Egyptian Journal of Petroleum xxx (2017) xxx-xxx

HOSTED BY

Contents lists available at ScienceDirect

Egyptian Journal of Petroleum

journal homepage: www.sciencedirect.com

Full Length Article

Application of Seismic Attribute Technique to estimate the 3D model of Hydraulic Flow Units: A case study of a gas field in Iran

Mohamad Iravania, Mahdi Rastegarnia b, Dariush Javanic, Ali Sanatid'*, Seyed Hasan Hajiabadid

a Department of Petroleum Engineering, Islamic Azad University, Science and Research Branch, Tehran, Iran b Department of Petrophysics, Pars Petro Zagros Engg. & Services Company, Tehran, Iran cDepartment ofMining Engineering, Imam Khomeini International University, Ghazvin, Iran d Faculty of Petrochemical and Petroleum Engineering, Hakim Sabzevari University, Sabzevar, Iran

ARTICLE INFO ABSTRACT

One of the most important steps in evaluation and development of hydrocarbon reservoirs is the mapping of their characteristics. Nowadays, Seismic Attribute Technique is used to build parameters of hydrocarbon reservoirs in inter-well spaces. One of these parameters is the Flow Zone Index (FZI) that has a significant effect on different stages of evaluation, completion, primary and secondary production, reservoir modeling and reservoir management. The aim of this study is to introduce an equation using seismic attribute and FZI log in wells and then generalize it to predict FZI throughout the reservoir. For this purpose, acoustic impedance (AI) volume as an external attribute was created while internal attributes were computed from seismic data. After that, The best set of attributes was determined using stepwise regression after which seismic attributes were applied to multi-attribute analysis to predict FZI. Then, the attribute map resulted from multi-attribute analysis was used to interpret the spatial distribution of the gas bearing carbonate layers. Finally, the optimum number of Hydraulic Flow Units (HFU) was determined by analyzing the break point in the plot of cumulative frequency of FZI for wells and was generalized all over the reservoir by using the 3D HFU model. The results demonstrated that multi-attribute analysis was a striking technique for HFU estimation in hydrocarbon reservoirs that reduces cost and increases rate of success in hydrocarbon exploration. Distribution of producible hydrocarbon zones along with the seismic lines around the reservoir was characterized by studying this model which can help us in choosing the location of new wells and more economical drilling operations.

© 2017 Egyptian Petroleum Research Institute. Production and hosting by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Article history: Received 20 January 2017 Revised 6 February 2017 Accepted 14 February 2017 Available online xxxx

Keywords:

Multi-attribute analysis Seismic attribute Flow Zone Index Acoustic impedance volume Hydraulic Flow Units

1. Introduction

One of the most important steps in evaluation and development of hydrocarbon reservoirs is the mapping of their characteristics. Numerous empirical relations between seismic attributes and well log data have been introduced for estimation of physical properties such as porosity, permeability, etc [1-7]. Multi-attribute analysis is an effective method to use well logs in conjunction with seismic data for prediction of well log properties from the seismic responses. In this paper multi-attribute method is used to examine the prediction of Flow Zone Index (FZI) logs from seismic attributes. Permeability and FZI has a significant effect on different stages of evaluation, completion, primary and secondary produc-

Peer review under responsibility of Egyptian Petroleum Research Institute.

* Corresponding author. E-mail address: ali.sanati@yahoo.com (A. Sanati).

tion, reservoir modeling and reservoir management. Therefore, different methods have been used to evaluate FZI by petroleum engineers [8-13]. Rezaee et al. introduced a new approach to determine hydraulic flow unit (HFU) by the current zone indicator (CZI) and electrical flow unit (EFU) concepts [14].

Recently, Dezfoolian et al. presented an intelligent model based on probabilistic neural networks (PNN) to produce a quantitative formulation between seismic attributes and HFU in Kangan and Dalan carbonate reservoirs [15]. Rastegarnia and Kadkhodaie-Ilkhchi used seismic attribute analysis to predict FZI using seismic and well log data. They showed that it is an effective technique to predict FZI in an oil reservoir [16]. Moreover, Yarmohammadi et al. delineated high porosity and permeability zones by using the seismic derived flow zone indicator data at the Shah Deniz sandstone packages [17]. Rastegarnia et al. studied the application of seismic attribute analysis and neural Network to estimate the 3D FZI and electrofacies of a hydrocarbon reservoir. They found that these

http://dx.doi.org/10.1016/j.ejpe.2017.02.003

1110-0621/® 2017 Egyptian Petroleum Research Institute. Production and hosting by Elsevier B.V.

This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

M. ¡ravani et al./Egyptian Journal of Petroleum xxx (2017) xxx-xxx

methods are successful in modeling of FZI and electrofacies from 3D seismic data [18].

In this study, FZI was estimated by use of seismic attribute analysis in one of Iranian gas fields. For this purpose, acoustic

Fig. 1. Geographical location of the gas field A.

impedance volume as an external attribute was created and the internal attributes were computed from seismic data. The best set of attributes was determined using stepwise regression. Seismic attributes were applied to multi-attribute analysis to predict FZI. The attribute map resulted from multi-attribute analysis was then used to interpret spatial distribution of the gas bearing carbonate layer. Petrophysical data of two wells, 3D seismic data cube, and structurally interpreted data from a gas field in the southwest of Iran were used in this study.

2. Study area

The gas field A, is one of the largest gas accumulations in the world containing about 8.5 trillions cubic meters of gas. In the Persian Gulf zone, the Permian gas basin is known as the Khuff Formation. This sequence is composed mainly of carbonate rocks and is extended in Bahrain, Qatar, Abu Dhabi, Saudi Arabia, and Iran (Fig. 1) [19]. Data in this study came from two wells namely SP5 and SP9 penetrated into the specific reservoir in the field for which FZI data and petrophysical logs were available. Figs. 2 and 3 show the wells used in this work to build a spatial distribution of FZI. As can be seen, there is a good relationship between FZI with saturation gas and lithology. The Dalan formation is divided into

Fig. 2. Well SP9 used in this study to build a 3D distribution of FZI in K4 member of Dalan formation.

M. ¡ravani et al./Egyptian Journal of Petroleum xxx (2017) xxx-xxx

Fig. 3. Well SP5 used in this study to build a 3D distribution of FZI in K4 member of Dalan formation.

AGE FORMATION LITHOLOGY PET.

MESOZOIC IC S 5 £ Dashtak L.Sudair Shale and Clay with Dolomitic Intercalations Seal

VI < 3 H Aghar Mbr. Shale

« VI Kangan Mbr. Shale

J s CS K1 Dolomite+Anhydrite

M s Anhydrite+Dolomite Gas

a K2 Dolomite

Limestone

K3 Anhydrite+Dolomite

Pi E -- S s ri Anhydrie

y o N o H z < MI pi W PH ri D S K4 Dolomite Limestone Anhydrite+Dolomite Gas

- H Nar Mbr. Anhydrite

- - Ö

R S s R K5 Limestone+Dolomite

Fig. 4. Stratigraphic of Permian-Triassic sequence in the Persian Gulf field.

M. ¡ravani et al./Egyptian Journal of Petroleum xxx (2017) xxx-xxx

Fig. 5. Characteristic of the extracted wavelet.

Fig. 6. Calculated correlation coefficient between real trace seismic and synthetic traces for well sp9.

four zonations that are composed of biogenic limestone dolomite and thin bed of anhydrite (Fig. 4). In this study, we used the 3D seismic data to make a 3D HFU model for K4 section of Dalan formation.

3. Methodology

The aim of this study is to apply multi-attribute analysis of seismic data for prediction of FZI throughout the Dalan formation. For this purpose, post stack 3D seismic and well log data of two wells were used (Figs. 1 and 2). Moreover, porosity logs and FZI log were available for all wells, but check shot data were only available in one well. In order to build a 3D model of FZI the below procedure was followed based on Rastegarnia and Kadkhodaei-Ilkhichi [16]:

• First, an acoustic impedance model was created using modelbased inversion that used as an external attribute for the creation of a 3D FZI model;

• Then the equation for correlating seismic attributes to FZI log was determined for wells in the field. This equation is applied for estimation of the FZI log in the intervals between the wells.

• Finally, the optimal number of HFUs was determined using the plot of cumulative frequency of FZI for both wells and was generalized on the created 3D HFU model.

3.1. Model-based inversion

Seismic inversion is a preliminary study in reservoir characterization. Therefore, there is a continuous effort to optimize the inversion algorithm and improve the resolution of the inverted volume. The amplitude-based seismic data were processed through a

M. ¡ravani et al./Egyptian Journal of Petroleum xxx (2017) xxx-xxx

Fig. 7. Correlation coefficient resulted in initial model for well sp9.

Fig. 8. Broadband acoustic impedance inversion based on model.

M. ¡ravani et al./Egyptian Journal of Petroleum xxx (2017) xxx-xxx

Average Error ( un)

123456789 10

Number of Attributes

Legend

—|— AI Well Error —§— Validation Error-

Fig. 9. The multi-attribute analysis showing the average RMS and validation error; the optimum number of attributes is equal to 7.

Table 1

Multi-attributes extracted for predicting the FZI.

Target Final attribute Training error Validation

(mm) error (mm)

FZI Acoustic impedance 55.15 74.90

FZI Amplitude envelope 48.70 71.03

FZI Filter 15/20-25/30 45.17 68.171

FZI Filter 45/50-55/60 42.38 84.803

FZI Filter 55/60-65/70 38.6 96.03

FZI Quadrature trace 35.31 69.22

FZI Cosine instantaneous phase 29.94 49.84

model-based inversion algorithm to produce acoustic impedance volume which was used as an external attribute in the multi-

attribute analysis [20]. For this purpose, sonic log data were corrected via available seismic data (check shot), geological horizons were determined, picked and interpreted in seismic lines and then best synthetic wavelet and seismogram were extracted. Fig. 5 shows the extracted identification wavelet. Well logs and seismic were calibrated and the highest correlation coefficient between real traces seismic and synthetic traces seismic was obtained using calculated wavelet (Fig. 6). Afterwards, an initial model has been developed and inverted for error correction (Fig. 7).

To perform the model-based inversion, a geological model was compared with seismic data. In order to find a better match, the result of comparison is then used to iteratively update the model. Since this model does not use the direct inversion of the seismic data, it is quite applicable [21]. In current study, ''hard constraint" method was implemented. Indeed, this method considers additional information as a hard constraint that sets absolute boundaries on how far the final answer may deviate from the initial model. In the model-based inversion algorithm, average block size and the number of iterations are of prime importance. In this case, using average block size greater than seismic sample interval is a must [21]. The logic behind these recommendations is the assumption that there might be some false recorded readings from surrounded environment that are possibly mixed with the data. A larger number of iteration results in better accuracy, even though it is time consuming process. By investigating 3D seismic data from Dalan formation, the following results were obtained; constraint limit was 25%, average block size was 4 ms, and number of iterations was set to 10. The results of model-based inversion demonstrate that this type of modeling can clearly identify considered geological layers (Fig. 8).

3.2. Constructing a 3D model of FZI

In building a 3D model of FZI, the most important step is the determination of seismic attributes number. In this study, the appropriate seismic attributes were selected by stepwise regression and cross validation techniques. The stepwise regression

Fig. 10. The multi-attribute analysis result shows the average RMS error; the optimum operator length is equal to 8.

M. ¡ravani et al./Egyptian Journal of Petroleum xxx (2017) xxx-xxx

Fig. 11. Measured FZI logs (in black) and the predicted ones from the multi-attribute analysis (in red); the correlation of the training data is 0.8

Fig. 12. Measured FZI logs (in black) and predicted ones (in red); the correlation of validation data is 0.66.

8 M. ¡ravani et al./Egyptian Journal of Petroleum xxx (2017) xxx-xxx

Table 2

The results of the multi-attribute analysis.

Volume type Train data Validation data

Correlation Error Correlation Error

FZI 0.88 29.9 mm 0.66 49.8 mm

Fig. 13. FZI and AI volume slice below K4 member of Dalan horizon at 1650 ms.

method divides dataset into two parts that are training dataset (original wells, black line in Fig. 9) and validating dataset (predicted data, red line in Fig. 9). As shown in Fig. 9, the horizontal axis represents the number of attributes used in the prediction and the vertical axis is the corresponding root-mean-squared prediction error. This figure demonstrates that the minimum validating error occurs when only 7 attributes are used. The used seismic attributes and their corresponding prediction errors and correla-

tion coefficients are listed in Table 1. So, seismic attributes consist of acoustic impedance (AI), quadrature trace, amplitude envelope, filter 15/20-25/30, filter 45/50-55/60, filter 55/60-65/70, and cosine of instantaneous phase were used in this study to estimate flow zone indicator. Foregoing filters are frequency internal attribute that extracted from raw 3D cube seismic. Each row is related to a particular multi-attribute and the successive row accumulates the previous attributes. The value of amplitude envelope attribute

M. ¡ravani et al./Egyptian Journal of Petroleum xxx (2017) xxx-xxx

Fig. 14. FZI and AI volume slice below K4 member of Dalan horizon at 1670 ms.

is dependent on phase and correlated directly with the change in acoustic impedance. In addition, the instantaneous phase indicator represents the continuation of the layers. The instantaneous phase is an effective attribute in highlighting discontinuities on seismic and detection of reflectors, faults, pinch-outs, angularities, and bed interfaces [16]. AI attribute indicates reservoir properties such as porosity, permeability, and FZI. There is a reverse relationship between AI and FZI. FZI is defined as the relationship between the volumetric proportions of pore space to its geometric distribution. On the other hand, acoustic impedance is a function of both density and velocity. Since deeper layers are harder and denser, velocity increases with depth. Moreover, reduction in wave velocities and density resulted from porosity and fluid saturation is leading to decrease in AI. Such a correlation can be improved by applying the residual time-shift between the target FZI logs and the seismic data.

The length parameter is used to resolve the difference between the frequency content of the target log and seismic attribute because the log data have high frequency and seismic data have low frequency with some specified length to relate a group of

neighbouring seismic samples around the point target log (FZI). In fact, this parameter estimates target log (FZI) by using an average weight of a group of seismic samples on each attribute. So, the parameter operator length determines the length of the convolution operator. In this study, the optimum value of operator length was determined as high as 8. Fig. 10 shows where the number of seismic attribute is 7, the optimum value of operator length is 8 where the average error is minimum.

After determining the optimum number of attributes and length operator, an equation was found between target log (FZI) and seismic attributes in well locations. Fig. 11 demonstrates, for the 7th attribute and an eight-point convolution operator, the correlation exponent and RMS error between the predicted FZI and actual FZI log in well location are 88% and 29.9 mm respectively for the training data. Fig. 12 shows the target log in black versus the "predicted" log in red for each well for validation data. Finally, relationship between multi-attribute features and the target log was used to predict FZI in inter-well spaces by using seismic data. Results obtained from multi-attribute analysis are shown in Table 2.

M. Iravani et al./Egyptian Journal of Petroleum xxx (2017) xxx-xxx

inverted Amplitude at 1 700 ms with a window of 10 ms below and showing the Arithmetic Mean.

Color Key

2937500.0 -

(77 >.1)

¿P9 MM7 '0, 621)

200,621

622500 625000 627500 630000 632500 635000 637500 640000 642500 645000 647500

JLJ JJ

Legend

■ Inline

■ Well Positions

c<wripute<i_l1ow_xocw_r%dex Ampttude at 170O rrte with a window of 10 ms centered and showing ttie Arihmetic toeen.

Color Key

2347500.0 -

2940000.0 2837500.0 2935000 0 2932500.0

(770,1)

'0. 621)

' « !00,1)>

500,621

S22WC 62SOOQ 627500 630000 632500 635000 6375M 640000 642500 645000

jJ -Ü

Legend

■ We« Postions

Fig. 15. FZI and AI volume slice, below K4 member of Dalan horizon at 1700 ms.

Log FZI

Fig. 16. Cumulative frequency of FZI and optimal number of HFU categories.

M. ¡ravani et al./Egyptian Journal of Petroleum xxx (2017) xxx-xxx

Table 3

FZI ranges in each set of HFU.

HFU category FZI range

HFU1 < Log FZI < 0.9674

HFU2 -0.9674 < Log FZI < 0.15598

HFU3 0.155979 < Log FZI < 1.77864

HFU4 1.77864 < Log FZI < 2.52756

HFU5 2.52756 < Log FZI <

10000 1000 100 10

û 1 S

^ 0.1 0.01 0.001 0.0001 0.00001

R2 = 0.7932 R2 = 0.964

♦ HFU1 HFU2 HFU3 HFU4 HFU5

0 0.1 0.2 0.3 0.4

PHI(V/V)

Fig. 17. Plot of permeability against porosity in each category of HFUs.

m ■Я1№<т^" -

♦ HFU 1 ■ HFU 2 ▲ HFU 3 X HFU 4 Ж HFU 5

-HFU 1

HFU 2 HFU 3 -HFU 4

Fig. 18. RQI versus ®z in each category of HFUs throughout the reservoir.

Figs. 13-15 show the comparison between FZI and AI resulted from multi-attribute analysis. These figures demonstrates that an increase in AI causes decrease in the FZI.

3.3. Determination of HFU

HFUs are defined as zones within a reservoir having the potential to control the fluid flow. Each flow unit is characterized by a flow zone indicator, which can be understood in terms of the relationship between the volume of void space, 0z, and the geometric distribution of pore space (quantified as the reservoir quality index, RQI) as follows [10]:

log(RQI) = log(FZI) + log(Uz)

and u is the effective porosity.

RQI can be calculated using the following equations:

RQI = 0.0314</(|

where k is permeability in milli-Darcy and u is the fractional porosity. The FZI can be rearranged in terms of the measurable RQI as given below:

Rocks with a narrow range of FZI values belong to a single hydraulic unit, i.e., they have similar flow properties [13]. In this study, optimal number of HFU categories, which depends on tact of user, is determined by analyzing break point in the plot of cumulative frequency of FZI for both SP5 and SP9 (Fig. 16). The ranges of FZI in each set of HFU are shown in Table 3.

Fig. 17 plots the semi-logarithmic permeability against porosity. There is a good agreement between permeability and porosity in each defined HFUs. It can be seen that porosity and permeability of reservoir is increasing with respect to HFU categories. HFU1 demonstrates low quality reservoir with low amount of permeability and porosity whereas HFU5 represents a high quality reservoir with large amount of permeability and porosity.

Fig. 18 displays the RQI versus pore to matrix volume ratio in each set of HFUs throughout the reservoir of the gas Field A.

A combination of indexes which are derived in previous section to calculate the FZI, are utilized in order to construct a 3D model of HFU. Since the acoustic impedance and FZI have a reverse relation we conclude that increase in acoustic impedance results in decrease in HFUs and vice versa. HFUs volume slice of reservoir at various time are illustrated in Fig. 19.

4. Results and discussion

In this study, the correlation exponent and RMS error between the predicted FZI and actual FZI log in well location are 88% and 29.9 mm respectively for the training data. The resulted 3D FZI indicates high anomalies, which are in good agreement with the petro-physical properties of oil producing wells in the field of interest. Also, optimal number of HFU categories is determined using the plot of cumulative frequency of FZI for both SP5 and SP9. HFU1 demonstrates low quality reservoir with low amount of permeability and porosity whereas HFU5 represents a high quality reservoir with large amount of permeability and porosity. For inter-well spaces, where well logs and core data were not available, the intelligent model is applied. In this field, from NE to SW, quality of reservoir has been increased significantly and directional wells can be drilled in deeper interval of reservoir to produce more gas. Since HFU depends on the geological properties of the rocks, it helps in tracking hydrocarbon saturation changes over the Dalan and provides information respecting the locations of perforation and development wells.

5. Conclusion

In this study, multi-attribute analysis is successfully used to estimate FZI log from seismic attributes. This technique is a fast method to evaluate the reservoir characteristics as well as to reduce cost and increase rate of success in hydrocarbon exploration. The result of this study showed that multi-attribute analysis was effective to predict HFU deploying seismic. This model can be improved by neural network to be used as a guide model. Distribution of producible hydrocarbon zones along with the seismic lines around the reservoir was characterized by introducing this model which can help us in choosing the location of new wells and more economical drilling operations.

M. Iravani et al./Egyptian Journal of Petroleum xxx (2017) xxx-xxx

Fig. 19. Time slice from 3D HFU resulted at 1650, 1670 and 1700 MS, respectively.

References

[1] A.R. Brown, et al., Interpretation of three-dimensional seismic data, 1996.

[2] J. Calderon, J. Castagna, Porosity and lithologic estimation using rock physics and multi-attribute transforms in Balcon Field, Colombia, The Leading Edge 26 (2) (2007) 142-150.

[3] D.P. Hampson, J.S. Schuelke, J.A. Quirein, Use of multiattribute transforms to predict log properties from seismic data, Geophysics 66 (1) (2001) 220-236.

[4] D.J. Leiphart, B.S. Hart, Comparison of linear regression and a probabilistic neural network to predict porosity from 3-D seismic attributes in Lower Brushy Canyon channeled sandstones, southeast New Mexico, Geophysics 66 (5) (2001) 1349-1358.

[5] B. Russell et al., Multiattribute seismic analysis, Leading Edge 16 (10) (1997) 1439-1444.

[6] P.S. Schultz et al., Seismic-guided estimation of log properties (Part 1: a data-driven interpretation methodology), Leading Edge 13 (5) (1994) 305-310.

[7] S. Ronen et al., Seismic-guided estimation of log properties (Part 2: using artificial neural networks for nonlinear attribute calibration), Leading Edge 13 (6) (1994) 674-678.

[8] M. Abbaszadeh, H. Fujii, F. Fujimoto, Permeability prediction by hydraulic flow units-theory and applications, SPE Form. Eval. 11 (04) (1996) 263-271.

[9] F.A. Al-Ajmi, S.A. Holditch. Permeability estimation using hydraulic flow units in a Central Arabia reservoir, in: SPE Annual Technical Conference and Exhibition, Society of Petroleum Engineers, 2000.

[10] J.O. Amaefule et al., Enhanced reservoir description: using core and log data to identify hydraulic (flow) units and predict permeability in uncored intervals/ wells, in: SPE Annual Technical Conference and Exhibition, Society of Petroleum Engineers, 1993.

M. ¡ravani et al./Egyptian Journal of Petroleum xxx (2017) xxx-xxx 13

[11] E. Kazemzadeh, B.H.M. Nabi, M.R. Rezaei, The study of formation resistivity factor by using hydraulic flow units method in carbonate reservoirs, 2008.

[12] J. Kozeny, Über kapillare Leitung des Wassers im Boden:(Aufstieg, Versickerung und Anwendung auf die Bewässerung), Hölder-Pichler-Tempsky, 1927.

[13] M. Prasad, Velocity-permeability relations within hydraulic units, Geophysics 68 (1) (2003) 108-117.

[14] M.R. Rezaee, A.K. Ilkhchi, A. Barabadi, Prediction of shear wave velocity from petrophysical data utilizing intelligent systems: an example from a sandstone reservoir of Carnarvon Basin, Australia, J. Pet. Sci. Eng. 55 (3) (2007) 201-212.

[15] M.A. Dezfoolian, M.A. Riahi, A. Kadkhodaie-Ilkhchi, Conversion of 3D seismic attributes to reservoir hydraulic flow units using a neural network approach: an example from the Kangan and Dalan carbonate reservoirs, the world's largest non-associated gas reservoirs, near the Persian Gulf, Earth Sci. Res. J. 17 (2) (2013) 75-85.

[16] M. Rastegarnia, A. Kadkhodaie, Estimation of flow zone indicator distribution by using seismic data: a case study from a central iranian oilfield, Iran. J. Oil Gas Sci. Technol. 2 (4) (2013) 12-25.

[17] S. Yarmohammadi et al., Seismic reservoir characterization of a deep water sandstone reservoir using hydraulic and electrical flow units: a case study from the Shah Deniz gas field, the South Caspian Sea, J. Pet. Sci. Eng. 118 (2014) 52-60.

[18] M. Rastegarnia, A. Sanati, D. Javani, et al., A comparative study of 3D FZI and electrofacies modeling using seismic attribute analysis and neural network technique: a case study of cheshmeh-khosh oil field in Iran, Petroleum (2016).

[19] E. Galimov, A. Rabbani, Geochemical characteristics and origin of natural gas in southern Iran, Geochem. Int. 39 (8) (2001) 780-792.

[20] D.A. Cooke, W.A. Schneider, Generalized linear inversion of reflection seismic data, Geophysics 48 (6) (1983) 665-676.

[21] M. Haghighi, A. Javaherian, E.I. ABD, Model-based inversion on 3-d seismic data of Ab-Teymur oilfield, 2006.