Summary Reservoir simulation studies allow engineers to identify development strategies that can maximize the economic recovery of hydrocarbon resources. During the life of a field, the reservoir engineer confronts numerous decisions about the placement and operation of wells for production or injection. Traditionally, engineers evaluate the costs and benefits associated with competing development strategies by using reservoir simulation tools in different scenarios. The numerical reservoir simulator is used in forecasting recovery of reserves under existing production schemes and in evaluating the effects of changing operating conditions. This approach limits the overall benefit of a field study because it does not consider the effects that flows in wellbores and surface facilities have on the economic performance of the recovery programs. Yet, the costs of compression, separation, fluid injection, or water treatment can have a significant effect on the success of production strategies. When the goal is to identify an optimal hydrocarbon-recovery scheme, a compelling case arises for the computational integration of subsurface- and surface-simulation technologies.
In this paper, we discuss an integrated computational solution for management of reservoir-production strategies and field development. The approach is implemented by coupling a reservoir simulator (Eclipse) with a surface and production network simulator and optimizer (Netopt). Parallel Virtual Machine (PVM) interface provides the coupling communication mechanism that integrates the two simultaneously running simulators. At each timestep, the network simulator and the reservoir simulator provide a consistent integrated solution after rate and pressure convergence is achieved within a predetermined tolerance.
Introduction Reservoir simulators are designed to provide information on the behavior of the formation subdivided into a two- or three-dimensional (3D) network of gridblocks. The simulator calculates spatial distribution of fluid pressure and saturation, producing GOR, water cut, and injection or production rate for each well at each timestep of the computations. The reservoir model also determines the local inflow performance for each well at its gridblock as a function of flow rate and cellblock pressure within the timestep. The wells must be assigned a production- or injection-rate schedule and/or a flowing-bottomhole-pressure limit so that well productivity (injectivity) or deliverability can be calculated. What the reservoir can deliver, however, is controlled directly by the entire production network, which includes single- or multiphase flow through wellbores and surface facilities.
A simulation model that combines the reservoir with production-piping network offers increased accuracy in predicting reservoir deliverability and in forecasting full-field performance on the basis of many possible production-facility schemes. Startzman et al. reported one of the earliest implementations of this concept and described an interfaced program where a facility simulator calculated the capacity of producing network and passed the individual well flow rates to a reservoir simulator for use in solving the next time-step. Emanuel and Ranney presented a formulation that consisted of three separate systems that solved the reservoir model, well flow, and surface-network equations. The surface-network model, however, was limited to tree-structured networks with a fixed sink (separator) pressure, and wellbore-pressure drop was determined from pregenerated flow tables to reduce CPU time. To increase calculation efficiency, Schiozer and Aziz described advantages of using domain decomposition in the reservoir model. Breaux et al. and Stoisits et al. showed applications and impact of integrated simulations in field-development studies.
In this paper, we present an integrated solution that couples a 3D three-phase reservoir simulator with a general-purpose network simulator that does not impose any restrictions on the network structure or on network boundary conditions. The integrated model is divided into two domains:the reservoir equations, modeling the flow through porous media, and
the network equations, modeling flow in wells, surface piping, and equipment.
The convergence algorithm is different from past algorithms in that it includes the local inflow performance relationships (IPR's) for each well both in the reservoir and in the network equations, and both sets of equations converge to a consistent solution within every timestep. This approach results in a closely coupled interface and a robust convergence procedure with smaller number of iterations at each timestep. Well-control logic implemented in the network simulator and at the interface allows the wells to be on rate control, on deliverability, or shut in.
The surface and production network simulator incorporates all the components in the production stream starting at the reservoir formation sandface and ending at a sales point or a refinery inlet. The steady-state network solution includes well-completion models: IPR's; multiphase flow through wellbores and pipeline gathering or distribution systems; wellbore and pipeline heat-transfer details; critical or subcritical flow through chokes; and other equipment in surface facilities, such as compressors, pumps, heat exchangers, and regulators. The network simulator calculates flow rates, pressure and temperature profiles, and fluid properties on the basis of pressure or flow boundary conditions specified at source or sink nodes while accounting for reservoir behavior over time.
In the integrated solution, the surface and production network simulator acts as the master program sending and receiving messages through a PVM interface. For each well in the simulation model, flow rates, pressures, and local IPR data within a timestep are communicated between the reservoir and network simulators through a PVM daemon, which eliminates any need for file exchange. The rate- and pressure-dependent local IPR data provided by the reservoir simulator set the boundary conditions for the network simulator that determines the production-system and surface-facility performance. The convergence algorithm that ensures a consistent solution is incorporated into the network simulator. At every timestep, before the final rate and pressure solution is updated for the following timestep, each simulator must run with information updated by its companion simulator. This must be done a number of times successively with each simulator reaching its own converged solution. The overall convergence is achieved when rates and pressures agree within a predetermined tolerance.
In this paper, we present the integrated solution procedure after we examine each simulator and the PVM interface. A field example is also presented to show the application of the integrated system.