1-1 Correlating Heterogeneous Production to Lithology and Fractures in Coalbed Methane Exploitation Jeremy C. Fisk (University of Oklahoma)
AGENDA I. Geology II. Coal / Coalbed Methane III. Data Overview IV. Rock Property Predictions V. Fault, Fracture, Cleat Analysis VI. Conclusion 1-2
World s 2 nd Largest CBM Producer 1-3 (www.ga.gov.au/map/)
0 200 km N 1-4 (after Korsch et al, 1988)
Back-Arc Setting (inland of the orogen) North 1-5 (after Peter, J.M. 2009)
Stratigraphy Coarse to medium grained, high energy braided river SS Poor to moderately sorted sandstone with siltstone and claystone, oxidized floodplain, with fluvial episodes, cementation Low energy, fluvio /deltaic, well developed bright coal seams with siltstone and interbedded sandstone and tuffaceous beds 1-6 Low / moderate energy, fluvial, back swamp to deltaic, thin dull coal seams with siltstone and interbedded sandstone
COAL is a readily combustible rock containing more than 50 percent by weight of carbonaceous (organic) material, formed from compaction and induration of variously altered plant remains similar to those in peat [adapted from AGI's "Glossary of Geology"]. PEAT is an unconsolidated deposit of plant remains from a water-saturated environment such as a swamp or marsh [adapted from AGI's "Glossary of Geology"]. 1-7 (courtesy Brian Cardott, OGS)
Coal is formed from peat in swamps Example of swamp in Hoh Rain Forest, Washington 1-8 (courtesy Brian Cardott, OGS)
1-9 (courtesy Brian Cardott, OGS)
RANK refers to the physical and chemical changes that occur to organic matter as it is affected by increasing temperature and time. [LIGNITE SUBBITUMINOUS BITUMINOUS (High Volatile; Medium Volatile; Low Volatile) ANTHRACITE (Semianthracite; Anthracite; Meta-anthracite)] 1-10 (courtesy Brian Cardott, OGS)
1-11 (courtesy Brian Cardott, OGS after Boyer 1989)
Dual Porosity, Dual Permeability (courtesy Brian Cardott, OGS after Boyer 1989)
1-13 (courtesy Brian Cardott, OGS after Boyer 1989)
Rate Methane Water Time 1-14 (after Ayers, 2002)
1. Source Rock 15 (courtesy Brian Cardott, OGS after Diamond 1993)
FACE CLEAT (courtesy Brian Cardott, OGS)
Shot Interval (50 m) Receiver Line Spacing (200 m) * shot 31 km², 6-130 Hz (sweep freq) * receiver 4 x 15-ton Vibroseis trucks * active receiver Shot Line Spacing (200 m) Receiver Lines (96 channels) 1-17 Shot Lines Receiver Interval (25 m) (Modified after Edip Baysal, Paradigm)
1-18
3 mi. 9.9 km N 1-19
Flattened Variance of C2 Coal Seam Best wells in survey N Variance 1-20 3 mi. 9.9 km
Generating Predictive Property Volume 650 m 50 DT 180 1.00 RHOB 3.00 575 ms 875 m 1-21 725 ms
22 Full Wavelet extracted from 4 wells
Multi-Attribute Analysis L(t)= target log to predict at given time sample w=weights assigned, that give closest result to the log in a least-squares sense A=attribute value at given time sample Li=actual log value at given time sample 1-23 (Hampson et al, 2001)
Multi-Attribute Analysis 24 Seismic Density Log
Predicted Density [gm/cc] 2.5 2.0 Training Error = 0.208 [gm/cc] Validation Error = 0.234 [gm/cc] Correlation = 0.86 3 Attributes : Integrated trace, 2 nd Derivative Instantaneous Amplitude, Filter 25/30 35/40 4 pt Operator Top of C2 to Top of M. Baralaba 25 1, 1.25 Actual Density [gm/cc] 2.0 2.5
[gm/cc].26 Average Error vs. Number of Attributes Validation error Training error.19 1 5 26
[gm/cc].29 Avg. Error vs. No. of Attr. (per operator) 4-point operator.23 1 5 27
Linear Regression Neural Network 28 (Hampson et al, 2001)
Probabilistic Neural Network (PNN) W1 W2 W3 W4 Xj=Attribute values at well Xij=Attribute values away from well Li=Actual Log L(x)=predicted Log σ= smoothing parameter(empirical) 29 (Chart by Specht, 1988 and Equations by Hampson et al, 2001)
PNN Parameters Density Validation Error: 0.213 gm/cc Validation Correlation: 0.798 Window Analysis : Top of Rewan to Top of M. Baralaba Coal Measures Training Wells: Well A,C,D,F Validation Well: Well E (This well was left out of the network training to assess the prediction value) 30
PNN Computed Density Higher production Lower Production 3 mi. 9.9 km N 31
Density blended with Positive Curvature Higher production Lower Production 3 mi. 9.9 km N
Acoustic Impedance Inversion Higher production Lower Production 3 mi. 9.9 km N 33
A Depositional Environment 34 (Edmunds, 2002)
Linear Regression Applied for Coal Thickness map (meters) N 35 3 mi. 9.9 km
FACE CLEAT MAX. PALEO STRESS (courtesy Brian Cardott, OGS)
Fault and Fracture Analysis 37 3 mi. Flattened Top of C2 coal coherence blended with e_positive 9.9 km
Fault and Fracture Analysis 3 mi. 38 Flattened Top of C2 coal coherence blended with k_positive 9.9 km
1-39 (courtesy Brian Cardott, OGS after Saulsberry et al. 1996)
1-40 (courtesy Brian Cardott, OGS after Lucia 1983)
(courtesy Brian Cardott, OGS after Scott 1995)
Top of C2 -Rose N 1.5 mi. 42 (after Ha Mai, 2009)
Dip Azimuth 3 mi. 9.9 km N 43
Top of C3 -Rose 44 3 mi. 9.9 km N (Ha Mai, 2009)
Well H: C2 Seam: Image Log Core correlation, E-W striking, subvertical resistive (mineralized fractures)
The Way Ahead -Spectral Inversion -Refined PNN using more meaningful attributes -Best Well Completion Strategies for CBM (perforation strategy and fracture stimulation -Continued Fracture Analysis -Borehole Images -Azimuthal Anisotropy -Temperature logs -AVO Analysis 46 -Self Organized Maps : Correlating Coal Thickness and Productivity to Seismic Waveforms
47
3 mi. 9.9 km N 48
49
3 mi. 9.9 km N 50
51
3 mi. 9.9 km N 52
Resolution Conundrum 1.00 RHOB 3.00 876 m 883 m V p = 1,858 m/s f pred =40 hz λ/4 = 11.6 m 53 λ= 46 m
Kohonen SOM 54 (modified after Marcilio Matos)