Building Type Wells for Appraisal of Unconventional Resource Plays

You have access to this full article to experience the outstanding content available to SPE members and JPT subscribers.

To ensure continued access to JPT's content, please Sign In, JOIN SPE, or Subscribe to JPT

Although the application of statistical techniques to type wells is gaining acceptance, it is often unclear to evaluators how these techniques can be applied to capture accurately the full range of uncertainty in the average single-well estimated ultimate recovery (EUR) for a geologic subset. The objective of the complete paper is to present an integrated work flow that can be used to build P90, mean, and P10 type wells, which represent the range of potential outcomes for the geologic subset in an unconventional resource play.

Introduction

A common challenge that accompanies new technologies dedicated to the discovery of unconventional resources is how to forecast production and quantify EUR. Early in the life of a resource play, it can be difficult to build type wells because of limited production history and a small well count.

Traditional methods would use an analogous-field well model or decline methods to predict future production. Because of unconventionals being a relatively recent development, no late-life fields exist that can be used as direct analogs to understand mid- to late-time ­horizontal-well behavior in tight unconventional formations. For plays in the early stages of development, because of the relatively small well count and difficulty with a direct analog, the early-time well behavior is also not easily predicted with confidence. There is thus a high degree of uncertainty in both the shape and the magnitude of the type-well profile. Consequently, it is becoming more common for management to ask for an expected type well with a range to capture uncertainty, rather than a single ­deterministic estimate. The work flow in this paper applies such methods.

Methodology

Acknowledging Uncertainty. For unconventional resource plays, the two basic sources of uncertainty are the drilling-and-completion (D&C) design and the geological properties that characterize the reservoir. Ideally, one would select a statistically significant number of wells with identical (or nearly identical) geological properties, completions, lateral length, and drilling azimuth to construct type wells. However, this ­scenario is often far from reality.

One solution is to wait until enough wells exist with nearly identical D&C designs and geological properties before proceeding with type-well construction. Obviously, this solution is not practical if management needs to rank assets in the portfolio and justify capital allocation for development of some assets but not others. Therefore, alternative solutions to deal with varying D&C designs and geological properties are to normalize production data for D&C design and to define geologic subsets for areas with similar geological properties.

For the purposes of discussing the issue, one can assume that, after normalizing production data and defining geologic subsets, one has identical well designs and nearly identical geological properties. “Nearly identical geological properties” means that, along the length of a horizontal well, there is significant heterogeneity in geological properties but, within a geologic subset, there is enough homogeneity in the regional geological properties for the population of wells to behave consistently enough to be represented by a statistical distribution.

This consistency in behavior within a geologic subset is quite useful for understanding and predicting production in unconventional resource plays. Not surprisingly, it is also why many evaluators refer to an unconventional resource play as a “statistical play.” Unfortunately, the use of the term “statistical play” can cause serious problems in understanding the real range of uncertainty in future production.

Probably the most serious problem in evaluating so-called “statistical plays” revolves around defining the statistical distribution used to represent the play. For unconventional resource plays in an early phase of development, when very few wells exist, even though we can calculate a mean of the existing data, there is still uncertainty around the mean of the true distribution. If that uncertainty is not acknowledged, evaluators may run into problems when they apply statistical aggregation principles to booking proved-undeveloped locations in reserves.

Understanding the Impact of Geological and D&C Design Uncertainties. A useful method to understand the effect of such sources of uncertainty is to test different scenarios with a reservoir model and observe how changing geological or D&C parameters in the model affect the decline equation that can be used to model the production curve.

One of the more critical (and uncertain) parameters to consider is permeability. Even if an evaluator has defined geologic subsets with similar geological parameters, there can still be orders of magnitude of variation in permeability within that region. Higher permeability in the enhanced fracture region is a key driver for early time in this analytical model and will result in higher initial production rates and shorter times to the end of linear flow. The effect of permeability becomes even more apparent if one considers a liquids-rich fluid system, where multiphase flow and relative permeability come into play. Dropping below the saturation pressure in the reservoir will result in multiphase flow, and changes in the relative permeability of the different fluid phases. These changes in relative permeability can cause a change in flow regime observed by creating the appearance of a boundary caused by the reduction in permeability. Relative permeability effects can be significant, but, unfortunately, current technology creates limitations because it is extremely difficult to measure or calculate relative permeability curves reliably for low-permeability unconventional resource plays.

It is well-understood that volumetric parameters such as pay, porosity, and well length do contribute to well-­productivity differences. However, within the typical ranges of uncertainty for these variables, their effects are less significant than those of permeability and D&C design. If geological parameters are held constant, the D&C design can also affect the production forecast dramatically in low-permeability unconventional resource plays.

Unfortunately, while a well can be designed to achieve a particular lateral length, number of fractures, and fracture half-length, one cannot reach a high level of certainty as to whether goals have been achieved. Therefore, in practice, there is no perfect way to normalize production data between wells with different D&C designs. However, on the basis of the authors’ experience, normalizing production data by stimulated lateral length (SLL) is a simple method that does a reasonably good, though imperfect, job because many of the other important variables are positively correlated with SLL.

Building Uncertainties Into Type Wells

The approach proposed in the complete paper is to use a multisegment Arps decline and to apply ranges of uncertainty to different parameters in the Arps equation. These ranges of uncertainty can then be combined in a Monte Carlo or risk-tree analysis to build a distribution of production profiles, and this distribution of production profiles can be used to construct P90, mean, and P10 type wells.

In this instance, the authors propose a three-segment Arps equation to forecast the primary fluid, where the first segment represents linear flow, the second segment represents boundary-­influenced flow (i.e., the extent of the stimulated reservoir volume has been reached but there may still be matrix contribution), and the third segment represents boundary-dominated flow.\

Putting It All Together. The authors’ proposed work flow is defined by the following steps, each of which are detailed in the complete paper.

  1. Define the geologic subset.
  2. Normalize production data of analog wells.
  3. Statistically aggregate normalized production data to estimate initial production-rate range.
  4. Assess the range of uncertainty in late-time production behavior.
  5. Assess biases in the data used to estimate initial production rates and late-time production.
  6. Combine ranges of uncertainty in initial production with uncertainty in late-time production.
  7. Build initial estimates of P90, mean, and P10 type wells.
  8. Compare type-well EURs to hydrocarbons initially in place (HIIP).
  9. Iterate to refine uncertainty estimates if necessary.
  10. Evaluate key uncertainties that drive economic value.

While following this work flow, it is critical for evaluators with expertise in different areas to work together to identify biases and to arrive at reasonable assessments of uncertainty. Involving multiple evaluators (including evaluators from different asset teams) will result in type wells that capture the range of possible outcomes better and helps to provide consistency in assessments across an organization with several unconventional assets.

Conclusions

Relative to previously published methodologies, this work flow attempts to remove some of the bias that is introduced to EUR estimates by use of ­decline-curve analysis (DCA) alone. The methodology in this paper uses actual normalized well-performance data (6‑month cumulative) to establish early profile behavior.

For a given geologic subset in an unconventional resource play, the authors believe that this work flow allows for a more-realistic definition of the uncertainty range for the average single-well EUR. The following conclusions are ­offered in regard to the work flow:

  • The range of uncertainty for a geologic subset is likely larger than that estimated by analyzing existing production data with DCA alone to generate a range of type wells.
  • Production data should be normalized, and offset wells must be selected with care to ensure applicability of analog data.
  • Aggregation provides a more-realistic indication of the average well-results range for the play.
  • Type wells should be compared with HIIP to ensure consistency with volumetrics and dynamic flow behavior for a given reservoir quality and fluid system.
  • Key type-well uncertainties that drive variation in economic value can be identified and understood, which highlights the uncertainties that should be derisked in an appraisal and data-gathering program.
This article, written by JPT Technology Editor Chris Carpenter, contains highlights of paper SPE 185053, “Building Type Wells for Unconventional Resource Plays,” by P. Miller, N. Frechette, and K.D. Kellett, Repsol, prepared for the 2017 SPE Canada Unconventional Resources Conference, Calgary, 15–16 February. The paper has not been peer reviewed.

Building Type Wells for Appraisal of Unconventional Resource Plays

01 December 2017

Volume: 69 | Issue: 12

STAY CONNECTED

Don't miss the latest content delivered to your email box weekly. Sign up for the JPT newsletter.