Summary History matching of a steam-assisted-gravity-drainage (SAGD) reservoir requires a large ensemble size for proper uncertainty assessment, which ultimately results in high computational cost. Therefore, it is necessary to reduce the number of realizations for SAGD-reservoir simulation purposes. In this paper, a novel sampling method (based on the probability-distance-minimization method) to generate an initial ensemble of reduced size is discussed. This method considers multiple static measurements and geological properties and uses Kantorovich distance to quantify the probability distance between the original ensemble and the reduced ensemble, which is later optimized by use of the mixed-integer linear-optimization (MILP) technique. To show the effectiveness of the method, we have shown history matching of an SAGD reservoir using the smaller size initial ensemble derived from the proposed method and compared with the original ensemble. For history matching, the ensemble Kalman filter (EnKF) has been used because of its ability to assimilate data for large-scale nonlinear systems. Results are compared with several other methods, such as importance sampling, kernel K -means clustering, and sampling by use of orthogonal ensemble members. The robustness and usefulness of each method for generating an improved initial ensemble of reduced size are analyzed on the basis of two criteria: (1) Does the smaller ensemble retain the same statistical distribution characteristics as the original ensemble, and (2) does the smaller ensemble improve the performance of history matching? In general, we conclude that the improved, smaller initial ensemble created by use of the proposed method retains the best statistical characteristics of the original ensemble. Also, it provides better performance compared with other ranking methods in sampling and history matching using EnKF. Finally, the proposed method can reduce the computing cost significantly without compromising uncertainty in the forecast model, which allows for real-time updating at smaller time intervals.

Summary This paper presents the applications of the ensemble-Kalmanfilter (EnKF) inverse-modelling technique to petroleum-reservoir characterization of thermally operated oil fields in northern Alberta, Canada. The EnKF is applied to 2D- and 3D-case studies based on the steam-assisted-gravity-drainage heavy-oil-extraction method. The modelling technique integrates effectively both static and dynamic data (petrophysical core data from wellbores, continuous temperature data measured by thermocouples, and 4D-seismic attributes) into a petroleum-reservoir model. Assimilated secondary information provides better insight into geologic properties of a reservoir and improves production forecasting. The method performs well for linear or slightly nonlinear systems that follow a Gaussian distribution, but shows worse performance for nonlinear or non-Gaussian systems. Integration of a large amount of data with a small number of realizations leads to ensemble collapse because of insufficient degrees of freedom. Increasing the ensemble size is a solution, but by increasing forecasting time. To overcome these issues and reduce computational time, matrix localization techniques and a shortcut based on replacement of the model realizations with their mean in forecasting are suggested. Even though the EnKF has been proved to be simple in implementation and effective for modelling of continuous linear systems, special care should be taken for modelling nonlinear systems, including categorical variables (e.g., geologic facies).

Summary A new recovery process, thermal-assisted gravity drainage (TAGD), is described for the recovery of bitumen from heterogeneous carbonate reservoirs. The process is shown to be thermally efficient in lower-porosity formations and is less impacted by the complex porosity distributions, that characterize carbonate rock. The TAGD process heats the reservoir by means of a pattern of horizontal heater wells to allow bitumen to flow by gravity into a producer placed at the base of the reservoir. High thermal efficiency is achieved by heating the reservoir to 120–160°C. The energy for this process is transferred to the reservoir by heat conduction from electrically heated mineral-insulated cables. Precise control of reservoir heating is accomplished by optimally placing heater wells within the formation and adjusting the power in each heater for efficient energy usage. Voidage replacement is achieved by solution-gas evolution and connate-water vapourization. The target of this new technology is the thick, highly permeable, and vertically continuous Leduc bitumen reservoir located in the Liege area of north-central Alberta. In addition to improved energy efficiency, the TAGD process does not require water for steam generation, thus reducing the size and initial capital cost of the surface-processing facilities.

Summary Thermal gas/oil gravity drainage (T-GOGD) is an attractive enhanced-oil-recovery method applicable to naturally fractured reservoirs (NFRs). The process is applied successfully in Qarn Alam, a heavy-oil field in Oman. This paper presents a case study featuring dynamic-modelling optimization and uncertainty-analysis workflow for T-GOGD in a bitumen-bearing fractured reservoir on the basis of realistic 3D fracture characterization. In T-GOGD, the fractures are displaced to steam to provide a (matrix) gravity-drainage potential while heating the reservoir at the same time. One of the recovery mechanisms associated with T-GOGD is thermal expansion, which can result in high initial rates, but may cause plugging of the fracture system in the case of extraheavy oil (bitumen) if the expanded oil cools down before it is being produced. This situation requires short-distance well configurations and/or steam-stimulation cycles to establish communication. In an NFR, steam vapour occupies the fracture system while oil drains through the matrix, increasing the area for heat transfer with respect to the steam-chamber case; the process therefore differs significantly from steam-assisted gravity drainage, and a different production function applies. Shell’s in-house reservoir simulator MoReS with advanced dual-permeability capability is used to model development of T-GOGD in a bitumen reservoir employing 3D element-of-symmetry (EOS) models. A realistic fracture-characterization and -modelling process is described. The geometrical well configuration and operating schedule and strategy are optimized on an economic function for a deterministic subsurface realization. Using an uncertainty-analysis work flow, cumulative distribution functions of recovery and steam/oil ratio (SOR) are generated. Finally, robust optimization is explored by use of an economic objective function. The study concludes that (1) relatively large well spacing is feasible while injector/producer horizontal well offset is necessary to avoid steam channeling to the producer well, (2) live steamproduction control is a robust operating strategy, (3) performance is most sensitive to matrix permeability and oil viscosity, and (4) vertical fracture connectivity plays an important role in the process performance. T-GOGD has significant potential to develop the bitumen resource in naturally fractured carbonates.

Summary Shale-gas production has gained worldwide attention over the past several years. Production from shale-gas reservoirs requires horizontal drilling with multiple hydraulic fracturing to obtain the most economical production. However, there are high cost and uncertainty because of many inestimable and uncertain parameters (e.g., reservoir permeability, porosity, fracture spacing, fracture half length,fracture conductivity, gas desorption, geomechanics, and existing natural fractures). Therefore, the development of a way to quantify uncertainties and optimization of shale-gas production in an efficient and practical method is clearly desirable. In this paper, we present a user-friendly and efficient framework to obtain the optimal gas-production scenario by optimizing the uncertain factors by integrating several commercial simulators, an economic model, design of experiment (DoE), and response-surface methodology (RSM) with a global optimization search engine. Specifically, we use factorial design to screen insignificant factors and find the most influential design and uncertain factors; then, we use RSM to design over those most influential factors to fit a response surface using net present value (NPV) as the objective function; finally, we identify the most economical production scenario under conditions of uncertainty. Eight uncertain parameters [i.e., porosity, permeability, reservoir thickness, reservoir pressure, bottomhole pressure (BHP), fracture spacing, fracture half-length, and fracture conductivity] with a reasonable range on the basis of Barnett-shale information are investigated. Also, different gas prices are considered for the optimization process. This framework is effective and efficient for hydraulic-fracturing- treatment design and production-scheme optimization in unconventional gas reservoirs. It can contribute to providing guidance for engineers to modify the design of a hydraulic-fracture treatment before the actual fracture treatment.

Summary Every simulation engineer wishes to simulate large full-field models, but historically reservoir simulation of the steam-assisted-gravity- drainage (SAGD) process has been constrained to single-well models up to a single pad. Models of these sizes provide valuable information and have helped to assess the development potential of reservoirs. These models may be used for reservoir management and to support the decision-making process for the design of the initial completion, operating strategy, and multipad wind-down evaluations, and also qualitatively assess the uncertainty in the SAGD forecast under different geological settings. However, in many cases we are left with the question of how multiwell and multipad communications ultimately affect performance at the well-pair scale. Because of technological constraints with computer hardware and simulation technology, running extremely large multipad models has been until recently largely impractical, especially when trying to run multiple scenarios to better understand the impact of geological and operational uncertainty. In this paper, we present a new and practical workflow that makes running extremely large multipad, multimillion-grid-cell SAGD models a reality. The three major steps of the workflow are (1) generating simulation-friendly geomodels, (2) use of experimental design and 3D submodels on the basis of SAGD performance index (SPI) for numerical tuning, and (3) use of 2D cross sections and SPI to develop dynamic grid-refinement-parameter values for the full 3D model. All of these steps are intended to improve the numerical stability and run time of multipad SAGD simulation models. A 24-SAGD-well-pair model with 2.52 million gridblocks was simulated for 10 years of forecast. The reservoir is geologically complex and highly heterogeneous. We discuss some of the important aspects that need to be accounted for when simulating large-scale SAGD models. Using this new workflow, the simulation run time was reduced from 42 days to 7 days on eight central processing units (CPUs)—a six-time speedup. The resulting run time is short enough to facilitate multirealization simultaneous runs using eight CPUs, hence maximizing the throughput and minimizing the simulation cycle time. This new workflow can be easily replicated and, more importantly, automated to reduce engineering time requirements. While this paper focuses on the SAGD process, this methodology is completely generic in that it can be applied to any large data set for any process. Details will differ depending on the process, but the workflow will be the same.

Summary This paper describes a new technique for effective placement of delineationwells on the basis of the change of the uncertainty in the key global-reservesvariable. Uncertainty is summarized through the geostatistical framework. Theauthors develop a numerical and analytical methodology that is tested onsynthetic and real petroleum case studies. The implementation isstraightforward, and the results are promising. A methodology is developed toassist in delineation-well placement. Decisions for new-well locations areassisted with a quantitative measure of the expected reduction in globaluncertainty in the volume of original oil in place (OOIP). The availablerealizations are analyzed and processed to quantify the impact of wellplacement. Variograms and other required statistics are inferred from therealizations. As a result, a gridded map of impact values is produced, fromwhich locations with the highest impact are suggested for new-well locations.Numerical and analytical approaches for the impact-map calculation are proposedand compared. Pros and cons of each approach are summarized. The numericalapproach requires a large number of realizations for effective implementationof the impact map, which might not be practically achievable. On the otherhand, the analytical approach does not require many realizations and producesstable results. In most cases, only variogram models and current well locationsare needed for the analytical impact-map computation. Although computationaltime of this approach largely depends on the model size, some options aresuggested to reduce the cost. The analytical impact calculation is developedfor the OOIP model response, in which the petroleum reservoir is defined as acomplex geological architecture with multiple structural surface constraints.Several case studies, including a real-petroleum-reservoir example, demonstratethe use of the impact map for the assessment of new delineation-well locations.The developed tool is of significant help for well placement.

Summary Placement of steam-assisted-gravity-drainage (SAGD) surface production pads and subsurface drainage areas in the McMurray formation to maximize the economic potential of an area is a challenging problem. The location of surface pads (SPs) and drainage areas has a large impact on their production performance because of several factors, including variation of bitumen in place, variation of the reservoir base surface, vertical conformance, areal conformance, interaction between different drainage areas and pads, and surface hazards. An optimization algorithm is presented to determine the positions and orientations of SPs and drainage areas over a reservoir area; therefore, the potential for economically recoverable bitumen is maximized. The optimization considers either a deterministic model of the relevant properties or multiple realizations to account for uncertainty. Optimization considers all drainage areas simultaneously to ensure joint optimality of an entire set. The algorithm is demonstrated using two realistic examples that show a significant improvement in potential recovery. The algorithm executes in a reasonable amount of computation time, considering the complexity of the problem.

Summary Numerous steam-assisted gravity-drainage (SAGD) optimization studies published in the literature combined numerical simulation with graphical or analytical techniques for design and performance evaluation. Efforts have integrated the simulation exercise with global optimization algorithms. Some studies focused on optimization of cumulative steam/oil ratio (cSOR) in SAGD by altering steam-injection rates, while others focused on optimization of net cumulative energy/oil ratio (cEOR) in solvent-additive SAGD by altering injection pressures and fraction of solvent in the injection stream. Several studies also considered total project net-present-value (NPV) calculation by changing total project area, capital-cost intensities, solvent prices, and risk factors to determine the well spacing and drilling schedule. Optimization techniques commonly used in those studies were scattered search, simulated annealing, and genetic algorithm (GA). However, applications of hybrid GA were rarely found. In this paper, we focused on optimization of solvent-assisted SAGD using various GA implementations. In our models, hexane was selected to be coinjected with steam. The objective function, defined on the basis of cSOR and recovery factor, was optimized by changing injection pressures, production pressures, and injected solvent/steam ratio. Techniques, including orthogonal arrays (OA) for experimental design (e.g., Taguchi's arrays) and proxy models for objective-function (F) evaluations, were incorporated with the GA method to improve computational and convergence efficiency. Results from these hybrid approaches revealed that an optimized solution could be achieved with less central-processing-unit time (e.g., fewer number of iterations) compared with the conventional GA method. Sensitivity analysis was also conducted on the choice of proxy model to study the robustness of the proposed methods. To investigate the effects of heterogeneity in the design process, optimization of solvent-assisted SAGD was performed on various synthetic heterogeneous reservoir models of porosity, permeability, and shale distributions. Our results highlight the potential application of the proposed techniques in other solvent-enhanced heavy-oil-recovery processes.

Summary An ensemble-based history technique has been applied to implicitly estimate three-phase relative permeability curves from production data. A power law representative of relative permeability curves is used. Both endpoints and shape factors of relative permeability curves are included in state vectors that are updated sequentially by assimilating observation data. This method has been validated by accurately evaluating relative permeability in a synthetic reservoir with 2D, three-phase flow. It is shown from the synthetic case that good estimation of relative permeability curves can be obtained by assimilating the observed oil rates, gas/oil ratios, and bottomhole pressures of production wells. Both shape factors and endpoints of relative permeability curves are accurately evaluated; however, a larger ensemble size is needed to avoid filter divergence. Compared with the existing implicit methods, the ensemble-based history matching technique does not require the gradient of the objective function, which makes the technique easy to implement.

CO 2 flooding has gained momentum in the oil and gas industry and might be suitable for approximately 80% of oil reservoirs worldwide based on the oil recovery criteria alone. In addition to miscibility, production performance needs to be optimized to achieve higher sweep efficiency and oil recovery. Although many techniques have been made available for production optimization in the upstream oil and gas industry, it is still a challenging task to optimize production performance in the presence of physical and/or financial uncertainties. In this paper, a new technique is developed to optimize production performance in a CO 2 flooding reservoir under uncertainty. More specifically, potential uncertainties influencing production performance are analyzed and assessed by using the geostatistical technique. This enables us to integrate the available information within a unified and consistent framework and to generate multiple geological realizations accounting for uncertainty and spatial variability. Subsequently, the net present value (NPV) is selected as the objective function to be optimized by using the genetic algorithm, while well rates of the injectors and the flowing bottomhole pressure for the producers are chosen as the controlling variables. In addition, corresponding modifications have been made to accelerate the convergence speed of the genetic algorithm. A field case is used to demonstrate the procedures of the newly developed technique and the optimized results show that the oil recovery and the NPV can be increased by 6.4% and 9.2%, respectively. It is also found that the genetic algorithm is a powerful and reliable search method to optimize production performance of reservoirs with complex structures. Introduction CO 2 flooding is considered as a promising and practical enhanced oil recovery (EOR) process because it not only increases oil recovery, but also reduces greenhouse gas emissions by sequestrating CO 2 in the depleted reservoirs. In practice, CO 2 flooding performance can be greatly affected by the reservoir heterogeneity, which can severely reduce the sweep efficiency, result in early CO 2 breakthrough at the producers, and thus, leave a large amount of bypassed oil in the reservoir(1). Therefore, it is of fundamental and practical importance to optimize production performance of a CO 2 flooding reservoir.

Summary A pragmatic method has been developed to efficiently design the production-injection parameters to optimize the water-alternating-gas (WAG) performance in a field-scale CO 2 -miscible flooding project. The net present value (NPV) is selected as the objective function, while the controlling variables are chosen to be the injection rates, ratios of gas slug size to water slug size (WAG ratio) and cycle time (i.e., the injection time for each gas or water slug) for the injectors and bottomhole pressures (BHPs) for the producers. A hybrid technique, which integrates the orthogonal array (OA) and Tabu technique into the genetic algorithm (GA), is then developed and employed to determine the optimum WAG production-injection parameters. Sensitivity analysis of the WAG parameters on oil recovery is conducted and a field case is finally presented to demonstrate the successful application of the newly developed technique.

Summary In this paper, a new drilling optimization procedure is presented that is designed to improve the drilling efficiency with positive displacement motors (PDMs) and PDC bits. This developed optimization method is based on predicting rate of penetration (ROP) from PDM outputs for any PDC bit design. More specifically, optimization is done for a hole section and optimum values of weight on bit (WOB) and surface RPM are obtained for the section. For given flow rates, estimated values of optimum WOB and surface RPM are used to calculate the corresponding motor differential pressures and the foot by foot ROP values. Also, the method is used to show how improper operational parameter selection can affect total drilling time. A case study was done to consider different PDMs with different lobe configurations and a set of fixed operational parameters. The presented method is verified by generating a confined rock strength log based on drilling data for a previously drilled well in Alberta. This foot-by-foot strength log is compared to a confined rock strength log generated as a follow-up analysis by a commercially available drilling simulator package. Also, a PDM differential pressure log is generated and compared to field-recorded on-bottom differential pressure values. The method's application is best demonstrated by simulating the drilling operation of the Alberta well with three different PDMs. It is shown that consideration of PDM performance/selection in the drilling planning phase will help to perform a safe and cost-effective operation by preventing motor stalls and maintaining highest average ROP for the section. It is also shown that by optimizing WOB and surface RPM values for a constant mud flow rate and predefined bit wear at total depth, a maximum average ROP for the section can be reached for any PDM.

Summary Geostatistical models of reservoir properties are high resolution with many grid cells; it is impractical to use them directly in flow simulation because of computational costs. Upscaling techniques are applied to average small-scale permeability values to larger flow simulation blocks. In cases where unstructured grids are used or the geological features inside the gridblock are not aligned with the block geometry, symmetric or full-permeability tensors arise instead of a simple diagonal tensor. A method is presented to calculate the permeability tensor for an unstructured gridblock to account for the small-scale heterogeneity information inside the gridblock. Single-phase flow-based upscaling is performed based on random boundary conditions and the effective tensor permeability is found through an optimization technique. Full, symmetric and diagonal permeability tensors are calculated for 2D and 3D blocks and sensitivity analysis is performed.

Summary This paper describes the modeling of the main processes for CO 2 trapping in saline aquifers, namely solubility trapping, residual gas trapping and mineral trapping. Several important aspects of CO 2 storage are presented. It has been found that the total amount of CO 2 trapped as a soluble component, and as residual gas, can be enhanced by injecting brine above the CO 2 injector. An optimization technique is used to adjust the location and rate of the brine injector to maximize the total amount of CO 2 trapping. The security of the trapping process is then evaluated by taking into account the leakage of mobile CO 2 through the caprock. For long-term CO 2 storage, the conversion of CO 2 into minerals is found to depend on the pre-existing minerals in the aquifer that provide the necessary ions for the reactions to occur.

Summary Accurate characterization of sub-surface oil reservoirs is an essential prerequisite to the design and implementation of enhanced oil recovery (EOR) scenarios. Specifically, in reservoir characterization, integrating static and dynamic data into reservoir models to construct accurate and realistic models has received considerable attention. Unlike most of the conventional geostatistical approaches of integrating data into reservoir models that are based on semi-variograms (two-point statistics) as a measure of spatial connectivity, a complete multiple-point (MP) statistic framework is presented in this paper. In contrast to two-point statistic methods, MP statistics-based methods are capable of reproducing curvilinear geological structures. The algorithm starts with extracting MP statistics from training images (TI) using an optimal spatial template. After collecting different patterns and building the MP histogram, the pattern reproduction process commences. This process begins from data locations and then grows to fill the whole reservoir domain. The algorithm accounts for three main practical issues: uncertainty in geological scenarios, scanning template and non-stationarity. The MP statistics algorithm (growthsim) is capable of integrating data from multiple data sources. Among these data types is dynamic data or flow history. The conventional approach to integrate production information into reservoir models is by iterative perturbation of the reservoir model until the production history of the reservoir is matched. Iterative methods have been applied till date to random fields that are completely characterized by a two-point co-variance function. In contrast, this paper presents a forward modelling approach that investigates history matching within a MP modelling framework. A novel technique implemented in this research is based on the merging of MPs inferred from history matched and geological models. Pattern growth is performed subsequently by sampling from the merged MP histograms. History matched models using the presented approach show an excellent agreement with underlying geological descriptions and match production history.

Summary Near-shore oil reservoirs have become significantly depleted, forcing oil companies to explore deep-sea reservoirs with huge investments and the latest technology. However, these projects are often risky. Thus, the optimal solution is to explore shallow sea oil fields before proceeding to deep, high-risk areas. The Lower Miocene reservoir of the White Tiger field is a sedimentary reservoir with high heterogeneity and complex geological characteristics. This reservoir was discovered 22 years ago. There is an urgent need to study procedures for an increased and maximum oil recovery. A detailed geological understanding of the reservoir, along with a reservoir simulation, is needed to gain a detailed reservoir description and determine the optimal recovery method for this oil reservoir. These are essential to having a successful operation, as well as reducing uncertainties and improving the efficiency of oilfield management. With a large database collected from initial production stages of over 50 wells, the authors developed an integrated static and dynamic workflow to forecast oil production under several production scenarios for this reservoir. These integrated results served as input data for simulation with IMEX TM , which will be useful for the economical and technical evaluation of this study. In addition, the authors introduced history matching and pointed out the main reasons for the significant errors between actual data and simulation results. Based on the reservoir modelling, the authors optimized the wells? network locations and obtained good results in oil recovery. The oil recovery factor increased from 24.21% - 37.26% for the Lower Miocene reservoir. Specifically, oil recovery for the Southern dome structure was increased from 14.83% - 33.76%.

In China, annual natural gas consumption is over 67 billion cubic meters with an expected growth rate of 10% per year. Most of the gas is transported from well heads to markets over cross-country gas networks, which requires construction of the West to East Gas Network--one of the largest gas networks in the world. Presently, the network is comprised of four large-diameter pipelines and will include most major gas pipelines in China in the future. The network distributes approximately 30 x 109 m 3 gas per year, of which 3% to 5% is burned to power the gas transportation. At current gas prices, gas transportation costs are roughly 350 million per year, which is a considerable cost that could be reduced by improvements in network design and operation. This paper reports on a study aimed at optimizing the network to minimize its energy consumption and cost. The large size and complex geometry of the network required breaking the study down into simple components, optimizing operation of the components locally, re-combining the optimized components into the network and optimizing the network globally. This four-step approach employed four different optimization methods, penalty function method, pattern search, enumeration and non-sequential dynamic programming, to solve the problem. The results show that cost savings, because of global optimization, are reduced with increased throughput. For example, increasing the gas rate from 67 - 90 million m 3 /d would reduce operational cost savings because of optimization from 23% - 1.15%. Moreover, the study shows that if the compressors were fully loaded at their maximum rating, the optimized operation would approach the one being presently practiced. Thus, the optimization is effective and much needed when the system does not work at its maximum capacity, a typical case in the present operations of Chinese gas networks.

Failure of drill-strings is very costly in terms of money and time and occurs more frequently than oil companies would like to see. There are many reasons for drill-string failure, such as vibration, fatigue, and buckling. This paper addresses the problem of suppression of stick-slip oscillations in oil well drill-strings using proportional integral derivative (PID) and lead-lag controllers in conjunction with genetic algorithms. Simulation results are presented to validate the proposed control schemes. Introduction The drilling process is affected by the dynamically induced vibrations caused by design imperfections as well as material elasticity. Because of these vibrations, premature wear and damage of drilling equipment may occur. Vibrations can decrease the rate of penetration (ROP), and thereby increase the cost of the well(1,2,3). Moreover, vibrations can interfere with measurement-while-drilling (MWD) tools or even cause their damage. Another major problem caused by vibrations is the induced wellbore instabilities which can worsen the condition of the well(4,5). The literature on drill-strings can be classified in three groups: drill-string modelling, drill-string control, and drill-string technology. In this paper, we will focus on drill-string dynamics and control. There have been a significant number of research publications on drill-string dynamics and a representative few are considered here. Many different models for drill-strings are available in the literature, which may include one or more of the following phenomena: bit-bounce, stick-slip, forward and backward whirl, axial, lateral, and torsional vibrations; see, for example, Christoforou et al.(6), Jansen and van den Steen(7,8), and Leine et al.(9). Stick-slip is a major cause of torsional vibrations and many researchers have tried to minimize its effect on the behaviour of the drill-strings using active damping techniques(10,11,12). These techniques reduce the torque fluctuations and torsional drill-string vibrations affecting in this manner the stick-slip conditions. The underlying concept is to reduce the amplitude of the downhole rotational vibrations using torque feedback. The feedback is used by the rotational drive, which slows down the rotary rate when the torque increases and speeds it up when the torque decreases(8,13,14).

As SAGD is being increasingly used as a commercial technology to recover heavy oil and bitumen, it is essential to determine the most economical operating conditions for a SAGD operation by reservoir simulation. Furthermore, to support the decision-making process of a SAGD project, it is also important to quantitatively assess the uncertainty of its economicforecasts. In this paper, the application of global optimization, experimental design, response surface generation and Monte Carlo simulation techniques in the workflow of SAGD simulation studies were demonstrated with a real field case example. The field case is an infill SAGD project with two planned SAGD well pairs and eight existing primary production wells which have 5 years of primary production. A bottomwater zone is also present. Three major steps of the workflow are: history matching primary production data; optimizing SAGD performance; and quantifying uncertainty of the SAGD forecasts. Firstly, experimental design and DECE (Designed Exploration and Controlled Evolution) optimization methods were used to achieve a faster and better history match than the traditional manual history match. Secondly, SAGD performance was optimized by adjusting the steam injection rate and producer liquid withdrawal rate during different SAGD operation periods. Finally, experimental design and response surface generation techniques were applied to build a polynomial response surface through which the net present value (NPV) of the SAGD project is correlated with uncertain parameters and a SAGD design parameter. Monte Carlo simulation was then performed to quantify the uncertainty of SAGD forecasts in terms of cumulative probability distribution of the NPV at different values of the SAGD designparameter. The results show that the economics of this project are improved considerably through optimization. The optimum operating conditions obtained use a high initial steam rate and high production rate to develop the steam chamber. After the instantaneous steam-oil ratio reaches a certain value, both steam rate and production rate are lowered to prevent steam breakthrough to the bottomwater. The uncertainty of the project NPV was assessed, taking into consideration the uncertainties in high temperature relative permeability endpoints and the variation of the SAGD design parameter. Introduction Steam-assisted gravity drainage (SAGD) is a thermal oil recovery process which consists of pairs of two parallel horizontal wells drilled near the bottom of the pay (1) . Typically, the length of the wells are between 500 and 1,000 m, the inter-well distance of the two parallel wells is between 5 and 10 m and inter-well pair spacing is between 90 and 120 m (2, 3) . The top horizontal well is used to inject steam, while the bottom horizontal well is used to produce reservoir fluids. The steam injected from the top well rises into the formation, forming an expanding steam chamber around and above the injection well. The rising steam eventually loses its latent heat near the boundary of the steam chamber, heats the oil and allows it to drain to the bottom production well by gravity. Successful field tests have proven that SAGD is a viable technology for in situ recovery of heavy oil and bitumen (4–6) .