Thesis Defense College Station, TX (USA) — 05 September 2013 An Integrated Well Performance Study for Shale Gas Reservoir Systems — Application to the Marcellus Shale Landon RISER Department of Petroleum Engineering Texas A&M University College Station, TX 77843-3116 (USA) landon.riser@pe.tamu.edu Outline ●Purpose of the Study: ■Apply modern well/reservoir analysis techniques to field cases. ■Present methods used and challenges encountered in our pursuit. ●Validation of the Study: ■Illustrative cases of non-uniqueness in model interpretations. ■Ramifications of non-uniqueness in long-term performance. ●Rate-Time and Model-Based Production Analyses: ■Initial analyses performed contemporaneously, but independently. ■Integrated analyses based on initial parameter/property correlations. ■Adjustments made to "tune" parameters based on initial correlation. ■Observe effect the "tuning" has on EUR. ●Pressure Transient Analysis: ●Summary & Conclusions: ■Summary of the work done. ■Discussion on the key takeaways from the study. Thesis Defense — Landon RISER — Texas A&M University College Station, TX (USA) — 05 September 2013 Slide — 2/30 ■Illustrative cases with high-frequency bottomhole pressure gauges. ■Cases of daily surface pressures and their potential utility. Purpose of the Study ●Our Primary Objectives: Source: beckenergycorp.com Thesis Defense — Landon RISER — Texas A&M University College Station, TX (USA) — 05 September 2013 Slide — 3/30 ■Present a specialized workflow for modern dynamic data analyses. ■Apply the workflow to production data history of Marcellus shale wells. ■Discuss challenges encountered in unconventional reservoir analysis. ■Demonstrate a correlation/"tuning" concept from analysis integration. ■Address literature void of unconventional PTA with illustrative cases. Figure 1 — Schematic of non-interfering fracture behavior for a horizontal well with multiple vertical fractures. Thesis Defense — Landon RISER — Texas A&M University College Station, TX (USA) — 05 September 2013 Slide — 4/30 The Physical System Validation of the Study ● Issue of Non-uniqueness: ■ We can model a single-well diagnostic with infinite combinations. — (i.e. k, xf, Fc, etc.) Slide — 5/30 ■ Constraint on value ranges is our own scientific intuition. ■ The case shown below serves as a type-well for the region. Thesis Defense — Landon RISER — Texas A&M University College Station, TX (USA) — 05 September 2013 Validation of the Study ● Long-term Performance Ramifications: ■ The ultimate result is reliable EUR values. ■ We can "bound" (or constrain) our EUR predictions using parameters that Slide — 6/30 adhere to results/analogs gathered from independent sources (e.g., core analysis, pre-frac tests, etc.). EUR Variance = 0.36 BSCF (or 24 percent) for this case. Thesis Defense — Landon RISER — Texas A&M University College Station, TX (USA) — 05 September 2013 Thesis Defense College Station, TX (USA) — 05 September 2013 Rate-Time Analysis Landon RISER Department of Petroleum Engineering Texas A&M University College Station, TX 77843-3116 (USA) landon.riser@pe.tamu.edu Rate-Time Analysis ●Rate-Time Concepts: ■ Diagnostic Data — Continuous calculation of loss ratio (D-1) and loss ratio derivative (b). — Qualitative evaluation of characteristic behavior. — Adjust model parameters to match diagnostic data (D and b). ■ Flow Rate Data diagnostics, we shift the initial flow rate (qgi). Thesis Defense — Landon RISER — Texas A&M University College Station, TX (USA) — 05 September 2013 Slide — 8/30 — Upon matching Rate-Time Analysis ●We Used Two "Modern" Rate-Time Relations: ■ Modified Hyperbolic Relation — Adaptation of Arps’ hyperbolic model with an exponential "tail." — Captures early-time hyperbolic decline behavior. — Avoids indefinite extrapolation of early-time behavior. ■ Power-Law Exponential Relation qi D D limit 1/ b q (t ) 1 bDi t ………… Modified Hyperbolic Relation q exp[ D t ] D D limit limit i q(t ) qi exp[ D t Di t n ] ………..… Power-Law Exponential Relation Thesis Defense — Landon RISER — Texas A&M University College Station, TX (USA) — 05 September 2013 Slide — 9/30 — Developed empirically based on observed "power law" behavior. — Provides adequate representation for transient and transition flow. — Conservatively forecasts EUR (serves as a lower bound). Rate-Time Analysis ●Field Case #1 ■ Modified Hyperbolic Relation — We focus on data > 60 days. — Hyperbolic D(t) character. — Relatively constant b(t). ■ Match Parameters qgi = 2029 MSCFD Di = 0.0047 b = 1.9 Dlimit = 10% (default). ■ EUR — 2.88 BSCF Thesis Defense — Landon RISER — Texas A&M University College Station, TX (USA) — 05 September 2013 Slide — 10/30 — — — — Rate-Time Analysis ●Field Case #2 ■ PLE Relation — We focus on data > 20 days. — Power law D(t) and b(t) character. — Excellent qg(t) match. — qgi — Ďi — n — D∞ ■ EUR = 1715 MSCFD = 0.068 = 0.45 = 0 (default). — 1.63 BSCF Thesis Defense — Landon RISER — Texas A&M University College Station, TX (USA) — 05 September 2013 Slide — 11/30 ■ Match Parameters Thesis Defense College Station, TX (USA) — 05 September 2013 Model-Based Production Analysis Landon RISER Department of Petroleum Engineering Texas A&M University College Station, TX 77843-3116 (USA) landon.riser@pe.tamu.edu Model-Based Production Analysis ●Production Analysis Concepts: ■ Diagnostic Plot — Rate-normalized pseudopressure — — calculated continuously. Plotted against te. Diagnostic analog to well testing. — Constant-rate equivalent. ■ Method of Use Thesis Defense — Landon RISER — Texas A&M University College Station, TX (USA) — 05 September 2013 Integral of rate−normalized pseudopressure: 𝑰 𝒕𝒆 = 𝟏 𝒕𝒆 𝒕𝒆 𝟎 𝒎 𝒑𝒊 − 𝒎 𝒑𝒘 𝝉 𝒒𝒈 𝝉 𝒅𝝉 Derivative of the integral of rate-normalized pseudopressure: 𝑰 ′ 𝒕𝒆 = 𝝏𝑰 𝒕𝒆 𝝏𝒍𝒏 𝒕𝒆 Slide — 13/30 — Load pressure and rate histories. — QA/QC. — Extract flow period(s) of interest. — Qualitative evaluation (diagnostics). — Incorporate subsurface data. — Build analytic model(s). — Forecast model(s) to obtain EUR. Model-Based Production Analysis ●Field Case #1 ■ Diagnostic Discussion — Early skin effect (common). — Stabilization @ 100 days, te. — Linear Flow (1/2 slope). — Moderate conductivity fracture. ■ Model Parameters ■ EUR = 260 = 180 =1 = 36 — 1.92 BSCF Thesis Defense — Landon RISER — Texas A&M University College Station, TX (USA) — 05 September 2013 nD ft md-ft (# of fractures) Slide — 14/30 — k — xf — Fc — nf Model-Based Production Analysis ●Field Case #2 ■ Diagnostic Discussion — Very similar to Case #1. — Noisier data (operations issues?). — Stabilization @ 200 days, te. — Moderate conductivity fracture. ■ Model Parameters = 230 nD = 100 ft = 0.42 md-ft = 36 (# of fractures) ■ EUR — 1.41 BSCF Thesis Defense — Landon RISER — Texas A&M University College Station, TX (USA) — 05 September 2013 Slide — 15/30 — k — xf — Fc — nf Model-Based Production Analysis Relative Analysis Exercise: "Normalized" Data Plot Vertical Shift Factor = 1.7 (increasing permeability) Horizontal Shift Factor = 1.05 (increasing flux area) Thesis Defense — Landon RISER — Texas A&M University College Station, TX (USA) — 05 September 2013 Slide — 16/30 Raw Data Plot Thesis Defense College Station, TX (USA) — 05 September 2013 Integration of Rate-Time Analysis and Model-Based Production Analysis Landon RISER Department of Petroleum Engineering Texas A&M University College Station, TX 77843-3116 (USA) landon.riser@pe.tamu.edu Integration and Correlation of Well/Reservoir Metrics ●The Workflow: ■ Independently analyze rate-time data with modern rate-time relations — Power-Law Exponential and Modified-Hyperbolic relations. — Model based on the D- and b-parameter behavior (diagnostic). — Tabulate model parameter results. ■ Independently analyze pressure-rate-time data with analytical models — Inspect the pressure-flowrate relationship for consistency. — Evaluate the diagnostic response from RNP output. — Create analytical well models that represent the data. — High-quality flowrate data with minimal interruptions is crucial. — Constrain the integration to the wells with the highest quality data. — Crossplot model results from rate-time with well/reservoir analysis. — Iteratively refine initial correlations by imposition. — Observe resultant change in correlation(s). Thesis Defense — Landon RISER — Texas A&M University College Station, TX (USA) — 05 September 2013 Slide — 18/30 ■ Combine the key results from the two analyses Integration and Correlation Correlation of Modified Hyperbolic b(t); and k from Diagnostic Plot: b-parameter b = 2.4 Thesis Defense — Landon RISER — Texas A&M University College Station, TX (USA) — 05 September 2013 correlate k = 170 nD Slide — 19/30 k from derivative Integration and Correlation ● Tuning Exercise: ●Concept: ■ Based on idea of interrelatedness of flow properties and decline parameters. — Rate-decline a function of — pressure distribution. Pressure distribution according to rock/formation properties. ●Process: etc.) accordingly to obtain new match. ■ Re-forecast updated model for new EUR value. ■ Observe changes in updated EUR correlation. Thesis Defense — Landon RISER — Texas A&M University College Station, TX (USA) — 05 September 2013 Slide — 20/30 ■ Crossplot k and hyperbolic b(t). ■ Tune k values to linear trend. ■ Adjust flow properties (xf, Fc, Integration and Correlation ●EUR Crossplot: ●Graphical Observations: ■ We observe a >1:1 relationship. ■ R-squared value = 0.78. ●Conceptual Comments: ■ Pre-tuning R-squared value on the order of 0.6. ■ Error increases with increasing model-based EUR. ■ Slope or intercept adjustment most appropriate model? ●Hypothesis: to initial flow rate (qgi). ■ Decline character could be captured, but area-under-the-curve impacted by erroneous initial point. Thesis Defense — Landon RISER — Texas A&M University College Station, TX (USA) — 05 September 2013 Slide — 21/30 ■ Rate-time EUR values proportional Integration and Correlation ●EUR Histogram (PA and Rate-Time) ■ Alternate Graphic to Correlation Plot time. ■ Bin Selection — "Like" binning for comparison. — Manipulative binning could produce more similar continuous curve (w/ offset). ■ Conundrum — We’re still left uncertain precisely why rate-time analysis consistently overestimates EUR w.r.t. model-based forecasting. Thesis Defense — Landon RISER — Texas A&M University College Station, TX (USA) — 05 September 2013 Slide — 22/30 — Pseudo-Gaussian distribution. — Narrower range for PA. — Two "outlier" EURs from Rate- Thesis Defense College Station, TX (USA) — 05 September 2013 Pressure Transient Analysis Landon RISER Department of Petroleum Engineering Texas A&M University College Station, TX 77843-3116 (USA) landon.riser@pe.tamu.edu Pressure Transient Analysis ●Brief Rundown: ■ Challenges faced in pressure transient analysis in shale reservoirs — Non-uniqueness — Expense (in terms of money and time) — Technology ■ Benefits realized from PTA — Independent source of information. — Confirmation of model parameters from production analysis. ■ What follows Slide — 24/30 — An illustrative example of a traditional pressure buildup test. — Discussion of potential use of daily surface pressure data. — Demonstration of static and dynamic flow dichotomy. Thesis Defense — Landon RISER — Texas A&M University College Station, TX (USA) — 05 September 2013 Pressure Transient Analysis ●26 Day Buildup Test ●Diagnostic Attributes: ■ Half-slope (High FcD). ■ Minimal Wellbore Storage. ■ Minimal skin effect. ●Model: obtain match. ■ Requires lower xf, but greater Fc (than PA) to obtain match. ■ This is a common theme: — We observe higher conductivity response during shut-in than in drawdown. Thesis Defense — Landon RISER — Texas A&M University College Station, TX (USA) — 05 September 2013 Slide — 25/30 ■ Modeled with k from PA. ■ Adjusted xf, Fc, and skin factor to Pressure Transient Analysis ●The Case for Daily Surface Pressure ■ Surface Pressures Overlay — Both derivative and pressure drop ■ For Dry Gas — Pressure drop largely conserved — Liquid dropout a non-issue ■ Qualitative/Quantitative — If we don’t feel comfortable Slide — 26/30 modeling surface buildups, we can potentially benefit from diagnostics (qualitative). Thesis Defense — Landon RISER — Texas A&M University College Station, TX (USA) — 05 September 2013 Pressure Transient Analysis ● Buildup – Drawdown Dichotomy: ●Diagnostic Dichotomy: ■ Half-slope (1/2) Buildup. ■ Quarter-slope (1/4) Drawdown. ■ Minimal skin effect. ●Fracture Behavior ■ All buildups display linear flow (1/2). — High fracture conductivity ■ Most drawdowns are bilinear (1/4). — Low (finite) conductivity appreciably on effective stress? ■ How can we account for this dichotomy? ■ What are the long-term implications of a stress dependent conductivity? Thesis Defense — Landon RISER — Texas A&M University College Station, TX (USA) — 05 September 2013 Slide — 27/30 ■ Does fracture flow depend Thesis Defense College Station, TX (USA) — 05 September 2013 Summary and Conclusions Landon RISER Department of Petroleum Engineering Texas A&M University College Station, TX 77843-3116 (USA) landon.riser@pe.tamu.edu Summary and Conclusions ●Summary: ■ Performed independent production data and rate-time analyses. ■ Integrated the two analyses with an iterative correlation scheme. ■ Discussed challenges in unconventional well performance analysis. ■ Presented a workflow that attempts to reduce non-uniqueness. ■ Introduced PTA as an analysis tool in unconventional reservoirs. ●Conclusions: ■ From this work we conclude the following: character for our 55-well data set. — PLE relation produces the most conservative EUR estimates. — Bilinear flow (1/4 slope) is the predominant flow regime. — Linear flow (1/2 slope) is the exclusive PTA diagnostic. — Correlation scheme using a "tuning" technique improved the EUR relationship between model-based and rate-time analyses. — Model-based production analysis is an effective tool for cases of erratic production history, while rate-time analysis requires smooth, lightly-interrupted flow periods. Thesis Defense — Landon RISER — Texas A&M University College Station, TX (USA) — 05 September 2013 Slide — 29/30 — Rate-time diagnostics exhibit primarily hyperbolic decline Thesis Defense College Station, TX (USA) — 05 September 2013 An Integrated Well Performance Study for Shale Gas Reservoir Systems — Application to the Marcellus Shale Landon RISER Department of Petroleum Engineering Texas A&M University College Station, TX 77843-3116 (USA) landon.riser@pe.tamu.edu