Suplacu de Barcau, a heavy oil field in Romania with more than 2,700 wells and over 50 years of air injection history, is considered one of the world's largest in-situ combustion projects. This paper describes a 3D simulation study of a sector of the field. The sector chosen spans the entire history of the field, from a short period of cold production in the early 1960s to the current production method of cyclic steam stimulation followed by air injection. To date, about 200 wells have been drilled in this sector, which covers approximately 1.1 km2. We present the different stages of the numerical study and review some of the difficulties involved in modeling in-situ combustion in a large model. We describe how we managed the uncertainties in the reservoir and in the fluid description, and how we overcame limitations in the available data. Kinematic data from laboratory experiments were used as a starting point, and we present details of the simplified kinematic formulation that was needed to improve the numerical performance of the simulation model. The model is heterogeneous, so there is an uneven propagation of the combustion front. We compare historical production rates with the model predictions and estimate areas of oil that were bypassed due to this heterogeneity and to gravity override.

No abstract

An investigation is presented on the use of Flow Control Valves (ICVs, FCVs) to control steam placement in the early stages of a Steam Assisted Gravity Drainage (SAGD) process. The two parts of this process that are examined in this paper are the steam circulation preheating period and the early stages up to one year of injection/production in which the steam chamber is beginning to form. Steam injection and production in this and other thermal processes can be difficult to control because steam has a high mobility ratio and tends to establish flow paths that may be difficult to break once established. This is especially pronounced in heterogeneous reservoirs. Two SAGD case studies have been designed that accurately model the initial preheating period in which both wells circulate steam through an inner tubing and outer annulus in order to conductively and, to a lesser extent convectively, heat the region around the well pair in order to establish communication. After this initial circulation period, the wells switch to injection and production. Both cases have the same base configuration but differ in the degree of reservoir heterogeneity. In the injection well, ICV devices are placed to control steam/water flow through the outer screens. In the producer, FCV valves are used to flatten the production profile along the well. Two methods are examined to change valve apertures. One uses proportional-integral-derivative (PID) controllers while the second applies an optimization algorithm directly on each individual connection productivity index. A preliminary investigation is presented here into using feedback controllers and optimization with instantaneous reservoir parameters to improve a SAGD process in the presence of reservoir heterogeneity.

This paper describes a 3D thermal reservoir simulator that models the injection of steam into heavy oil. The variables are pressure, energy and molar densities, and the equations are volume balance and component and energy conservation. The use of energy and molar densities means that there is no variable-switching, and the formulation combines the advantages of volume balance with a fully-implicit code. A number of example problems are presented, including a large full-field fine-grid model run on parallel processors. Introduction We describe a full-featured thermal simulator that works in a number of coordinate systems: cartesian or radial, rectangular or corner-point geometry, and PEBI grids. It calculates non-neighbor connections, and will solve both single- and dual-porosity models. Local grid refinements can be defined, and these are treated simultaneously and fully coupled in the linear solver. A nine-point method can be used to reduce grid orientation effects. Computer memory requirements are minimized by only storing properties and derivatives for active cells. The new simulator is based on an existing simulator that runs in either compositional or black oil mode. A choice has been maintained and we can either model dead or live oil. We define live oil as having at least one volatile component and dead oil as having none. This paper will describe the general formulation then give examples for the dead oil case. In practice, the dead oil model can be applied to oils that have a small proportion of volatile components as long as these components do not have a significant effect on the simulation. To solve the thermal effects in the reservoir we choose energy rather than temperature as our new variable. The main advantage of an energy variable is that we avoid the problems of variable switching. In the dead-oil case, for instance, we can use energy to describe the boiling of water at a constant pressure and temperature. By adding an energy variable and an energy conservation equation to the existing formulation, we can simulate the injection of steam or hot or cold water (or hot or cold gas) into the reservoir. The formulation differs from many others in calculating energy bounds for each state in each grid block. The energy in the grid block then defines its state and a flash is performed within that state. We use a conventional fully-implicit discretization of the flow equations. Once the non-linear equations have been set up, they are iterated using Newton's method. The linear equations are solved using nested factorization. The simulator can be run either in scalar or parallel mode, and results are presented for both cases. Formulation Variables and Equations The primary solution variables X are the pressure P, the reservoir molar densities mc and the bulk internal energy density e. The system is made up of N-1 oleic components and one aqueous component. There are N+2 variables. The N+2 equations are one volume balance equation, one energy conservation equation, and N component conservation equations. P. 97^

Ponting, D.K., Ponting, D.K., Atomic Energy Research Establishment Foster, B.A., Atomic Energy Research Establishment Naccache, P.F., Atomic Energy Research Establishment Nicholas, M.O., Atomic Energy Research Establishment Pollard, R.K., Pollard, R.K., Atomic Energy Research Establishment Rae, J., SPE, Atomic Energy Research Establishment Banks, D., British Natl. Oil Corp. Walsh, S.K., Atomic Energy Establishment Abstract This paper describes Program for Oil Reservoir Simulation (PORES), an efficient general-purpose black-oil simulator now in production use for modeling North Sea fields. The frilly implicit finite-difference equations are solved for each time step with a Newton-Raphson procedure. The resulting large sets of linear equations are procedure. The resulting large sets of linear equations are usually solved simultaneously by a new and powerful iterative method that uses a preconditioned conjugate gradient algorithm with an enforced column-sum condition to accelerate convergence. A sequential solution option is available, and direct matrix inversion methods also are provided. Gas condensate problems are handled by a variable switching technique. Four examples are presented to illustrate the power and efficiency of the presented to illustrate the power and efficiency of the program. program. Introduction PORES is a general-purpose three-phase fully implicit PORES is a general-purpose three-phase fully implicit finite-difference oil-reservoir simulator. It was developed to exploit the stability of the fully implicit equations to provide a package capable of efficiently solving widely differing types of problems without the danger of numerical instability. The program has been in production use for the U.K. Dept. of Energy (DEn), the British Natl. Oil Corp., and the British Gas Corp. for the past 2 years. This paper describes the main features of the PORES program. The finite-difference equations describing the flow of oil, water, and gas are formulated in terms of residuals associated with each phase for every cell of the reservoir. The resulting fully implicit equations are linearized by a Newton-Raphson iteration scheme. Convergence of the nonlinear iterations is monitored by substituting successive iterates into the finite-difference equations and evaluating the root mean square (RMS) residual. The linear equations that arise at each nonlinear iteration can be solved by iterative or direct methods. The iterative method is a preconditioned conjugate gradient algorithm with an enforced column-sum condition to ensure material conservation and to accelerate the iteration process. The form of preconditioning used is particularly process. The form of preconditioning used is particularly suited to reservoir simulation problems that contain an inherent directionality. The direct solver is based on Gaussian elimination with D4 ordering. Both types of solution include the effects of off-band terms arising from the inclusion of multilayer wells. The effect of flow between non-neighboring pairs of reservoir gridblocks can also be included. The linear equations may be solved fully simultaneously or a number of phases may be treated sequentially. The sequential method performs an approximate factorization of the Jacobian matrix while employing a column-sum condition to maintain material balance. Although the sequential option is efficient on easy problems, we have found that many North Sea applications require the extra stringency of a simultaneous solution. Computer core requirements can be reduced by allocating storage for only the active cells in the reservoir. For large problems this can reduce memory requirements by 50% and in exceptional cases by as much as 75%. The program contains a number of additional features to improve user convenience.A micro- or minicomputer-based color graphics facility enables interrogation of an interface file produced from a simulation run. produced from a simulation run.One option generates directional pseudo functions on a prescribed coarse mesh from the results of a fine-grid simulation.An aquifer model based on fault connections between large aquifer blocks and any number of reservoir gridblocks is available, in addition to analytical models. JPT P. 544