Case studies using new field-development software

Easy-to-use modeling software that does not rely on grid-based algorithms is helping EOG Resources, Inc. make effective drilling and field-development decisions in record time

Chuck Smith, EOG Resources, Inc. and Dick Barden, Vertex Petroleum Systems

hat is the best, most cost-efficient way to develop the field? What combination of well spacing and completion methods will drain the reservoir in the shortest time at the lowest cost? These perennial questions concern everyone engaged in drilling and production; their answers will determine a company’s success.

To enhance the probability of making good decisions, the industry is increasingly turning to emerging technologies that provide faster access to better information. This article describes the experience of EOG Resources, Inc., formerly known as Enron Oil & Gas Company, with implementing one such emerging technology: a new computer modeling tool for field development. Cases in which the software was used and results that were achieved — as well as its impact on the company — are discussed.

New Software / Modeling Approach

Production and exploitation engineers need to model the inflow performance of vertical, horizontal and hydraulically fractured wells. The problem is that this requires models that are, so far, either too simplistic or too complex (reservoir simulation). Now, an easy-to-use method has been developed for modeling these well types. It also accounts for well interference between fractured or horizontal wells in irregularly shaped reservoirs. The approach allows engineers to adjust wellbore configuration to produce wellhead deliverability curves, as well as run economic comparisons to test the application of different well configurations and reservoir-development scenarios.

Green’s function theory for unsteady-flow problems is the basis for a domain and boundary, integral-equation method for modeling the reservoir. This mathematical approach was chosen for its ability to model sophisticated problems with relative ease. The reader should consult the bibliography for a thorough understanding of the mathematics involved.

The result of this effort is Vertex 1000, a new field-development and reservoir modeling package. It uses an approach that is completely different from the finite-element and finite-difference methods on which all current, grid-based simulation technology is founded.

It runs on conventional PCs under Windows or NT and is based on what the developer, Vertex Petroleum Systems, calls SWARM (Simplified Well, Area and Reservoir Modeling) technology. This dramatically simplifies the modeling process, saving considerable time and affording non-specialists in reservoir simulation the ability to make highly accurate productivity forecasts quickly, easily and at low cost. The current release of the model covers mainly single-phase, primary-depletion problems; but a waterflood version is expected to be released by next summer.

This system was specifically designed as a tool for production and reservoir engineers who have not had special training in complex simulation practices. This enables them to evaluate completion options, predict flow potential, identify optimum well spacing and field-development strategies, as well as consider the economic consequences of each "what if" case.

Background

The software is currently being used by E&P companies in the U.S., Canada and the Middle East. One of these, EOG Resources, Inc. (EOGR), is a medium-size independent with about 5,000 gas wells and 900 oil wells in 500 fields, which collectively produce about 1 Bcfgd and 20,000 bopd. EOGR’s staff includes about 40 production and reservoir engineers. In the U.S. and Canada, EOGR drilled 563 wells in 1998 and expects to drill nearly that many again in the current calendar year, about 90% of which will be development wells.

The company has an extensive library of technical software applications to work with, including a powerful grid-based reservoir simulator and packages for doing well-test analysis, nodal analysis, type-curve analysis, material balance and 3-D geological modeling. However, even with this full array of tools, EOGR found that there was often not enough time to use them effectively due to the volume of problems at hand. To further complicate matters, there were often gaps in their functionality, and data could not always be shared conveniently among these applications.

It was clear that EOGR needed some new technology to generate fast, reliable solutions to its increasing number of field-development problems. In these days of downsized operations, smaller staffs and larger workloads, the company felt that if a good answer — at least an 80% solution — could be quickly and reliably obtained, the benefit would be huge.

Problems often did not warrant a lot of analysis. Management could be comfortable allocating a day’s work to a particular problem, but could not justify spending a full week on some evaluations. It was a matter of assessing business risk and allocating resources accordingly. Unfortunately, this forced local engineers to take shortcuts, basing decisions on analogous solutions or general experience. While this worked some of the time, there was a need for a new type of modeling software that would allow engineers to make multiple reservoir-simulation runs quickly and with more technical accuracy than has been available in the past, resulting in the best solution available within the time constraints.

Real-World Experiences

After satisfying himself that the model worked correctly, i.e., that the math was sound and was calculating correct answers, EOGR’s senior reservoir engineer soon recognized the model’s potential. The single greatest benefit was speed. A new case could usually be built from scratch, a good history match made between observed flowrates and pressures, and field-development scenarios made ready to run — all in about one hour. At the same time, a much better understanding of the reservoir was obtained.

The questions that a manager is going to ask an engineer when he wants to drill or develop a prospect are:

• What can this well be expected to do?
• How many wells must be drilled to maximize drainage and minimize costs?
• What spacing should be used?
• How should they be completed?
• What is the best timing — when should they be drilled?

An engineer needs to have answers to all of those questions. Until now, those answers were frequently based on experience, analogy or some quick calculation, using whatever tools were readily available. With this new technology, people can see what is most likely to happen. They can answer the above questions and have a basis to validate or contradict assumptions made about the reservoir. It gives a good idea as to whether a new well in a defined reservoir will perform up to expectations or, if not, why not.

EOGR has been pleased with how easy the system is to learn and use, and it has now been installed company-wide in the U.S. and Canada. At each site, formal training took only 3–5 hours, and the users / engineers were running their first cases the same day, although it takes a few weeks to really get good with it.

The new model offers a fast, accurate way to interpret and evaluate data for individual wells or an entire reservoir. Because it can predict interference effects among any number of wells, completed in any manner and in any shape reservoir, engineers can model many well-completion, spacing and field-development options within a short time.

EOGR works with many fractured wells, and that can create problems when setting up grid-based models. Fracs require a lot of grid refinement, and it is difficult to generate a grid that gives the correct analysis so, most of the time, people do not even attempt it. An approximation would be to run a normal case and put a negative skin on the well to simulate the frac, but that is not truly modeling the fracture geometry. The new software does not have these limitations and gives good answers for any combination of radial, fractured or horizontal wells, regardless of orientation.

Example Software Use

Described here are several applications of the new software.

Determination of tank size. In the process of establishing a match, the model displays the reservoir volumetrics and allows comparison with well-test analysis and other data. In most cases, it generally confirms prior ideas, but on a recent case, a nearly perfect match showed a tank size that was only about 9% of what the geologists thought it should be. That represented a huge difference and was very valuable information to have when making plans on how to develop the remaining block.

Like everyone else, EOGR’s AFEs are based on expectations of certain well performance. This model facilitates realistic drilling and development decisions and provides greater confidence that those expectations will be met.

High-pressure, relatively low-perm reservoir. This reservoir was modeled with confidence in the parameters. However, after adjusting reservoir volume to get a tubing-head-pressure match with six days’ flowing data, the reservoir size was found to be smaller than expected. A closer geological evaluation indicated that production could all be coming from a small compartment within the main reservoir.

The model was run to predict well performance during the next two months, and results matched the observed data extremely well, Fig. 1. This valuable insight forced everyone to re-examine the reservoir geology; whereas, before, such unexpected results might have been blamed on a bad completion or some other unknown factor. The beauty of it for EOGR was that this evaluation only took an hour of work to see a problem with the original tank-size estimation, which has since been confirmed.

High-rate, radial, gas well. In this well, the objective was to determine the amount of gravel-pack deterioration during the past 18 months. Shortly before discovering the new software, the grid-based simulator was used to solve this problem. In that study, it took a full week to set up the model and determine that the gravel pack was still in good shape. However, using the same set of well and reservoir parameters, the problem was quickly reconstructed in the new model. It matched 450 days of observed data (Fig. 2) and obtained a nearly identical result — all in just 1.5 hours. The ability to obtain such answers with that kind of speed is a strong advantage.

Determining completion results. The next study involved a fracture stimulation of a tight gas well. There was a lot of information about this well, but a lack of confidence as to how the frac job had turned out. The basic objective was to analyze the initial flow-back data to determine effective fracture half-length. In less than two hours, the new software produced a good match with the observed well performance and predicted the fracture half-length, Fig. 3a.

Before solving this problem with the new model, the same analysis was run using the older reservoir simulator. After four days of setting up and tuning the gridded model, it predicted a fracture half-length of about 148 ft. The new model gave the same result (about 145 ft), but in less than two hours.

To test the new model’s predictive capabilities, EOGR also did a pressure-buildup test on this same well and compared observed results with those forecast by the model. As shown in Fig. 3b, the program predicted the actual pressure buildup with an extraordinary degree of accuracy.

In a very tight reservoir — 0.1 to 0.02 md — EOGR engineers were able to run quick iterations comparing net-pay thickness and fracture length to see the effect of each change. For example, what if there was 100 ft of net pay in this zone instead of 228 ft, as previously thought? How would that affect the calculated frac length? By allowing a fast calculation of how that would impact the well, hourly flow data can be used to do history matching with very good results.

Low-pressure gas sand. The question was how best to develop the formation, which covered about 4 mi2. At the time of the study, only two wells had been drilled: Wells 6-10 and 16-10. After getting a good history match using observed flowrates and tubing pressures, the model projected total production for 10 years, based on development plans with 320-acre vs. 160-acre spacing, Figs. 4a and 4b.

When the flow differential between these two cases was plotted, it was discovered that virtually all the additional production would be accelerated into, and realized within, the first four years, Fig. 4c. After that, each section would yield about the same amount of gas, regardless of whether it was being produced by two or four wells. This insight clarified the situation and brought into sharp focus the true nature of the drilling decision. Other field-development cases were subsequently run. These determined the optimum drilling pattern and completion methods that would allow the field to be drained with only two wells per section.

Making drilling decisions. There were several other situations where EOGR found the new model to be highly useful. A recent case was a prospect with a single well in a fault block. The question was whether to drill a second well. The new software predicted the initial pressure of the next well and the resulting production rate. That capability made the model very valuable as an evaluation tool to either verify or postpone drilling. In this case, it indicated that the well would have been a marginal producer. By preventing the drilling of a marginal, yet expensive well, the system had paid for itself many times over.

Using inexact matches. Even when the model does not provide a good match or a conclusive result, the information it generates may be valuable. If the first try is way off the mark, then right away the user knows something is wrong, either with the data entry, the data itself or with his field characterization.

In one case, after quickly running the new model, it became apparent that the frac job did not turn out very well. Although the frac may have been good to begin with, its flow capacity had degraded for some reason. It may have been a 500-ft frac initially, but now it was fluid packed, had sand-grain embedment or something else preventing it from performing. Even when there was not a close match, the model was valuable, because it forced a re-evaluation of the reservoir and / or the completion.

This is actually a good, real-world example of the model’s value, because it can be used to look at the data and see if it makes sense. On one occasion, the model caught some completely erroneous field data. In that case, it was unknown that the data was bad until running it through the model and seeing the outlandish results on the history match. Upon questioning the data, it was discovered that it did not even belong to the well being studied.

Summary

While EOGR has experienced extremely good results with the Vertex software in the cases presented above, it is important to bear in mind that when using this, or any, computer modeling system, one must be realistic. The model is not a black box. A user cannot just plug in ridiculous perms, or other arbitrary data, just to make the model work. EOGR has seen this happen all too often when using its finite-difference (grid-based) models.

This is where knowledge and experience really come into play and the engineer has to use reasonable data and sound judgement to ensure that the model reflects the physical realities of the reservoir. A numerical model will always generate an answer. It is up to the engineer to interpret that answer, determine whether it is correct and what adjustments to the model, if any, are needed to make it right.

Overall, EOGR has found that implementing this new technology has enabled its staff to make much better field-development decisions on a regular basis, saving considerable time and money. It has also been able to shift the focus of model building away from one or two people and to distribute that task to district-level production and reservoir engineers. This has streamlined that part of EOGR’s operations, allowing the field-management, decision-making process to be much more efficient.

For at least 80% of what engineers do every week, this software is a tool that saves significant time and effort. The model’s biggest advantages are that it easy to learn, and it gives good, accurate solutions very quickly. At EOGR, a group of people can get together in an office or conference room, set up a field model, get a match, run some "what if" cases and make a decision that everyone is happy with — all in just an hour or two.

Bibliography

Wong, D.W., J. S. Archer and J. M. R. Graham, "Modeling flow through heterogeneous porous media with boundary integrals using higher-order surface singularities," 2nd European Conference on the Mathematics of Oil Recovery, Arles, 1990, pp. 113–120.

Kikani, J. and R. N. Horne, "Pressure transient analysis of arbitrary shaped reservoirs with the boundary element method," paper SPE 18159, presented at the 63rd Annual Technical Conference and Exhibition of the Society of Petroleum Engineers, Houston, Texas, Oct. 2–5, 1988.

Lee, S. H., "Analysis of productivity of inclined wells and its implication for finite-difference reservoir simulation," SPE Production Engineering, May 1989.

Oguztoreli, M. and D. W. Wong, "VERTEX: A new modeling method used to direct field development," paper SPE 39806, presented at the 1998 Permian Basin Conference, Midland, Texas, March 25–27, 1998.

The authors

Chuck Smith is the chief reservoir engineer for EOGR Resources, Inc., Houston. He graduated from the Colorado School of Mines in 1970 with a professional engineering degree and began his career in West Texas. He has spent the balance of his 30-year career working as a reservoir engineer on domestic and international oil and gas projects.

Dick Barden, with degrees in business and law, is president and CEO of Vertex Petroleum Systems, Englewood, Colorado. VPS is the third company Mr. Barden has formed since entering the oil and gas industry in 1973, and the second to be involved with developing software applications specifically for the petroleum sector.

Copyright © 1999 World Oil
Copyright © 1999 Gulf Publishing Company