SPE 84055 Geostatistical Prediction of Sand Distribution of Gas Reservoir in Jilin, China

Li-Wei Qiu, SINOPEC, AnPing Yang, Applied Computer Engineering, Sheng-Xiang Long and Zhi-Jiang Kang, SINOPEC

Copyright 2003, Society of Petroleum Engineers Inc. This paper was prepared for presentation at the SPE Annual Technical Conference and Exhibition held in Denver, Colorado, U.S.A., 58 October 2003. 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, as presented, have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material, as presented, does not necessarily reflect any position of the Society of Petroleum Engineers, its officers, or members. Papers presented at SPE meetings are subject to publication review by Editorial Committees of the Society of Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper for commercial purposes 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 where and by whom the paper was presented. Write Librarian, SPE, P.O. Box 833836, Richardson, TX 75083-3836 U.S.A., fax 01-972-952-9435.

resistivity is a good indicator of lithology, the study results demonstrate that geostatistical inversion of seismic data to resistivity is useful in predicting sand distribution, and for optimizing future drilling locations. Introduction The Gujiazhi and Houwujiahu fields are located in Lash County of Jilin Province, China. The Gujiazhi structure is actually part of the broad, gentle Houwujiahu anticline, and are situated in the western end of the Shiou Central structural belt. To date, there are 74 wells in this area, producing 35104 m3 per day in Gujiazhi and about 40104 m3 per day in Houwujiahu, and supplying gas to cities in the region. The main producing strata are in the fan-delta complexes within the Dengloku formation, where the Chuan-I and Chuan-II members are deltaic depositional sequences. The main pay zones are the delta-front distributary-channel sandstones. The structures in the area are complicated by the many faults, the multiple gas pays are widely dispersed, and individual gas sands are thin. The physical properties of the gas sands vary greatly, while the prominent gas sands have moderate to low permeability, and are tight reservoirs with strong inhomogeneity. Our study involves stratigraphic analysis, sequence identification and correlation, depositional microfacies, detailed 3-D structure interpretation and log interpretation. From this basis, we performed reservoir modeling to describe the reservoirs in detail and the spatial variation of physical properties. The reservoir modeling study was divided into two parts: For the Gujiazhi field, generate 3-D, horizontal and cross-sectional models of porosity, permeability and shale contents for the primary reservoir sands: N-IX, N-X, N-XI, X-I, and X-II, using well log data from 12 wells and core analysis data from 4 wells. For the Houwujiahu field, generate 3-D models of porosity, permeability and shale contents for the Quan-I member, using well log data from 24 wells. This allows better three-dimensional visualization of the reservoirs, and provides enhanced basis for better delineation of the internal distribution of gas pays within the reservoirs section. Reservoir Modeling Of Gujiazhi Field Input data. The input data for the Gujiazhi field are as follows: (1) Data for individual pay zones (Table 1).

Abstract This case study involves sand distribution prediction for the Gujiazhi and Houwujiahu gas fields in Jilin Province, China. Initial data analysis involved correcting and calibrating the well logs. Log data from 74 wells were available from the Gujiazhi field and the Houwujiahu field. This initial process also established 3-D porosity, permeability and clay volume models for each field. The Gujiazhi grid has a cell dimension of 100x100x2 meter, Houwujiahu grid 100x100x1 meter. Geostatistical inversion was conducted for a 3-D seismic volume covering both fields. Synthetic resistivity logs were generated and converted to pseudo-impedance in time, the derived reflectivity was convolved with a wavelet, and the synthetic seismogram correlated with actual seismic traces. This process was allowed to iterate until a satisfactory correlation was reached. One VSP was used for the initial time-depth conversion, and the time-depth relationship for each well was subsequently time-adjusted until the synthetic seismograms matched the actual seismic, allowing a reasonable lateral velocity variation. A common wavelet was extracted for all the wells. With the optimized time-depth control and common wavelet, resistivity inversion was conducted for the 3-D seismic grid. A 3-D reservoir geology model was generated by co-kriging, using the well log data as hard data and the inverted seismic data as soft data. The 3-D reservoir geology model from well logs showed high vertical resolution but poor lateral resolution, while the 3-D seismic-inverted model showed higher lateral resolution but poor vertical resolution. The co-kriging model showed favorable resolution both vertically and laterally. The resulting reservoir geology model provided insights into assessing the distribution of pay sands. Since in this area,

SPE 84055

(2) Well log interpretation: the source of the data comes from the log interpretation results for this study, including porosity, permeability and shale contents. (3) Stratigraphic and fault data: the reference plane is the N-X horizon according to the interpretation of the Northeast Petroleum Bureau. The depths to other tops are derived based on bed thickness. Data Quality Control. In order to create a high-quality reservoir geology model, it is of paramount importance to check data and control quality. (1) Using different tables and figures in checking the validity of well data and data points to rectify data quality problems on geographic coordinates, elevation, well deviation, stratigraphy, and log curves. (2) We can use the polygonal method or the limiting data range method to delete well-log data. The limiting data range can be data value, depth (or calibrated time) or horizon delineation and correlation. This may involve deletion or combination of log curves. Compute variograms for spatial analysis. The spatial continuity in different directions can best be expressed by the variogram function. The variogram can help make full use of data and information, and allows integration with raw data to produce best possible 3-D model. Figure 1 shows the following: (1) Vertical variogram (upper right, Figure 1) represents average reservoir thickness from vertical variogram computed along the well bore. (2) Well-to-well Variogram (upper and lower left, Figure 1) represents the lateral reservoir continuity. Horizontally the continuity of each of the reservoirs may be different in different direction. Also if the variograms from in two closely spaced wells are vastly different, this may reflect in errors in the log data, coordinates, or stratigraphy in one well or both of the wells. (3) Mapview Variogram (lower right, Figure 1) expresses well-to-well variogram with parameters such as distance and direction grouping. Model Generation. For our model the cell dimension was assigned as 100m100m2 m, making up a total of 5763211 = 757,701 cells. We may then generate different 3-D models through 3D interpolation (Figure 2). From this, we can cut along different directions, or depth, to create cross-sections or depth slices. Since the accuracy of permeability interpretation is low, we used core data as hard data (from 4 wells) and log data as soft data to increased the accuracy, specifically we performed co-kriging of the data sets. From the cross-plot of core-derived permeability and log-derived permeability, where the correlation coefficient was 0.8368, we obtained the function for converting data. Starting from least-square fit of core data, we can model permeability and expand it into the 5763211 grid cells to generate a 3-D model. A permeability fan-section is shown in Figure 3.

Reservoir Modeling of Houwujiahu Input Data. The input data for the Houwujiahu field are as follows: (1) Data for individual pay zones (2) Well log interpretation: data come from log interpretation of porosity, permeability and shale contents. (3) Stratigraphy and faults data: the reference horizon is T30 according to the interpretation of Northeast Petroleum Bureau. The depths to other tops are derived the bed thickness. Modeling Steps. The steps for modeling in the Houwujiahu field are the same as Gujiazhi Field, while the cell dimension is 100m100m1m, totaling 2,426,031 cells. Seismic Inversion Study A reservoir geology model based on limited well control would likely lead to inaccurate reservoir description. In order to improve the accuracy of lateral prediction and to assist future operations in peripheral areas of the two fields, we created models utilizing seismic inversion and integrated well-log reservoir modeling. Seismic data serve two purposes in reservoir description: structure and attribute. Horizon and faults are used to generate the structural framework of the reservoir geology models. When the distances between wells are large or the stratigraphic variation is great, seismic data will help to improve the 3-D attribute models. There are two approaches to combine seismic attributes with lithology or porosity for an integrated evaluation: one is statistical calibration, and the other is seismic inversion. Accurate time-depth relationship and sufficient seismic resolution to resolve necessary layer thickness are key factors for successful seismic inversion. Having an appropriate wavelet is also a necessary condition for success. We used GridSTAT Pro to generate a single composite wavelet (derived from wavelets from many wells), since it allows, at the same time, very fine adjustment of time-depth relationship for fine-tuning of the match between synthetic seismograms and actual seismic data. Successful application of seismic inversion can be found in other examples [1] [2]. Seismic Inversion for Resistivity and Reservoir Modeling. (1) Well-log data Processing. In the area, the R2.5 curves can be used to differentiate lithology with favorable result. Therefore, we employed the R2.5 curves to evaluate distribution of reservoir sands. Also, an initial step to edit the R2.5 curves is necessary to delete sections with questionable quality, to combine repeated sections, and normalize the processing of the log data. (2) Wavelet extraction and time-depth conversion. The original seismic data was loaded and converted to a seismic grid of 25m25m1m. The seismic grid, the R2.5 curves from 11 wells, 8 horizon and 1 VSP were all loaded to GridSTAT Pro. The seismic traces at the well sites were extracted for use together with the R2.5 curves from 11 wells and 1 VSP in the generation of synthetic

SPE 84055

seismograms. An initial wavelet was provided and iterations in refinement of the wavelet were repeated until a satisfactory correlation of the synthetic trace and actual seismic was reached. The correlation of the synthetic seismogram with actual seismic data was a 0.7 (Figure 4). (3) An initial R2.5-based inversion. The T30 horizon derived from seismic interpretation was correlated with the corresponding Quan-I well markers at eight different wells to generate a horizon structure model as a constraint. An initial R2.5 log model was produce through interpolation with simple kriging of the R2.5 curves from 11 wells. The cell dimension of the initial model was 50m50m1m. (Figure 5). Using the initial model as reference and the actual seismic as inversion target, the extracted wavelet and the accurate time-depth function were used to iterate 400 times to derived a R2.5 inversion model (Figure 6). The grid dimension for the inversion model was the same as the initial model, where this was used as soft data in the ensuing model generation step. (4) Generation of Reservoir Geology model. The inversion grid was used as soft data, and the R2.5 curves from the 11 wells as hard data, and utilizing the T30 structure surface model (derived from seismic correlated with Quan-I marker from eight wells) as constraint, an inverted R2.5 model was generated by co-kriging (Figure 7). Assessment of Results. Three different models were shown in Figures 5, 6 and 7, an initial model derived from log data, a model from seismic inversion, and the final model integrating well data and seismic inversion results. From the figures, we can observe that log-derived model has higher vertical resolution but poor lateral resolution; the seismic inversion model has higher lateral resolution but poor vertical resolution; whereas the finial model integrating wells data and seismic inversion has favorable vertical and lateral resolution. Conclusions From the case study, we can draw the follow conclusions: 1. Geostatistical inversion is a technology that can be used to combine with various types of log data, to realize effective integration of seismic and log data. 2. Integrated seismic inversion modeling can greatly enhance the accuracy of reservoir predictions. 3. Calibration of seismic horizons with and generation of initial model are critical tasks, and they are the core processes in seismic inversion. 4. Obtaining accurate time-depth relationship and appropriate wavelets are key to the success of seismic inversion. References
1 "Geostatistical Modeling of Porosity with Seismic Data A Case History", A.P. Yang, T.G. Harris, and C. Li, AAPG Hedberg Research Conference on Applied Reservoir Characterization Using Geostatistics, The Woodlands, Texas, Dec. 3-6, 2000.

"Geostatistical Evaluation of QHD 32-6 Oil Field: a DRO in China", Tai-chang Shih, An-Ping Yang, Charles T. Cheng, Youli Quan, Gou Fan Lu, Ben Wang, D. Lee Martin, and Eric M. Hussey, SPE 48879, SPE International Meeting on Petroleum Engineering, Beijing, China, 2-6 November 1998

SPE 84055

Well Name SN76 SN84 SN85 SN86 SN87 SN89 SN90 SN91

N-IX 1517 1484 1462 1318 1500 1555 1405 1190

N-X 1600 1572 1512 1400 1582 1630 1467 1278

N-XI 1668 1643 1580 1470 1665 1702 1538 1350

X-I 1785 1765 1703 1583 1774 1815 1656 1460

X-II 1925 1903 1838 1712 1865 1961 1785 1590

X-III 2056

Basal X-III 2175

1972 1857 2001

2097

2135

1915 1718

2030

Table 1 Individual-layer data from Gujiazi Field

2.5 2.5

Variogram

Variogram

1.5

0.5

Vert. Vert.fit N-S N-S fit NE-SW NE-SW fit E-W E-W fit NW-SE NW-SE fit

1.5

0.5

1000

2000 Separation Distance (M)

3000

4000

100

200 300 400 Separation Distance (M)

500

600

700

All Variograms por

-

Vertical Variogram por

-

1.8

GK3_SN91 SN86_SN91

4000

1.5
GK8_SN76 SN87_SN89 GK8_SN86 SN89_SN90 GK3_SN84

2000 North-South (M)

1.2 Variance

GK8_SN84 SN84_SN89

SN86_SN90SN76_SN89

0.9

GK8_SN87 GK8_SN89 GK1_SN89 SN85_SN86 SN84_SN87 GK3_SN90 GK3_GK8 GK2_SN84 GK8_SN90 GK2_SN76 GK1_SN84 SN76_SN84 SN87_SN90 GK1_SN76 SN76_SN90 GK2_SN87 SN84_SN90 GK3_SN86 GK1_SN90 GK1_GK3 GK1_GK8 GK3_SN85 SN76_SN85 SN85_SN90 GK2_SN90 GK2_GK8 GK2_SN89 GK1_GK2

0.6

1.980 1.720 1.550 1.380 1.210 1.040 0.870 0.700

-2000 0.3

300

600

900 Distance (M)

1200

1500

1800

-4000 -5000

-3000

-1000 1000 East-West (M)

3000

5000

Well Pair Variance por

-

Areal Variogram
-
GridstatPro 9-25-99

Figure 1

Variogram calculation.

SN87 GK2 GK12 SN84 SN90 GK1 GK8 GK6 SN76 SN89 SN85 GK5 GK3 SN86

SN91

-1300

34.6 23.9 16.5 11.4 7.835 5.404 3.728 2.571

-1800

4847000

4845000

Northing

21599000 Easting 21601000 21603000 4843000

3d Grid View
Porosity

Figure 2

Porosity Model

GridstatPro

9-21-99

SPE 84055

SN89

SN84

SN90

-700

-966.67

-1233.3 -1766.7 1229 238 46.3 8.988 1.745 .339 .066 .013 Subsea Depth

-1500

-2033.3

-2300

-400

480

1360 Distance

2240

3120

SN91 4000
GridstatPro

GK8

FanSection
9-21-99

Figure 3: Permeability Fan Section of Guojiazi Field

SN76

SN89

SN84

SN90

SN91

GK3

SN85

-665.18

-762.70 q3-2top

-860.22 q3-1top Subsea Depth -957.75 q2-4top

-1055.3 q2-3top q2-3-2 q2-2top -1250.3 q2-1top q1-6top -1347.8 -67.323 134.22 335.76 Distance 537.29 738.83

-1152.8

Figure 4
GK1 GK2 -700 q2-4top q2-3top q2-3-2 q2-2top q2-1top Subsea Depth -1140 q1-6top q1-5top -1360 q1-4top q1-3top q1-2top -1580 q1-2-2top d3-3top

Composite Seismogram
GK3 GK8

-920

SN86

GK2

GK1 GK8

GK3

2.2 1 .029

-1800

-100

100 Distance

200

300

400

FanSection

Figure 5 Model Based on Wells

SPE 84055

GK1

GK2

GK3

-700 q2-4top q2-3top q2-3-2 q2-2top q2-1top Subsea Depth -1140 q1-6top q1-5top -1360 q1-4top q1-3top q1-2top -1580 q1-2-2top d3-3top 2.2 1 .4

-920

-1800

-100

100 Distance

200

GK8 300

400

FanSection

Figure 6 Model Based on Seismic D

GK1 GK2 GK3 GK8

-900

q2-2top q2-1top

-1050

q1-6top -1200 Subsea Depth q1-5top q1-4top q1-3top -1500 q1-2top q1-2-2top -1650 d3-3top

-1350

2 1 .4

-1800

-100

100 Distance

200

300

400

FanSection

Figure 7 Model Based on Seismic and Wells