1. Development of an Advanced Approach for Next-Generation, High-Resolution, Integrated Reservoir Characterization Performed by: Advanced Resources International Houston, Texas September 9, 2003 Tulsa, Oklahoma DOE Award No. DE-FC26-01BC15357

2. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 1 Presentation Outline  Background and Project Description  Data Availability and Preliminary Processing  Model Building  Accomplishments & Next Steps

3. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 2 The Industry Need  Effective oilfield management requires 3-D, high resolution reservoir characterization to identify reservoir heterogeneity.  SOR/IOR  EOR  CO 2 sequestration Surface seismic is the most cost-efficient method to obtain inter-well volumetric reservoir information, but vertical resolution is insufficient (>50 ft) for optimized injection management. Surface seismic is the most cost-efficient method to obtain inter-well volumetric reservoir information, but vertical resolution is insufficient (>50 ft) for optimized injection management.

4. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 3 ARI’s Technology Development Goals  Using existing data acquisition capabilities, improve vertical resolution and reduce uncertainty of reservoir characterization.  Present result in engineering terms required for flow modeling and performance forecasting ( , k).  Reduce time/cost requirements.  “Better, faster, cheaper”.

5. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 4 Approach  Apply advanced pattern recognition technologies (artificial neural networks) to integrate multi-scale data (cores, logs, seismic).  Simple  Data-Driven  Deterministic  Generate high-frequency reservoir description at each 3-D seismic trace location.  Incorporate intermediate-scale data (crosswell seismic) to bridge resolution gap and reduce uncertainty.

6. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 5 Proposed Pathway to High-Resolution Reservoir Characterization Borehole Seismic (X-Well, VSP) Conventional Well Logs 3-D Surface Seismic High-Resolution Reservoir Description (Core, MRI, etc.) Objective Model #3 Model #2 Model #1

7. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 6 Benefits over Inversion  Directly predicts permeability in addition to porosity (more robustly than  /k relationship).  More deterministic outcome (not series of equi-probable outcomes).  Simple & fast.  Data-driven modeling, not analytic modeling.

8. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 7 Project Objective  Demonstrate and validate the integrated (virtual intelligence) procedure at a single field.

9. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 8 Analytic Work Flow Acquire and QC Data 3D Seismic X-well, VSP Log Core Well Completion Production Clustering Logs (Seismic) (Log/seismic) (Log/core) Seismic Processing Depth/time 3D/borehole Attribute extraction Rock Physics Modeling Critical attributes Engineering Model Log/core , k Broadband Transform Function 3D/X-Well/Log Validation Simulation? Statistical? Uncertainty analysis? Acquire Site McElroy Field ChevronTexaco

10. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 9 Presentation Outline  Background and Project Description  Data Availability and Preliminary Processing  Model Building  Accomplishments

11. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 10 McElroy Field, West Texas Seismic Survey Crosswell Profiles Study Area N

12. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 11 Summary of Data Received Data TypeAmount Available Well Locations192 Seismic Survey, 3D P-P2.5 sq. mi. (2000, post-steam) migrated stacked time Crosswell Profiles8 crosswell profile data files (1997, pre-steam) Well LogsComplete modern log suites for 59 wells (1984 – 2001) Sonic Logs84 sonic logs over survey area Formation TopsInterpreted formation tops (5) in 150 wells Image Logs8 image log files within the survey area Core LogsCore analysis logs for 13 cored wells in survey: approx. 325 ft. of whole core each with core porosity, saturation, and permeability measurements on ½ foot intervals.

13. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 12 Sample Surface Seismic Line ~ 2.25 mi. W E Inline 165

14. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 13 Comparison of Surface & X-well Data ~ 430 ms. ~ 457 ms. X-line 1295 X-well Profile S N S N Illustrates need for intermediate-scale crosswell data!

15. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 14 Three Main “Data Processing” Elements  Rock Physics Modeling  Seismic Processing & Attribute Extraction  Log Clustering

16. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 15 Rock Physics Modeling - Objectives  Identify and prioritize seismic attributes most likely to be influenced by reservoir properties of interest ( ,seismic facies thickness).  Results will be used to select attributes to include in seismic models.  See Topical Report.

17. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 16 Rock Physics Workflow  Build stratigraphic models of varying vertical resolution based on McElroy data  Conduct seismic experiment  Generate suites of synthetic seismic as investigation layer properties vary  Lithology (over range encountered at McElroy)  Fluid content (saturations, type)  Seismic Facies Thickness (over range encountered at McElroy)  Porosity  Use results to guide selection of seismic attributes most affected by reservoir parameters of interest at McElroy.

18. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 17 Example Seismic Facies Models Model 1 Model 12 Model 8 Model 17Model 4 3 Layers 23 Layers 50 Layers 114 Layers 250 Layers

19. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 18 Ranking of Attributes Affected by Biot-Gassmann Layer Thickness

20. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 19 “High-Priority” Attributes  Trace Differentiation  Hilbert Transform (complex part of the analytic tract)  Perigram (zero mean of the complex amplitude of the trace)  Cosine of Phase (cosine of the instantaneous phase)  Perigram * Cosine of Phase (product of these two attributes)  Instantaneous Phase  Instantaneous Frequency (time derivative of instantaneous phase)  Median Smoother (3 point)  Absolute Value of Trace  Response Phase (instantaneous phase at the trace envelope peaks in degrees)  Seismic Amplitude

21. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 20 Seismic Processing and Attribute Extraction - Objectives  Data QC, time/depth conversion, tie (collocate) surface & X-well traces.  Calculate and extract “high priority” attributes from depth seismic.  See Topical Report.

22. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 21 Seismic Data QC  Two datasets to reprocess  2000 vintage surface P-P seismic cube  ~2.5 square mile reflection seismic survey  176 inlines, 176 crosslines  55 ft. bin spacing CMP gathers  Central frequency 65 Hz.  Zero to 2 sec. data @ 2 ms. sampling  1997 vintage X-well surveys  8 Crosswell seismic surveys  Lengths from 443 ft to 758 ft. in six surveys  Two surveys approx. 1,250 ft. very poor quality  Shot depths approx 1700-2950 ft.  Receiver depths 2200-2900 ft.  Sample interval 0.15 ms.  Varying Data Quality

23. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 22 Seismic Frequency Spectra  Surface Seismic  Good S/N 20 to 100 Hz  Implied vertical resolution ~40 ft.  Crosswell Seismic  Source Sweep 100-2,000 Hz.  Signal to Noise as high as 20 db.  Falloff above 2000 Hz. – Poor data quality  Implied vertical resolution ~ 5 ft.  Overall Good Quality Surface Data  Signal to Noise Difficult in Crosswell Ten Traces0 Freq. Hz 3000 Normalized Signal Db Ten TracesFreq. Hz03000 Normalized Signal Db S/N Ratio Poor Above 2 KHz.

24. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 23 Crosswell Data  Most data of poor quality.  Tube-wave noise  Two profiles suitable for analysis.  Well DY0386 – B03826 (CM319 – CM423)  Well B03826 – DY4441 (CM423 – CM314) DY0386 B03826 DY4441

25. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 24 Clustering - Objectives  Identify trends in log data; adds an important input parameter to ANN analysis.  Facies definition  To be used in log-core model.  See Topical Report.

26. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 25 The Log Clustering Process  Clustering uses a technique known as Self Organizing Maps, or unsupervised neural networks  The object is to group data with similar characteristics into bins, or “clusters”  Determining what each cluster represents (e.g., facies) requires a priori knowledge  Overlap can exist between adjacent clusters

27. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 26 Frequency Distribution Curves for Logs

28. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 27 Example Multi-Dimensional Crossplot

29. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 28 Clustering Facies Type Codes

30. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 29 Comparison to Core Data DY0534 (CM315C) Facies Porosity Permeability

31. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 30 Presentation Outline  Background and Project Description  Data Availability and Preliminary Processing  Model Building  Accomplishments & Next Steps

32. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 31 Three Models  Log-to-Core (complete)  X-well-to-Log (underway)  Surface-to-X-well (underway)

33. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 32 Log-to-Core Model  +/- 6500 datapoints; (10 wells x 325 ft (@ ½ ft increments).  6 logs as input; core porosity and permeability as output.  Applied “depth windowing” to account for uncertainty introduced by different sample intervals.  Use 60% data for training, 20% for testing, 20% for validation.  See Topical Report.

34. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 33 28-14-2 Network Architecture Inputs 6 logs x 3 depth windows 10 fuzzy facies codes

35. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 34 Actual vs. Predicted Crossplots for Porosity & Permeability (Training/Testing Data) Results show model tends to “smooth” extreme values. Results show model tends to “smooth” extreme values. This is an expected outcome. This is an expected outcome. Unit Slope

36. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 35 Actual vs. Predicted Porosity & Permeability Logs (Training/Testing Data) “Smoothing” Porosity/ Permeability “Streaks”

37. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 36 Benefits of ANN Model PredictorCorrelation R 2 PorosityPermeability CNL log0.540.09 GR log0.02<0.01 LLD log0.100.01 PE log0.240.02 RHOB log0.580.13 DP log0.670.17 ANN Model0.740.71

38. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 37 Presentation Outline  Background and Project Description  Data availability and Preliminary Processing  Model Building  Accomplishments & Next Steps

39. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 38Accomplishments/Progress QC Data 3D Seismic X-well, VSP Log Core Well Completion Production Clustering Logs (Seismic) (Log/seismic) (Log/core) Seismic Processing Depth/time 3D/borehole Attribute extraction Engineering Model Log/core , k Broadband Transform Function 3D/X-Well/Log Validation Simulation? Statistical? Uncertainty analysis? DONE DONE DONE DONE Acquire Site McElroy Field ChevronTexaco DONE Rock Physics Modeling Critical attributes DONE

40. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 39 Current Focus  X-well-to-Log Model  Surface-to-X-well Model  Automate Predictive Workflow  Generate Hi Res 3D Ø and k Volumes.

41. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 40 X-well-to-Log Model  +/- 2,000 datapoints (3 wells x 325 ft/well x ½ ft sample interval).  Increasing to almost 5,000 by using multiple traces.  11 X-well attributes as input; 6 logs as output.  Will evaluate applicability of “depth windowing”.  60% for training, 20% for testing, 20% for validation.

42. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 41 Preliminary Results Actual vs. Predicted (all logs) Sample Density Log

43. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 42 Surface-to-X-well Model  +/- 180,000 data points (135 traces/line x 2 lines x 325 ft/trace x ½ ft sample interval).  11 surface seismic attributes as input; 11 X-well seismic attributes as output.  Will evaluate application of “depth windowing”.  60% for training, 20% for testing, 20% for validation.

44. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 43 Workflow Automation  Need to generate 37 million prediction values  Data cube is 1.8 mi x 1.8 mi x 300 ft  55 ft horizontal resolution (bin spacing)  ½ ft vertical resolution (core sampling interval)  176 inlines x 176 crosslines (≈ 31,000 traces)  300 ft interval  ½ ft increments  Ø, k  Scripts being used to feed each 3D seismic trace through each model:  11 seismic attributes → 11 X-well attributes  11 X-well attributes → 6 logs  6 logs → 2 core values (Ø, k)  Timing: Expect to be finished and have 3D Ø/k cube by end of September.

45. ADVANCED RESOURCES INTERNATIONAL SP09092003-2272 44 Validation  Compare predictions to two cored wells not used in study.  Compare predictions to all logs/core, and as a function of distance from central study area.  Compare predictions to results of time-lapse crosswell survey (pre- vs. post-steam)  Compare predictions to ChevronTexaco’s existing inversion model.  Error Analysis.

46. AdvancedResourcesInternational