SPE 86999 Advances in Generating and Ranking Integrated Geological Models for Fluvial Reservoir

M. A. Al-Khalifa, SPE, Saudi Aramco, ASP, The University of Adleaide

Copyright 2004, Society of Petroleum Engineers Inc. This paper was prepared for presentation at the SPE Asia Pacific Conference on Integrated Modelling for Asset Management held in Kuala Lumpur, Malaysia, 29-30 March 2004. 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.

Introduction
Development plans for oil and gas reservoirs require huge financial investment. The decision for making these investments is based on many factors, among them the reservoir performance predictions. These predictions answer questions such as how much this reservoir can produce and for how long and they are generated from a simulation model, which is based on a geological model. Therefore to get reliable reservoir performance predictions, a reliable geological model is needed. However, there are many challenges encountered in the process of building a reliable geological model. An important challenge is the limited amount of information available to geologists. Another challenge is the nonuniqueness problem, where several models can fit the same data and give different future forecasts. To overcome these challenges, methods have been developed that integrate different data such as well logs, 3-D seismic and conceptual geological models into a comprehensive geological model. Noticeably, the use of conceptual geological models is increasing in integrated geological modeling, due to the recognition of their role in adding geological knowledge in geological models. Also, it is well known that stochastic modeling methods alone cannot accurately predict lithological distribution. The conceptual geological model is based on the geologist's knowledge, which results from interpretations of well data using principle geological rules. Unfortunately, due to integrated geological methods alone are not enough to solve the non-uniqueness problem. For these reasons new methods have been developed to rank geological models to help in selecting the best possible one. Advances in Reservoir Characterisation Studies Field development projects require huge financial investment and they are typically based on predictions of future reservoir performance. These predictions are generated from a reservoir simulation model, which is based on a geological model. Consequently, the reliability of the simulation model is highly dependent on the accuracy of the geological model (Jian et al. 2002). Several challenges, however, are encountered in the creation of a reliable geological model. An important challenge is to build a model with a very limited amount of subsurface

Abstract
Development plans for Oil and gas reservoirs require huge investments. The decision for making these investments is based on many factors among them the reservoir performance predictions. To get reliable reservoir performance predictions, a reliable geological model is need. However, there are a lot of challenges encountered in the process of building a reliable geological. An important challenge is the limited amount of information available to geologists. Another challenge is the non-uniqueness problem, in other words several models can fit the same data and give different future forecasts. Moreover, stochastic modeling methods alone cannot predict accurately lithological distribution. New advances have been made in building and ranking geological models to over come all of these problems. This paper reviews the current status of integrated reservoir modeling in general and present advances in integrated geological modeling for fluvial reservoir. It covers advances in three important areas: reservoir characterisation of fluvial reservoirs, stochastic modeling and geological models ranking. These new advances have made it possible to generate the best possible geological model to be used in reservoir simulation. The first part focuses on advance in building a fully integrated reservoir characterisation study and exploring the different possible reservoir depositional settings for fluvial settings. In the second part covers advance in the utilization of stochastic modeling such as the use of object oriented algorithms to generate multi realizations that match modern analogs. The third part covers ranking geological models by static and dynamic methods. Static methods include the match of input and output statistics, while the dynamic methods include streamline simulation methods.

www.petroman.ir

SPE 86999

information available. Other challenges include defining flow units in the reservoir and identifying reservoir heterogeneities that effect fluid flow (Jian et al. 2002). A flow unit is defined as a volume of reservoir rock, which has very similar geological and petrophysical properties. Fluid flow properties within this rock are different from fluid flow properties in other rock (Aminian et al. 2002). To overcome these challenges, geoscientists conduct integrated reservoir characterisation studies. The literature contains well documented integrated reservoir studies which have helped in selecting the appropriate plan for reservoir development by improving the quality of the geological model (Meng et al. 2002). These studies have demonstrated their importance in reservoir management and future development. One of the important breakthroughs in reservoir characterisation in recent years is the use of high-resolution sequence stratigraphy, 3D seismic and modern analogues to came up with a better 3-D reservoir image which helps in exploration and development (Lang et al. 2002) and (Strong et al. 2002). Reservoir characterisation studies are usually done to address existing reservoir problems such as unexpected water production and optimisation of future development plans. Moreover, adequate reservoir characterisation studies may result in an increase in oil reserves, which are usually referred to as original oil in place (OOIP). Marquez at al. (2001) conducted such a study which resulted in an increase of 44% in OOIP. Furthermore, they have defined new areas with significant reserves, which will be targeted for future development. The success of their study was dependent on integrating data from different sources. This process is referred to as data integration in reservoir characterisation. Data integration has become an important part of reservoir characterisation. It involves merging of information from different branches of geosciences and petroleum engineering (Tahmhane et al. 1999). Data with different measurement scales such as conceptual geological models, 3-D seismic, flow-meters and well tests have been effectively used in reservoir characterisation studies (Landa et al. 2000; Qassab et al. 2000). Usually the outcome of most integrated reservoir characterisation studies is an integrated 3-D geological model that represents the geologists understanding of the reservoir. This model will have a large impact on the calculation of OOIP and the expected recovery. Additionally, it plays an important role in locating new producing and injecting wells, and the accuracy of this model is in part dependent on the accuracy of the conceptual geological model. Understanding Uncertainties in Geological Modeling Uncertainty is defined as the error difference between estimation and truth and it is very difficult to measure directly, because in most cases the truth in not known (Isaaks and Srivastava 1989). Uncertainty in 1-D can be very low, whereas in 3-D it increases significantly (Massonnat 2000). This idea is illustrated by the following example. Well logs at the well location can be considered as 1-D, whereas a conceptual geological model can be considered as 3-D. There can be few disagreements between geologists while

interpreting data at well location; however, disagreements will increase while interpreting the conceptual geological model. The increase in disagreement is the result of the increase in uncertainty. Uncertainty in the integrated geological model may come from many sources such as structural framework and petrophsycial properties (Ligero et al. 2003). However, an important source of uncertainty is the conceptual geological model, which is dependent on geologist knowledge. This knowledge is different from one geologist to another and it is highly influenced by experience and background (Tahmhane et al. 1999). In most of the integrated geological studies, such as in Qassab et al. (2000), only one integrated geological model is presented and used later in building the reservoir simulation model. This integrated geological model does not necessarily represent the actual reservoir; in fact it represents the geologists imagination of the reservoir. Therefore, there is a need to consider several possible models before deciding on which model to use. Stochastic modeling methods have the ability to create many equi-probable realisations from the same geological data. Unfortunately, in the petroleum industry this feature has not been fully utilized and only a few number of realizations are selected for history matching. This is due to several limitations, among them the long time required to do history matching, which make it impossible to do it on all realizations. Moreover, usually geological models are upscaled for simulation, which results in the exclusion of fine sale critical geological heterogeneities (Ates et al. 2003). This requires the section of the best possible realization with the least uncertainty before upscaling. Currently there are no agreed-upon procedures to rank geological models. Ranking could be made by static methods such as pore volume or by dynamic methods such sweep efficiency (Ates et al. 2003). Static ranking methods are widely used compared to dynamic methods, because dynamic methods require the availability of a significant production history. However, some static ranking parameters such as the match of input and output statics might not give an indication of flow behaviour of the reservoir. Alternatively, dynamic ranking based on matching of existing production profiles will give a good indication of a reservoirs future performance. Dynamic ranking can be done by conventional simulator; however this is not practical due to the extensive time needed for such a process (Scheepens 2003). An alternative for a conventional simulator is a streamline simulator. Hence, in cases where production history is available, a streamline simulator is a very valuable tool that can help in ranking geological models. Advances in Streamline Reservoir Characterisation Simulation Appliations in

Streamline simulator has the advantage of being much faster than a conventional finite difference method simulator, which has made it an ideal tool to rank large geological models. One factor that contributes to its speed is the solution method for

www.petroman.ir

SPE 86999

fluid saturations, which is done along 1-D streamlines and mapped into the model cells by simple interpolation. These streamlines are generated based on time of flight principle, which is defined as the time required for a particle to move from an injector to a producer along a single streamline (Ates et al. 2003). Streamline simulation process is conducted through several steps. The first step in streamline simulation is distributing fluid saturation and computing reservoir pressure at each cell in the model, which is the same first step done in any conventional simulator (Ates et al. 2003). This step is followed by generation of streamlines between each injector and producer in the simulation model. Afterwards, saturation calculation is performed along these streamlines and mapped to every cell in the model. With the completion of this step another time step is taken to update pressure and remap streamlines and execute saturation calculations. This entire process is repeated until all time steps are completed (Lolomari et al. 2000). As the number of time steps required for updating pressure increases, computer run time will increase (Ates et al. 2003). The number of time steps required to update the pressure will depend on the production profile for the wells. If most wells have constant production profile, then fewer time steps are required and vice versa, so selecting an optimum number of time steps will need a statistical analysis of production data. Lolomari at al. (2000) believe that streamline simulation methods are appropriate tools for ranking geological models for several important reasons. An important one is the capabilities of streamline simulators to handle a geological model as it is without upscaling it. This preserves fine scale geological heterogeneities in the model, which may have significant impact on fluid flow. Another reason results from the combination of two important developments in streamline algorithms and computers. Streamline simulations algorithms became more accurate and at the same time computers became more powerful. This means that larger geological models can be simulated in shorter times with much higher degrees of accuracy. Gilman et al. (2002) have used streamline based simulation methods to rank large-scale geological models and to address uncertainty issues. The authors stated that streamline simulation methods are being recognized as important components of integrated reservoir characterisation studies and believe that these methods are well suited to solve largescale problems provided that a number of simplifying assumptions can be made in regard to the flow physics. In their paper they have used this technique to rank a number of geological models. However, they did not use true well locations or true production and injection rates. Also, the ranking has been done by recovery only, which is dependent on many factors, such as number of wells and their location. Integrated reservoir characterisation studies have become an essential part of hydrocarbon reservoir management and development. Generating and ranking integrated geological

models would have a huge impact on the accuracy of reservoir simulation models.

Advanced Methodology for Reservoir Characterisation for a Fluvial Reservoir

In order to achieve a reliable and adequt geological model for a fluvial reservoir, the following steps should be done. First, a reservoir characterisation study is conducted, and then several integrated geological models are generated. The final step would be ranking the newly generated geological models to select representing geological models for reservoir simulation. Reservoir Characterisation The first step in reservoir characterisation is to collect and quality check all available static data such as well logs and cores and dynamic data such as well tests. This is an important step in the study, because the subsequent analysis and work will be dependent on the accuracy of these data. It is very crucial to identify any problem with the data before the startup of this project. If problems are not recognized, they will alter the interpretations and will be very difficult to correct. Afterwards, all quality-controlled data will be loaded into a database for subsequent geological analysis. The second part in this phase involves the examination of the quality-controlled data to develop a depositional model and a sequence stratigraphic framework for this reservoir. This includes isopach and structure mapping of genetic intervals. A key outcome is to determine the relationship between reservoir depositional facies and reservoir flow units to achive understanding between reservoir geometry, connectivity and depositional architecture. Flooding surfaces and sequence boundaries can be used to package the important successions into genetic intervals, bounded by unconformities and flooding surfaces, equivalent to depositional system tracts. Facies maps are generated for each sequence and they can be combined to generate a deterministic conceptual geological model. Stochastic Modeling The aim of any Stochastic Modeling is to generate a 3-D petrophysical properties models for the reservoir based on the conceptual geological model generated in the phase-1. The petrophysical properties include porosity, permeability and water saturation. Different stochastic modeling algorithms are used to generate a range of petrophsycial properties. Object based modeling algorithms are preferred to generate facies models based on specific conceptual geological models (e.g. meandering channels) compared to pixel based modeling methods (Cosentino 2001). Porosity models will be generated using a facies based geostatistical algorithm. Permeability models will be based on

www.petroman.ir

SPE 86999

an appropriate transform between core porosity and core permeability. Furthermore, it will be controlled by the facies and the porosity models, and conditioned well test data, if available. The goal for this is to make permeability values in the model as close as possible to permeability calculated from well tests. Permeability values from well tests give better estimates for reservoir permeability than core plug measurements. This is due to the scale difference in measurements between core plugs, which are a few inches in diameter, and well tests, which represent a much larger reservoir volume. Ideally, saturation will be mapped based on J-function, which are facies dependent. If J-functions are not available, saturation will be mapped based on available saturation logs. Ranking Geological Models Just building models in not enough, these models should be ranked and theire uncertainty should be assessed based on static and dynamic methods. One of the common ranking static methods is the statistical correlation between input and output data. Theoretically all stochastic modeling algorithms are able to reproduce statistics of the input data (Isaaks and Srivastava 1989). If there is a poor correlation between input and output data statistics and this difference cannot be justified, then this means there is a problem with the model. Although this is a very simple and straightforward checking method, not every geological modeler does it. Another simple ranking method is 3-D visualization of the geological model. During this process, the models will be checked for correct geometry and orientation of main geological bodies and for any abnormal features. Ranking can be also done on pore volume numbers. Even though all stochastic models are generated from the same data and have the same probability of occurring, they have different pore volumes. Some models will have a very high number and others will have a very low number and it is a common practice to avoid extreme values. The dynamic ranking will be done using a streamline simulator. The first step will be selecting an adequate time step number to optimize the time needed for simulation runs. This will be done by simplifying well production profiles by taking a moving average for production and injection rates, while capturing the major production events. Several time steps numbers will be tried first to get the optimum number. After this, different models such P(10), P(50) and P(90) will be simulated through streamline simulator. The simulated results will be analysed according to their match with observed field data. The analysis will be conducted by generating cross-plots between observed and simulated data for water break through time per well. Also, fluid movement, sweep efficiency and total production in the reservoir will be plotted for all simulated models. Finally, the best geological model will be selected for use in future filed development and management.

Conclusions
To achieve a reliable geological model for a fluvial reservoir, the following steps are highly recommended: In order to meet these aims, the following objectives need to be addressed: 1Determine and describe the geological depositional setting, important facies types through geological analysis and high-resolution sequence stratigraphy 2Identify and map possible facies types, distribution and geometry for every reservoir sequence. This requires the examination of all available static and dynamic data and the use of modern depositional setting analogs 3Build several fully integrated 3-D equiprobable geological models based on conceptual geological model and using stochastic modeling methods with object based modeling methodologies 4Rank and assess the uncertainty of these integrated geological models by using static and dynamic methods. Select different models such P(10), P(50) and P(90) for future use in conventional simulation and for future field development and management uses.

References Aminian, K., Thomas, B. and Bilgesu, H. I. (2002). A New Approach for Reservoir Characterization. SPE 78710, SPE Eastern Regional Meeting, Lexington, Kentucky, USA. Ates, H., Kelkar, M. and Datta-Gupta, A. (2003). The Description of Reservoir Properties by Integrating Geological, Geophysical and Engineering Data, The University of Tulsa and Texas A&M University, Joint Industry Project: pp. 3-16. Cosentino, L. (2001). Integrated Reservoir Studies. Paris, Editions TECHNIP. Gilman, J. R., Meng, H.-Z., Uland, M. J., Dzurman, P. J. and Cosic, S. (2002). Statistical Ranking of Stochastic Geomodels Using Streamline Simulation: A Field Application. SPE 77374, SPE Annual Technical Conference and Exhibition, San Antonio, Texas, USA. Isaaks, E. H. and Srivastava, R. M. (1989). An Introduction to Applied Geostatistics. New York, USA, Oxford University Press. Jian, F. X., Larue, D. K., Castellini, A. and Toldi, J. (2002). Reservoir Modeling Methods and Characterization Parameters for a Shoreface Reservoir: What Is Important for Fluid Flow Performance? SPE 77428, SPE Annual Technical Conference and Exhibition, San Antonio, Texas, USA. Landa, J. L., Home, R. N., Kamal, M. M. and Jenkins, C. D. (2000). "Reservoir Characterization Constrained to Well-Test

www.petroman.ir

SPE 86999

Data: A Field Example." SPE Reservoir Evaluation and Engineering 4(3): 325-334. Lang, S. C., Ceglar, N., Forder, S., Spencer, G. and Kassan, J. (2002). "High Resolution Sequence Stratigraphy, Reservoir Analogus, and 3d Seismic Interpretation-Application to Exploration and Reservoir Development in the Baryulah Complex, Cooper Basin, Southwest Queemsland." APPEA Journal 42: 512-521. Ligero, E. L., Maschio, C. and Schiozer, D. J. (2003). Quantifying the Impact of Grid Size, Upscaling and Streamline Simulation in the Risk Analysis Applied to Petroleum Field Development. SPE 79677, SPE Reservoir Simulation Symposium, Houston, Texas, USA. Lolomari, T., Bratvedt, K., Crane, M., Milliken, W. J. and Tyrie, J. J. (2000). The Use of Streamline Simulation in Reservoir Management: Methodology and Case Studies. SPE 63157, SPE Annual Conference and Exhibition, Dallas, Texas, USA. Marquez, L. J., Gonzalez, M., Gamble, S., Gomez, E., Vivas, H. A., Bressler, H. M., Jones, L. S., Ali, S. M. and Forrest, G. S. (2001). Improved Reservoir Characterization of a Mature Field through an Integrated Multi-Disciplinary Approach. Ll-04 Reservoir, Tia Juana Field, Venezuela. SPE 71355, SPE Annual Technical Conference and Exhibition, New Orleans, Louisiana, USA. Massonnat, G. J. (2000). "Can We Sample the Complete Geological Uncertainty Space in Reservoir-Modeling Uncertainty Estimates?" SPE Journal 5(1): 46-59.

Meng, H. Z., Godbey, K. S., Gilman, J. R. and Uland, M. J. (2002). Integrated Reservoir Characterization and Simulation for Reservoir Management Using a Web-Based Collaborative Technical Workflow Manager. SPE 77673, SPE Annual Technical Conference and Exhibition, San Antonio, Texas, USA. Qassab, H. M. A., Fitzmaurice, J., Al-Ali, Z. A., Al-Khalifa, M. A., Aktas, G. A. and Glover, P. W. (2000). Cross-Discipline Integration in Reservoir Modeling: The Impact on Fluid Flow Simulation and Reservoir Management. SPE 62902, SPE Annual Technical Conference and Exhibition, Dallas, Texas, USA. Scheepens, C. C. J. (2003). 3d Modeling Using Multiple Scenarios and Realizations for Pre-Reservoir Simulation Screening. SPE 82022, Hydrocarbon Economics and Evaluation Symposium, Dallas, Texas, USA. Strong, P. C., Wood, G. R., Lang, S. C., Jollands, A., Karalaus, E. and Kassan, J. (2002). "High Resolution Palaeogeographic Mapping of the Fulvial-Lacustrine Patchawarra Formation in the Cooper Basin, South Australia." APPEA Journal 42: 65-81. Tahmhane, D., Wang, L. and Wong, P. M. (1999). The Role of Geology in Stochastic Reservoir Modeling: The Future Trends. SPE 54307, SPE Asia Pacific Oil and Gas Conference and Exhibition, Jakarta, Indonesia.

www.petroman.ir