Hi ngh khoa hc và công ngh ln th 9, Trng i hc Bách khoa Tp. HCM, 11/10/2005 145 XÂY DNG MÔ HÌNH CU TRÚC 3 CHIU CHO CU TO DU KHÍ DA VÀO TÀI LIU A CHN VÀ A VT LÝ GING KHOAN CONSTRUCTING A 3-D STRUCTURAL MODEL OF AN OIL & GAS PROSPECT BASED ON SEISMIC AND WELL LOG DATA H Trng Long*, Bùi Th Thanh Huyn**, Keisuke Ushijima1*** * Khoa K thut a cht và Du khí, i hc Bách Khoa Tp.H Chí Minh, Vit Nam ** Department of Civil and Earth Resources Engineering, Kyoto University, Japan *** Exploration Geophysics Laboratory, Graduate School of Engineering, Kyushu University, Japan TÓM TT S minh gii tài liu đa chn 3 chiu cho c hi đ đa ra các bn đ cu trúc di sâu mt đt. Ngoài ra, s kt hp minh gii tài liu đa chn vi tài liu đa vt lý ging khoan s cung c p thêm nhng thông tin đáng tin cy đ thông hiu tt các cu trúc sâu, đc bit là xác đnh các đt gãy và các đi nt n. Trong nghiên cu này, chúng tôi đã s dng mt k thut tính toán da vào máy tính gi là “Mng Nron” đ tính đ rng ca va vi đ chính xác cao. Các giá tr đ rng có th thành lp đc các bn đ phân b đ rng cho mt cu to du khí. Chúng tôi nhn thy rng, các đi có đ rng cao gn lin vi các đt gãy và các đi nt n. Vì vy, s hiu chnh gia các bn đ phân b đ rng và kt qu minh gii tài liu đa chn có th xác đnh các đt gãy và các đi nt n vi đ tin cy cao hn. T đó, mô hình cu trúc 3 chiu s đc thành lp, th hin các hình dng cu trúc và ki n to cho vic đánh giá tim nng hydrocarbon. Chúng tôi đã s dng tài liu ca cu to du khí A2-VD  thm lc đa phía Nam Vit Nam cho bài báo này. Các kt qu thu đc đã cung cp nhng thông tin rt có giá tr cho vic nhn din v trí các ging khoan và khai thác, cng nh cho s phát trin ca cu to này trong tng lai. ABSTRACT Interpretation of three-dimensional (3-D) seismic data gives an opportunity to generate deep subsurface structure maps. Furthermore, combination of seismic with well-logging data interpretation will provide more reliable information for good understanding of deep structures, especially faults and fractured zones prediction. In this study, we used a computing technique based on computer program named “Neural Network”, to predict porosity of reservoirs with high accuracy. Porosity values can build porosity contribution maps for an oil & gas prospect. We found that, the zones with high porosity relate to the faults and fractured zones. Therefore, the correction between porosity distribution maps and results of seismic data interpretation can used to predict faults and fractured zones with higher reliability. Hence, 3-D structural model will be constructed, revealed structural and tectonic configurations for hydrocarbon potential assessment. We used data of A2-VD oil & gas prospect, southern offshore Vietnam, for this paper. Achieved results provided very valuable information for the identification of drilling and production well location, as well as development of the prospect in the future. Hi ngh khoa hc và công ngh ln th 9, Trng i hc Bách khoa Tp. HCM, 11/10/2005 146 1. INTRODUCTION A2-VD oil prospect, located in Cuu Long basin (Figure 1), southern offshore Vietnam is a main target area for oil and gas exploration in Viet Nam with the major reservoir is fractured granite basement (PV, 1998). The Cuu Long basin that was formed during Cenozoic Era under the influence of India-Eurasian collision generating the South China Sea spreading, is the most prospective hydrocarbon basin in offshore Vietnam (Phuong, 1997), especially the A2-VD oil prospect in Block 15-2 is of particular interest. The sedimentary stratigraphy of this basin is divided into several sequences: basement (Pre- Tertiary), sequence E (Lower Oligocene to Eocene), D (Upper Oligocene), C (Early Miocene), B1 (Middle Miocene), and younger sequences (B2 and A). The stratigraphy correlates with wells VD-1X, VD-2X in the study area as presented in Figure 2 (JVPC, 2000 and 2001). 2. THREE-DIMENSIONAL (3-D) SEISMIC DATA INTERPRETATION OF A2-VD PROSPECT In this research, we conducted seismic interpretation of a volume cube for 3-D seismic data in the area 12.5 x 6 km 2 with 345 inlines and 320 crosslines. The major seismic sequences in each section were determined by correlation with stratigraphy derived from the wells in the study area (JVPC, 2000 and 2001). The interpretation was carried out using the basic concepts for seismic stratigraphy interpretation (Badley, 1985; Vail et al., 1977). Figure 3 shows the seismic data interpretation in selected sections. Figure 1 Location of the A2-VD prospect (Modified from PV, 1998; JVPC, 2001) Figure 2 Stratigraphy and wells correlation of Block 15-2 (A2) (after JVPC, 2000) Figure 3 Seismic data interpretation in selected sections Hi ngh khoa hc và công ngh ln th 9, Trng i hc Bách khoa Tp. HCM, 11/10/2005 147 3. POROSITY DISTRIBUTIONS USING NEURAL NETWORK The architecture of NN we used as shown in Figure 4 with one input layer composed of six nodes. These six nodes represent the response of neutron, density, sonic, resistivity (LLS, LLD and MSFL). Figure 4 Architecture of neural network used in this study A single hidden layer has five nodes and the output layer has only one node represents porosity. With data of this study area, more hidden layers or more neurons of each layer is ineffective and make more complex calculation. For training NN, we used training data set which is a data set of 6 inputs parameters from well log data and 1 output parameter is porosity that was selected from core samples. During training process of NN, we applied the most common learning law, back-propagation, as a training law to reduce the errors (Lippman, 1987). However, back-propagation includes several kinds of paradigms such as on-line back-propagation, batch back-propagation, delta-bar-delta, resilient propagation (RPROP) and quick propagation (Werbos, 1994). The most successful paradigm used in this study are batch back-propagation. By using batch back-propagation paradigm, figure 5 shows the RMS errors as a function of training and testing data set patterns of NN, that all of them are lower than 0.1. The data used for the network design are taken from various wells in A2-VD oil prospect. We used derived NN to predict porosity from logs data of all wells in A2-VD oil prospect. Comparison of NN predictions and log predictions with core data are displayed in Figure 6 as a selection of well A2-VD-1X. It shows the results in the cored reservoir intervals, in that NN method is more efficient than conventional log method. Porosity values versus depth of all wells in study area were used to reveal the distribution maps of them. Figure 7 shows the porosity distribution in the upper 100 meters of the basement. The porosity distributions was correlated with seismic data interpretation for faults and fracture zones identification (Figures 7, 8 and 9) because the zones of good porosity are related to faults. Hence, 3-D structural models are able to constructed reliably. (a) 0.02 0.04 0.07 0.09 0.00 0.11 0.00 1 9 17 25 33 41 49 57 65 70 Pattern # Error Training Data RMS Error Vs. Pattern for all Nodes (b) 0.02 0.04 0.07 0.09 0.00 0.11 0.00 1 5 9 13 17 21 25 27 Pattern # Error Testing Data RMS Error Vs. Pattern for all Nodes Figure 5 RMS errors as a function of training and testing data set patterns of porosity NN for (a) the training data set; (b) the testing data set Density NPHI Sonic LLS MSFL LLD P P o o r r o o s s i i t t y y o o r r P P e e r r m m e e a a b b i i l l i i t t y y Hidden layer Output layer Connection weights Processing elements ( PE ) In p ut la y e r Hi ngh khoa hc và công ngh ln th 9, Trng i hc Bách khoa Tp. HCM, 11/10/2005 148 0 0.03 0.06 0.09 0.12 0.15 0.18 0.21 0.24 0.27 0.3 0.33 0.36 0.39 2165 2170 2175 2180 2185 2190 2195 2200 2205 2210 depth (m) p orosit y ( % ) CORE porosity NN porosity LOG porosity Figure 6 Comparison of porosity predicted by NN and conventional log method to that of core samples in a selected well (A2-VD-1X) Figure 7 Porosity distribution combined with seismic data to predict major faults and fractured zones in the upper 100 meters of the basement Figure 8. Structure of the top basement corrected with porosity distribution in A2- VD prospect Figure 9. Structure of the top D horizon correctedwith porosity distribution in A2-VD prospect 4. CONSTRUCTION 3-D STRUCTURAL MODELS OF A2-VD PROSPECT In this study, we focused to construct 3-D model of the top basement and E sequence, because that are main targets of oil and gas production in this prospect (JVPC, 2001). A 3-D structural model was prepared using a PC-based program. The basement is modeled as a Pre-Tertiary formation with a maximum depth of 3500 ms and minimum depth (highest point) of 2100 ms. Figure 10 shows the 3-D structural model for the top of the basement. The faults strongly segmented the basement with the location is nearly as the same as the location of high porosity distribution from NN. Re-activation of the faults in the Eocene and Lower Oligocene results in basement uplift, completely truncating the E sequence (Figure 11). Fault activities were interpreted meticulously from the seismic sections. This uplift shifts the top of the E sequence from 3000 ms to 2200 ms, and the truncation eliminates the E sequence from the basement high. Fault locations from these structural maps are quite coincident with the porosity locations obtained by NN. Hi ngh khoa hc và công ngh ln th 9, Trng i hc Bách khoa Tp. HCM, 11/10/2005 149 Figure 10. 3-D view of faults and the top basement in A2-VD prospect Figure 11. 3-D relationship between the basement high and the E sequence in A2-VD prospect 5. CONCLUSIONS By using neural network, reliability porosity values can be predicted directly from well log data. And then, porosity distribution maps were combined with seismic data interpretation to predict faults and fractures zones. Hence, 3-D structural models were constructed reliably. The 3-D structure models and structural maps prepared based on 3-D seismic data and well log data for the A2-VD prospect have revealed the detail subsurface structure of this area. This research provides useful data for oil field development in offshore Vietnam, and will be supplemented in the near future with more detailed research on the fault distributions in this area and also illustrated the influence of India- Eurasian to the tectonics of Vietnam. These studies thus form the basis for hydrocarbon potential assessment in this area, and provide fundamental data for planning of oil prospects. Acknowledgements Gratitude is extended to Japan Vietnam Petroleum Company (JVPC) and PetroVietnam for providing the data for this research. REFERENCES 1. Badley, M. E.,. Practical seismic interpretation. International Human Resources Development Corporation, Boston, USA (1985). 2. Japan Vietnam Petroleum Company (JVPC). Report for the Block 15-2 prospect, southern offshore Vietnam (2000), pp. 41- 42. 3. Japan Vietnam Petroleum Company (JVPC). Report for the Block 15-2 prospect, southern offshore Vietnam (2001), pp. 103- 104. 4. Lippman, R. An introduction to computing with neural nets, IEEE Transactions on Acoustics. Speech and Signal Processing, Vol. 4 (1987), pp. 4-22. 5. Long, H.T., Huyen, B.T.T., El-Qady, G., Ushijima, K. Porosity & permeability estimation in A2-VD oil prospect, southern offshore Vietnam using artificial neural networks. Proceedings of Second Annual Petroleum Conference and Exhibition, Egypt (2005), pp. 16. Hi ngh khoa hc và công ngh ln th 9, Trng i hc Bách khoa Tp. HCM, 11/10/2005 150 6. PetroVietnam. Report of Cuu Long basin, southern offshore Vietnam (1998), pp. 7-8. 7. Phuong, L.T. Lithofacies and depositional environments of the Oligocene sediments of the Cuulong basin and their relationship to hydrocarbon potential. Proceedings of an International Conference on Petroleum Systems of Southeast Asia & Australia, Jakarta, May 21-23, IPA (1997), pp. 531- 538. 8. Vail, P. R., Mitchum, R. M., Jr. and Thompson, S., III. Seismic stratigraphy and global changes of sea level, Part 2, The depositional sequence as a basic unit for stratigraphic analysis: in Seismic Stratigraphy Applications to Hydrocarbon Exploration, Payton, C. E. (Ed.). AAPG Memoirs, Vol. 26, (1977), pp. 53-62. 9. Werbos, P.J. The Roots of Back- Propagation. John Wiley & Sons, Inc (1994), pp. 115-127. . 11/10/2005 145 XÂY DNG MÔ HÌNH CU TRÚC 3 CHIU CHO CU TO DU KHÍ DA VÀO TÀI LIU A CHN VÀ A VT LÝ GING KHOAN CONSTRUCTING A 3- D STRUCTURAL. hc Bách khoa Tp. HCM, 11/10/2005 148 0 0. 03 0.06 0.09 0.12 0.15 0.18 0.21 0.24 0.27 0 .3 0 .33 0 .36 0 .39 2165 2170 2175 2180 2185 2190 2195 2200 2205