstract

Modern reservoir studies use reservoir simulation to predict future production and plan reservoir development. Due to the increase in production in unconventional reservoirs in recent years, the number of hydraulically fractured wells has risen drastically. The days of using simple analytic techniques to account for the production of a single hydraulic fracture in a vertical well are gone, and the need to be able to model numerous hydraulic fractures in many stages over long horizontals is required.

This brings up the study question of this research: What is the best way to model hydraulic fractures in a flow simulator? There are several methods to model hydraulic fractures that have been published, but there does not seem to be a clear consensus of what the best method might be or why a certain method is used.

To determine which hydraulic fracture modeling method might be the best, a selection of methods is chosen based on what is available and what is commonly used in various reservoir simulation software packages. This research was performed in the Petrel/Eclipse reservoir simulation environment. The methods being test were the Easy Frac, Schlumberger Correlation, LGR/Tartan grid, and the Conductivity Filter. These methods are tested on two real-world well datasets: A Vertical Dry Gas well and a Horizontal Oil Well model.

The Easy Frac and Schlumberger Correlation are a form of uniform conductivity rectangular fracture that is common in most modern reservoir simulation software. The user selects the location and the desired hydraulic fracture properties and the software changes various properties to increase production and model the hydraulic fracture. The LGR/Tartan grid breaks cells down into smaller sections to allow the user to specify a hydraulic fracture width that can then have a fracture permeability applied to model the hydraulic fracture conductivity. The Conductivity Filter also models the hydraulic fracture conductivity by recalculating the fracture permeability using the width of an entire grid cell as the hydraulic fracture width. This permeability is then applied to the grid cells that represent the hydraulic fracture.

Grid cell sensitivities were performed to determine the best grid cell sizes for both the Vertical Dry Gas Well and the Horizontal Oil Well models. Once these grid cells sizes were chosen, cumulative production was compared, a run time comparison was performed, and

a least squares sum analysis was performed to obtain a numerical value of how well a method matched the magnitude of observed production.

In the Vertical Dry Gas Well model the Schlumberger Correlation had the best results when visually comparing production to the observed data as well as having the smallest least squares sum value. It also managed to capture the production trend while maintaining a run time that was in the middle of the other methods.

The Horizontal Oil Well did not have the best results due to methods not accurately capturing the shape of the observed production. The method that captured the last 6 months of the observed production and had the lowest least squares sum value was the LGR/Tartan grid method.

Overall there was no clear better method than others. Some methods performed well in certain situations and worse in others. Ultimately it will come down to the desired outcome of the reservoir simulation model to determine which method will best suit that purpose as well as the user knowing the advantages and disadvantages of each method.

Comments

A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in Petroleum Engineering

Share

COinS