Uncertainty Analysis in Geomodeling: How Much Should We
Know About What We Dont Know?
Y. Zee Ma, Schlumberger, Denver CO, USA This paper has been selected for presentation for the 2014 Gussow Geosciences Conference. The authors of this material have been cleared by all interested companies/employers/clients to authorize the Canadian Society of Petroleum Geologists (CSPG), to make this material available to the attendees of Gussow 2014 and online. ABSTRACT As the demand for hydrocarbon resources continues to grow, reservoir modeling and uncertainty analysis have become increasingly important for optimizing field development. Optimal valuation and exploitation of a field requires a realistic description of the reservoir, which in turn requires reservoir characterization and modeling, and quantification of the uncertainty by integrating multidisciplinary data. An integrated approach for reservoir modeling helps bridge the traditional disciplinary divides and tear down interdisciplinary barriers, leading to better handling of uncertainties, and improvement of reservoir modeling for its use in the petroleum industry. Uncertainty analysis should be conducted for investigational analyses, and for decision analysis under uncertainty and risk. Constructing a realistic reservoir model, and reducing and quantifying the uncertainty are the topics discussed in this article. INTRODUCTION Reservoir characterization and modeling have seen significant leaps in the last two to three decades, driven by the development of computational horsepower, advances in seismic technology, logging tools, geological understanding of depositional systems and natural fracturing of subsurface systems, and applications of probabilistic methods. It has evolved from fragmentary pieces into a discipline of geoscience applications for the petroleum industry, from university research to value-added resource developments, from 2D mapping of structures and reservoir properties to 3D geocellular representations of hydrocarbon reservoirs, and from dealing with discipline-specific problems to integrated multidisciplinary reservoir modeling. The division of tasks between geologists and reservoir engineers in the early time was that geologists explored for hydrocarbon resources, and engineers produced hydrocarbons from the reservoirs. This separation of the tasks was based on the low usage of fossil fuel relative to the amount of the resources in the ground and high reservoir quality of formations. As hydrocarbon consumption has dramatically increased worldwide, reservoir management has become more and more important. Integration of geology with reservoir engineering has become critical for better reservoir management (Haldorsen and Lake, 1984, Ma et al, 2008), especially for unconventional reservoirs (Du et al., 2011; Cipolla et al., 2012). Geology has traditionally been considered as descriptive, although some quantitative branches including geophysics, 2 mathematical geology and geostatistics have significantly increased the breadth of geoscience. We believe that in the future, most geoscientists will conduct geologic or reservoir modeling as routine work. By performing geologic modeling, geoscientists can test and quantify their geologic concepts and hypotheses. In doing so, they use data to prove or disprove the concepts, and use statistics and geostatistics to resolve inconsistencies in various data and integrate them in a coherent manner (Ma, 2010). As a result of the convergence of descriptive geology and quantitative geology, geoscientists need to use the modeling as a process for understanding the reservoir, not just producing a numeric model. The convergence should make reservoir modeling a synonym of reservoir characterization. Reservoir management and field development planning are important for maximizing the economics of the field, which requires accurate reservoir characterization. Reservoir modeling was the missing link between geosciences and reservoir engineering in field development before the mid- 1980s. Since then reservoir characterization has shown significant values in identifying both prolific and marginal reservoirs, extending the production life of existing fields and increasing the hydrocarbon recovery of reservoirs. Successful reservoir characterization projects typically show high degree of integration. In fact, reservoir modeling is the best way to integrate all the data and disciplines, and the only way in which all the data and interpretations come together into a single 3D numeric representation of a reservoir. In integration, the data include not only quantitative data such as well-logs, cores, and seismic data, but also the geologic concepts and descriptive interpretations (Mallet, 2002; Ma, 2009; Cao et al., 2014). A reservoir is the result of geologic processes and is not randomly generated. However, the complexity of subsurface reservoir properties coupled with limited data leads to substantial uncertainty in a reservoir model. Uncertainties can be mitigated by gaining more information and/or using better science and technology. How much uncertainty should be mitigated depends on the needs of decision analysis for reservoir management and the cost of information. Uncertainty analysis should be conducted for investigational analyses, and for decision analysis under uncertainty and risk. Knowing what needs to be known and what can be known should be the main focal points of uncertainty analysis in reservoir modeling. RESERVOIR MODELING A good model can advance fashion by ten years. Yves Saint-Lauren The essence of reservoir modeling lies in using all the available data to build an accurate reservoir representation that is fit-for-purpose to the fields development needs. In a significant hydrocarbon resource asset, a good reservoir model can be an essential element for increasing the production and extending the field development life for years. Why build a reservoir model? The most common use of reservoir models is to provide a 3D numeric input to reservoir simulation. Reservoir modeling and simulation provide a basis for maximizing economic value for field development and operational decisions. The typical motivation for reservoir simulation is to increase profitability through better reservoir management. These include development plans for new fields and depletion strategies for mature fields. Reservoir modeling and simulation can address liquid (oil, and water) and gas volume forecasting, decline analysis, infill drilling uplift, secondary or tertiary recovery options, well management strategies, water/gas handling strategies and facility constraints, contact movement, liquid dropout, reservoir surveillance strategies, injection strategies, and well and completion designs. Reservoir modeling and simulation can also be used for reserve confirmation, equity determination, or support for funding large projects. Traditional mapping and cross-section methods worked relatively well for homogeneous reservoirs, but they tend to overestimate sweep efficiency for heterogeneous reservoirs. These methods may significantly under- or over-estimate inplace hydrocarbon resources because they lack 3D examination of reservoir heterogeneities. Reservoir modeling and simulation provide powerful tools for more accurate reservoir description and hydrocarbon production forecasting (Dubrule, 1989; Yu et al., 2011), and can help reservoir management and field development. Accurate reserve assessment through reservoir modeling and simulation could help reduce cost and increase recovery. 3 Besides reservoir simulation, reservoir modeling itself can be used as support for reservoir surveillance activities, such as monitoring fluid contacts and reservoir pressures, analyzing fault transmissibility and performing production fault seal analysis. It can also be used for an accurate determination of stock-tank original oil in place (STOOIP), for example, by incorporating capillary pressure effects, new opportunity identification and prioritization, well planning and well placement optimization, visualization and communication of the detailed 3-D reservoir architecture and properties between various disciplines, reviewing data and their quality controls, resolving inconsistencies between various disciplines, support for time lapse seismic analyses, for example, by identifying bypassed oil, and reservoir uncertainty and risk analysis. Reservoir modeling is critical to rapid successful commercialization of discovered and undeveloped hydrocarbon resources, as well as to optimizing depletion of mature fields. As a rapidly growing discipline, reservoir modeling has become an integral part of the field asset management. For large and capital-intensive development projects, reservoir modeling and simulation have almost become a necessity. Even for small to medium reservoirs, modeling and simulation can enhance efficient development, and depletion planning, and potentially increase reserves and yield cost saving. Modeling can also help in moving static resources to reserves. In some cases using traditional 2D mapping methods, reserves have originally been grossly overestimated, leading to false optimism. Expensive modern platforms may be installed, but later may be found under used because of the over-estimation of the resource. On the other hand, some large oil fields have been mistakenly farmed out because of the underestimation of the resource by traditional methods, leading to false pessimism. In many of these cases, reservoir modeling could have helped make the decisions more objectively and realistically. Reservoir modeling is a critical link between seismic interpretation and reservoir simulation. Without reservoir modeling, integrated approaches to E&P solution and accurate reservoir evaluation are almost impossible. Building a reservoir model used to be very costly, but availability of increasingly versatile and sophisticated software packages has made reservoir modeling much more efficient and affordable.