Abstract

A statistical study of "J" and Codell wells in the Denver-Julesburg (D-J) Basin was undertaken to determine which parameters influence well productivity. The parameters reviewed were open hole logging parameters and operational parameters. Parameters such as zone thickness, porosity, water saturation, shale indicators and the size of the hydraulic fracture treatment were reviewed. These parameters were chosen because they:

  1. may be related to well productivity;

  2. are easily obtainable from productivity;

  3. are easily obtainable from public sources; and

  4. are related to public sources; and

  5. are related to theoretical hydraulic fracture models.

A non-linear multi-variate regression analysis was used in this study. The variables used in the regression analysis can be related to the variables in the "uniform flux fracture" model.

The results of the regression analysis can be used; 1) to determine which parameters are most correlative to well productivity, 2) to determine the relative magnitude of the parameter's influence and 3) the degree of certainty of the correlation.

The application of regression analysis determined which log characteristics are correlative with well productivity, quantified the influence of each parameter and quantify the degree of uncertainty. Regression analysis can be applied to reservoirs which are not amenable to conventional analysis.

Introduction

Deterministic models and empirical correlation are the two most common methods to determine well productivity. With the advent of personal computers and statistical software, under utilized, but very powerful statistical methods, such as regression analysis should become more popular. Regression analysis offers a reliable method for determining well productivity.

Reservoir simulators are deterministic models. The major difficulties with deterministic models are:

  1. verifying that the model accurately simulates the physical process and

  2. measuring the variables which process and

  3. measuring the variables which are input into the model.

Different models simulating the same process have been offered, these models vary in the degrees of sophistication and accuracy of their results. Many variables which are used in hydraulic fracture reservoir simulators, such as far field stresses and permeability of the sand pack are not measured on a routine basis and may never be measured with any degree of certainty.

Empirical correlation such as hyperbolic decline curve fitting were developed by utilizing historical performance without attempting to determine the underlying factors causing such behavior. Often empirical models have gone awry when the factors causing the historical performance are not present in the case to which they are applied. Anecdotal inference is a form of empirical correlation. Statement such as "you need ten feet of pay to make a well ....." or "our fracture performance is better because..... "are examples of anecdotal inference.

P. 741

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