An Efficient Formulation For Simultaneous Solution Technical Paper
An Efficient Formulation For Simultaneous Solution Technical Paper
Network Equations
Document Type: Conference Paper
Document ID: SPE-93073-MS
Publication Date: 2005
Authors
G.S. Shiralkar (BP) | J.W. Watts (Landmark Graphics)
Link:
An Efficient Formulation for Simultaneous Solution of the Surface Network Equations
Abstract
We present a novel formulation for simultaneous solution of the surface network equations, which are
cast into a form similar to the reservoir flow equations. This form is more palatable to the efficient solvers
already developed for reservoir simulators.
Introduction
Integrated modeling of the reservoir with a surface network has assumed high business importance.
Because of the smaller timescale, surface flow is generally modeled assuming steady-state flow, leading
naturally to unstructured "network oriented" surface formulations.
Coupled models may require solution of the network equations, by themselves as well as together with
1
the reservoir equations in a global matrix, e.g., Coats et al . Typical network solution variables include a
mix of nodal quantities (pressures) and connection quantities (flow rates). For networks of complex
architecture with multiple outlet connections at nodes or looped flow paths, it is not generally possible to
order these variables in a regular way that reflects network topology. Such formulations may require the
use of direct (non-iterative) or inefficient iterative matrix solvers for simultaneous solution of network
variables.
When the network system is computationally "large", which can be caused by a large number of
wellbores, a complex surface network, a large number of hydrocarbon components, or any combination
thereof, the inefficient solution of the network equations can become a significant factor if not a bottleneck
within an integrated simulator. Additionally, non-iterative or highly unstructured solvers present problems
for parallelization.
The proposed formulation results from mapping connection based variables into an equivalent set of
nodal variables. This results in equations of a more familiar structure to allow the leveraging of existing
linear solvers that are efficient and parallelizable. The formulation is presented for multi-component,
multiphase flow, and does not depend on assumptions of network convergence or particular network
architecture.