SPE 135607 A Comprehensive Workflow for Early Analysis …€¦ · SPE 135607 A Comprehensive...

SPE 135607

A Comprehensive Workflow for Early Analysis and Interpretation of Flowback Data From Wells in Tight Gas/Shale Reservoir Systems D. Ilk, Texas A&M University/DeGolyer and MacNaughton; S.M. Currie, Texas A&M University/Devon Energy Corp.; D. Symmons, Consultant; J.A. Rushing, Anadarko Petroleum Corp.; N.J. Broussard, El Paso Corporation; and T.A. Blasingame, Texas A&M University

Copyright 2010, Society of Petroleum Engineers This paper was prepared for presentation at the SPE Annual Technical Conference and Exhibition held in Florence, Italy, 19–22 September 2010. This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents of the paper have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect any position of the Society of Petroleum Engineers, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract

Flowback data are defined as the early time-pressure-rate (TPR) data which are taken immediately following the completion and stimulation of a given well. Although typically discounted or even ignored, this data has a wealth of well/reservoir information. Analysis can provide forward-looking indications of long-term well deliverability as well as an early diagnosis of any potential production problems. Effective analysis of early TPR data, however, requires the utilization of all data sources and data types — for example: phase rates, pressures, choke settings, sand production, chlorides content, etc.

The purpose of this paper is to provide a comprehensive workflow for early-time flowback data which provides a unique visualization of these data, as well as a correlative and integrated analysis of these data (i.e., comparing different data types and functions, correlating one data type with another, etc.). The major steps in a generic "flowback" workflow include:

● Collection and quality control of TPR and other well performance/completion data. ● Construction/calibration of a base well/reservoir model (including well completion and stimulation parameters). ● Construction/application of specialized plots to identify features in the data at a very basic level (i.e., water unloading). ● Correlation of flowback data using empirical and non-parametric models. ● Extrapolation of parametric and non-parametric models to provide predictive estimates of early-time well performance. ● Implementation of an assessment loop to guide future production/completion practices.

The proposed methodology is demonstrated using case studies of wells completed in tight/shale gas reservoir systems. Extensive discussion is provided for the analysis/interpretation of flowback data; as well as our efforts to identify and quantify factors which affect and control production. The approach used in this work is that of a "dashboard" of (static) diagnostic plots as well as correlations/plots for multiple-well sets of flowback data. While the focus of this workflow is the diagnostic analysis of flowback data, we also provide guidance towards the use of flowback and long-term well performance data in "model-based" analyses.

Introduction

The concept of "flowback" is quite old — historically many (if not most) wells were "managed" during early production as to not damage the reservoir or the wellbore tubulars. In very recent times the focus on tight gas/shale gas resources has lead to extremely large (and expensive) well stimulation treatments (often one-half of the well cost), and as such, the process of bringing a well onto production has attained something of an art form — a balance of trying to maximize initial performance, but also to begin production carefully, so as not to damage the well and/or impair long-term production.

In this paper we focus on the analysis/interpretation of the process of flowback, not the objectives or the potential ramifications of poor flowback practices. In particular, our model-based analyses may provide an assessment of damage, but we do not pursue quantification of geomechanical effects (e.g., pressure-dependent permeability or proppant embedment), nor do we propose specific flowback design procedures, although we do provide some "holistic" guidelines, designed to ensure that the data obtained from a flowback procedure is relevant and can be analyzed/interpreted.

2 D. Ilk, S.M. Currie, D. Symmons, J.A. Rushing, N.J. Broussard, and T.A. Blasingame SPE 135607

Fig. 1 — Summary of wells considered in this flowback study (shale gas cases).

SPE 135607 A Comprehensive Workflow for Early Analysis and Interpretation of 3 Flowback Data From Wells in Tight Gas/Shale Reservoir Systems

Fig. 2 — Correlations for Well B flowback data.


In Fig. 1 we present a summary of the 5 wells used in this study — although these cases were selected for their consistency, we note that a "universal" correlation of the 5 wells (in combination) was not achieved, due in part (we believe), to the nature of the flowback data. Although the trends shown in Fig. 1 appear to be quite consistent, we suspect that there are some discrepancies in the data (perhaps the accuracy of the water rate data, as a possible example). As a result of our failure to achieve a universal correlation, we proceeded to create a series of correlations for "Well B" — where we note that Well B was selected because of the appearance of the data (it appears to be the most "correlated" of our data cases).

In Fig. 2 we present the correlation of data for Well B, and (somewhat) to our surprise, none of the matches are much better than "average," and the non-parametric (statistical) match suggest that the data are poorly correlated (on a per variable basis). Regardless of the correlation, this particular example case is valuable as a demonstration model — the data functions are relatively well-behaved and none of the correlations is uniquely outstanding (as it would be in practice). While the proposed correlations do tend to capture major features/events in the data, we are equally concerned about the relatively weak matches of certain features, but as it turns out, we believe that the water rate is the most likely culprit.

Field Example — Correlation of Well Performance Data

In this section we provide orientation and discussion regarding the workflow for the interpretation and analysis of flowback data — in particular, as we provide a demonstration using field data acquired from a modern shale gas play. As we noted earlier, our goal is the diagnostic process — not analysis leading to reservoir properties or reservoirs, that is another workflow.

In Fig. 3 we provide plots of the gas-water ratio (GWR) versus cumulative gas production (Gp) for each of the wells in this study, as well as a combined plot (a), which suggests the similarity of this process for all of these wells. There are apparent regimes of interest (a half-slope fracture depletion trend appears to be evident for several of the cases. In Fig. 4 we present the gas normalized productivity index (GNPI) versus cumulative gas production (Gp) and while we note significant "cleanup" trends in the early time data. This is expected. We also note in Fig. 4a a very strong trend where all of the wells have essentially the same "recovery" profile. We again note apparent 1/2-slope trends on several of the plot functions. Similarly, in Fig. 5 we present crossplots for individual wells in terms of the water normalized productivity index (WNPI) versus cumulative water production (Wp). There is no clear model trend in Fig. 5, and the "depletion" of the water is quite rapid.

In Fig. 6 we consider the individual well cross-plots of casing pressure (pcf) versus gas flowrate (qg), and we note with the possible exception of Well B (at that time), that all of the wells have "turned" down into the tubing/casing lift curve trend (i.e., the rollover feature). This is important because one can suggest that the dominance of water loading has diminished. In Fig .7 we provide the crossplots of casing pressure (pcf) versus water flowrate (qw) — and the story here is similar, we note that the tail for these trends are tending away from water production being dominant. In Fig. 8 the gas flowrate (qg) is plotted versus water flowrate (qw) in an attempt to assess the water rate/oil rate inter-relation, although admittedly, this rendering is not particularly effective, a better rendering of these functions is to consider the GWR function (see Fig. 3).

The crossplots of choke setting versus production time (t) are shown in Fig. 9, and there is little to discuss other that to suggest that the operator utilize a consistent schedule for choke settings. This is not necessary, but following a prescribed flowback schedule will help to diagnose and (possibly) analyze flowback data in more detail. In Fig. 10 we present the individual well crossplots for cumulative gas production (Gp) as a function of cumulative water production (Wp) to illustrate the early dominance of the water production.

Fig. 11 provides our first sequence of multi-well crossplots — these are basic crossplots; rate and cumulative functions versus time, as well as normalized production functions versus production time. The cumulative gas and water functions are of interest, as is the GNPI function versus time. In Fig. 12 we present the "multi-well crossplots" for gas and water productivity index functions versus production time (t). The primary observation in Fig. 12 is that all of the wells in our base review have effective cleanup trends. Our last sequence of crossplots is shown in Fig. 13, where we present multi-well crossplots of gas and water productivity index functions versus gas and water cumulative production. As with the productivity indices plots versus time, these plots (versus cumulative production) also confirm the well cleanup (and "rollover") to the expected trends for reservoir flow behavior.


Fig. 3 — Individual well crossplots: gas-water ratio (GWR) versus cumulative gas production (Gp).


Fig. 4 — Individual well crossplots: gas normalized productivity index (GNPI) versus cumulative gas production (Gp).


Fig. 5 — Individual well crossplots: gas normalized productivity index (GNPI) versus cumulative water production (Wp).


Fig. 6 — Individual well crossplots: casing pressure (pcf) versus gas flowrate (qg).


Fig. 7 — Individual well crossplots: casing pressure (pcf) versus water flowrate (qw).


Fig. 8 — Individual well crossplots: gas flowrate (qg) versus water flowrate (qw).


Fig. 9 — Individual well crossplots: choke setting versus production time (t).


Fig. 10 — Individual well crossplots: cumulative gas production (Gp) versus cumulative water production (Wp).


Fig. 11 — Multi-well crossplots: various plots/various functions.


Fig. 12 — Multi-well crossplots: gas and water productivity index functions versus production time (t).


Fig. 13 — Multi-well crossplots: gas and water productivity index functions versus gas and water cumulative production.


Summary, Conclusions, and Recommendations

Summary: In this paper we introduce the concept of correlating the flowback behavior of tight gas/shale gas wells in the form of a gas rate predictive relation. Such a relation could, in theory, be used to design a flowback procedure — obviously, we advise consistency in the flowback process as such correlations may not tolerate severe rate or pressure changes. We also propose and validate as sequence of single-well and multi-well "dashboard" plots for the qualitative assessment of the flowback process — again, this analysis/interpretation is not "quantitative" in the sense of yielding formation properties (that requires a separate, model-based analysis process). This process does yield a qualitative comparison of performance — in particular, a comparison of well performance on a well-by-well basis, based on a broad range of performance functions (time-rate-pressure (or TPR) functions). Again, in this paper, our goal is process and procedure; not estimation of reservoir properties.

We have successfully illustrated our flowback evaluation procedure using 5 (five) wells from the same shale gas play, which have similar flowback profiles. We have created a "dashboard" for each of the 5 wells as well as a common "multiwell" dashboard. We propose and encourage this practice as a "first stage" analysis.

Conclusions:

We state the following conclusions based on this work:

1. We have demonstrated the parametric correlation of gas rate performance with water rate, casing pressure, and the choke history (our "Well B" example). Although this particular case did not correlate uniquely, this case does serve as a relevant example, especially in a "lessons learned" sense. And in fairness, we are not trying to correlate these data to estimate reservoir properties, but rather, to test the methodology.

2. We have also successfully demonstrated our "dashboard" plot approach. In this work we utilized the same field data cases (i.e., the 5 shale gas wells). We demonstrated the correlation of time-rate-pressure (or TPR) data via the following plots.

Base Plots: Single Well Plots: Multi-Well Plots: ● qg vs. t. ● GWR vs. Gp. ● pcf vs. t. ● qw vs. t. ● GNPI vs. Gp. ● qw vs. t. ● pcf vs. t. ● WNPI vs. Wp. ● qg vs. t. ● Choke setting vs. time. ● pcf vs. qg. ● Gp vs. t. ● pcf vs. qw. ● GNPI vs. t. ● qg vs. qw. ● Wp vs. t. ● Gp vs. Wp. ● GPI vs. t. ● WNPI vs. t. ● WPI vs. t. ● GNPI vs. Gp. ● GPI vs. Gp. ● WNPI vs. Wp. ● WPI vs. Wp.

3. We have clearly shown the diagnostic value of accurate and consistent measurements of rates and pressures as a function of time. All of the diagnostic functions require these data, and while our approach is forgiving in terms of tolerance for data functions containing random noise, our approach (and any approach) will have issues with data which poorly measured and/or not taken in a proper synchronous manner. We must emphasize that good diagnostics require good data.

Recommendations:

The missing "ingredient" in the analysis of flowback data — in particular, the characterization of early-time flowback data which are dominated by water unloading is that of a model that can accurately and effectively represent the interaction between the well and reservoir. Our (admittedly weak) concept model is analogous to that of wellbore storage — i.e., a compressible volume of fluid which controls the interaction of the well and the reservoir, at least during early well performance. Future work should consider analytical/semi-empirical models for the "water unloading" phase of the flowback profile.


Nomenclature

GNPI = (pci2-pcf

2)/qg, psi2/MSCF/D GPI = qg/(pci

2-pcf2), MSCF/D/psi2

Gp = Cumulative gas production, MSCF pcf = Casing pressure, psig (typically) pci = Initial casing pressure (at t=0), psig (typically) qg = Gas production rate, MSCF/D qw = Water production rate, STB/D t = Production time, days WGR = Water-Gas-Ratio, STB/MSCF WNPI = (pci

2-pcf2)/qw, psi2/STB/D

Wp = Cumulative water production, STB WPI = qw/(pci

2-pcf2), STB/D/psi2

SPE 135607 A Comprehensive Workflow for Early Analysis …€¦ · SPE 135607 A Comprehensive...

Documents

Transcript of SPE 135607 A Comprehensive Workflow for Early Analysis …€¦ · SPE 135607 A Comprehensive...