Unconventional/complex reservoirs

Unconventional Resources: A Typical Well Is Hard To Define

A type curve is a quick way to answer a critical question—what does a typical well produce over time in a given place?

A plot of 50 Barnett shale wells in Tarrant County shows how the actual numbers (shades of gray) can diverge from the average (red line).
Source: SPE 178525.

A type curve is a quick way to answer a critical question—what does a typical well produce over time in a given place?

On the plus side this simple calculation can be done using only basic math skills. “The conventional approach is to determine the arithmetic average of production during a given month from different wells in a reservoir of interest to create a type well,” according to SPE 178525.

On the downside, the next sentence is: “This method is deeply flawed.”

The flaws include “different results by different evaluators” that are “overestimates or underestimates (usually over­estimates” of future production, according to the paper, whose authors include John Lee, a professor at Texas A&M University who is well known for his work on reservoir production analysis.

The negatives have become all the more pronounced by the rise of unconventional oil and gas developments, where the output varies, and is unpredictable from well to well, with decline curves that do not fit conventional norms.

“The problems are simply different (and usually harder to solve) in unconventionals,” Lee said.

That has lead Lee and other members of a committee within the Society of Petroleum Evaluation Engineers (SPEE) to set off on creating a new standard method to more accurately calculate type curves for wells. Lee said the 10 volunteers hope to complete SPEE Monograph 5, providing recommendations and workflows for constructing Typical Well Production Profiles (type wells) for unconventionals by 2018. But he would not be surprised if it takes longer.

Since those who do type curve calculations cannot wait, Lee recently offered some advice on how to do meaningful estimates of the reserves or production added by a typical well at a breakfast session during the Unconventional Resources Technology Conference.

One change he is sure about: He wants to change the widely used label, type curve, to type well to better describe how to create a model well, and avoid confusion with a method used to analyze field data to determine reservoir and completion properties.

After that, there are a lot of things that are not so certain. “It is not the final thinking; we have a lot to learn to bring in as much science and objectivity to type well construction” as possible, Lee said.

His presentation was clear, concise, well-organized, and generated enough questions to make it clear that the committee has some tough work ahead of it.

Those challenges range from recognizing the economic limits on production from high-cost tight reservoirs to the production decline likely when an older well is hit by nearby fracturing.

Lee acknowledged that to create a system that generates sound results based on a reasonable amount of time and data “we will have to have some clever approximations.”

Steps Toward a Type Curve

  • Identify the objective of the type well calculations.
  • Select wells to sample that are in keeping with the objective, with a total of 200 or more.
  • Divide the wells into bins of about 50 based on significant differences.
  • To limit the number of bins, choose ways to normalize well production when there are differences, such as wells with various lateral lengths.
  • Observe good practices to gather individual well data, such as forecasting each well separately rather than grouping wells.
  • Include abandoned rates with zero production rates in the total.
  • For wells with a short production history, forecast future production rather than leaving them out.
  • Validate results by comparing with similar wells not in the sample.

    Goal Oriented

    The goal is a statistically valid prediction of a typical well with an approach based on “what you are trying to do,” he said.

    The process will vary depending on whether the goal is predicting the estimated ultimate recovery for a typical well, a production profile for the first 5 years of a well’s life, or the discounted value of the critical early years of these fast-declining wells.

    Knowing the endpoint will guide the critical work of gathering a statistically significant sample of actual production data—his talk suggested 200 or more wells—sorted into bins of 50 or so wells with common characteristics.

    While the total needed will rise or fall based on the level of variance, if data are available from only a few wells the result will be less reliable.

    “I have seen examples of type wells based on three wells to produce thousands of wells to sell an investment,” Lee said, adding, “If that is all you have, that is all you have. But the uncertainty is just enormous.”

    Wells within each bin need to be grouped using characteristics that “really seem to matter.” Those groups in the large sample are needed to reflect the fact that within “a given resource play, wells in different areas are likely to follow somewhat dissimilar production paths from completion to abandonment,” Lee said.

    While a lot of things can matter, the characteristics used to sort the wells need to be limited to the most significant to make it possible to create bins large enough for reliable statistical analysis.

    A list of key variables could include the well location, spacing, the date of first production, geological similarities, and the operator. The time of the completion and the operator can be significant because “the wells are tending to get better and the best operator is really learning how to do that” (SPE 158867).

    A statistical check is advised to ensure that wells belong in the same bin, Lee said. When plotted the members should generate a log-normal distribution.

    In some cases, differences that significantly affect production, such as horizontal wellbores that are twice as long as others, can be eliminated by finding a common unit of measurement.

    One option for comparing the output from laterals lengths of 5,000 ft and those 10,000 ft and more is the oil and gas production per foot of wellbore. Lee said “that is an oversimplification we need to improve on” because it fails to reflect how the rate changes farther out on really long laterals.

    Lee said it is important to include the failures. The total needs to include the zero-production totals for wells that did not produce. Leaving them out will risk overestimates by introducing a “survivor bias.”

    He also said that with wells that have limited production histories, projections of future output should be included in the data used to calculate the type curve.

    A questioner asked how they would express production lost due to fracture hits. “We must find a way to do that. Frac hits are more common. In older data sets we did not see that, but we are seeing that today,” Lee said.

    Wells from one operator (company 4, purple) outperformed three other companies in four out of five Barnett wells, in two cases by a wide margin. Source: SPE 158867.

    Overly Optimistic

    A common flaw of type wells is that the numbers are higher than the ultimate reality. Some of that is a reflection of the difficulty in creating a methodology to generalize groups of wells, which sometimes vary widely and unpredictably.

    Unconventional development is focused on ultratight rock where the location targeted is critical, as are the skills of those planning and executing the well. This is still young technology where everyone in the business is working to improve the shape of production curves over time. And price changes affect how operators manage wells.

    One questioner pointed out that the cost of maintaining production in these fields tends to remain high. Lingering low prices can force operators to shut in those wells sooner than expected when doing the type curve.

    Lee agreed economic and technical considerations are needed because “a lot of people have had to de-book reserves.”

    And no matter how good the system, there are always going to be overly optimistic type curves where the outcome is shaped by a desire to make a case to investors or sweeten the value of assets for sale.

    “We need to be skeptical of what­ever we are presented,” Lee said, adding, “None of the practices address the problem of a deliberately overoptimistic bias. We can only deal with that by doing our own work.”

    When Monograph 5 becomes available, the hope is to offer help for checking the results.

    “We are not regulators, and we can’t enforce anything,” Lee said, adding, “We hope to be able to provide transparent recommendations that will allow investors, and possibly even regulators such as the SEC [US Securities and Exchange Commission], to ask the right questions of type well creators and determine whether a given type well satisfies reasonable criteria for confident application.”

    For Further Reading

    SPE 137748 An Unconventional Rate Decline Approach for Tight and Fracture-Dominated Gas Wells by A.N. Duong, ConocoPhillips Canada.

    SPE 178525 Methodology for Construction of Type Wells for Production Forecasting in Unconventional Reservoirs by Ayush Rastogi and John Lee, University of Houston.

    SPE 158867 A Practical Guide to Unconventional Petroleum Evaluation by Boyd Russell and Randy Freeborn, Energy Navigator.

    SPE 175967 Creating More Representative Type Wells by Randy Freeborn, SPEE, Energy Navigator, and Boyd Russell, SPEE, Energy Navigator.

    SPE 171658 Entering a Liquid-Rich Shale Play Near the Top of the Learning Curve: Early Quantification and Application of Development Best Practices by David Braun, Brad Powell, Rong Guo, Shell Canada et al.