Summary
This paper proposes a method based on Lagrangian decomposition for real-time
production optimization. It argues the method is well suited for the
well-allocation and routing problem in the upstream industries. A semirealistic
field case assuming steady-state conditions is used to compare the merits of
the method with current industrial practice. The method compares favorably and
is therefore an interesting option in real-time production optimization. There
are several reasons for this. An error bound on the solution of the production
optimization problem can be computed, which is clearly information of interest
to any user. Furthermore, the algorithm is efficient and can be parallized for
an even higher efficiency. Finally, the Lagrangian decomposition method
provides a tool for understanding how a limited resource, such as gas capacity,
is used in production optimization.
Introduction
Development of a field asset requires planning on multiple horizons. On a
long-term horizon, strategic decisions are made on field development (e.g.,
choice of technology and export options, investment strategies, and recovery
strategies despite uncertainties) (Saputelli et al. 2007). For offshore assets,
the choice of technology may include subsea solutions and the issue of
processing the reservoir fluid either offshore or onshore. Export choices
include pipelines or liquified natural gas tankers for gas, and pipelines or
oil tankers for heavier components. Furthermore, option management may point to
a flexible, and thereby more costly development to allow for future
development, such as tie-ins from possible neighboring assets. The analyses and
subsequent development plan seek to maximize the net present value of the asset
or maximizing oil recovery.
On a medium time horizon, typically 3 months to 2 years, production rates
and, if applicable, injection rates are decided. Depending on the life cycle of
an asset, decisions may also involve a drilling program. During the green field
stage, it is important to plan, drill, and commission new wells to reach some
predefined plateau rate as soon as possible. During plateau production, there
may be an in-field drilling program for production and/or injection wells. This
program involves decisions about the location and completion of wells. During
the decline phase of a field, lift technology may be an issue involving
decisions (e.g., the distribution of limited amounts of lift gas between
wells). A reservoir simulator is usually an important planning tool on the
medium time horizon. A reservoir simulator is quite complex if the geology is
complex, because of heterogeneities, like faults and shale layers, to represent
flow patterns accurately. Furthermore, complexity increases if phase behavior
is complicated. Gas-condensate reservoirs often necessitate the use of a
compositional simulator instead of a black-oil simulator to get accurate
prediction of phase behavior, which increases computational load significantly
because of thermodynamics.
On a short-time horizon, typically days to weeks, production optimization in
which both the subsurface part (i.e., reservoir and wells) and the surface part
(i.e., collection and downstream production equipment) of the system are taken
into account is important. We denote this optimization to be the real-time
production optimization (RTPO) problem and return to a more precise definition
later. Production may be constrained by reservoir conditions, such as coning
effects and/or the production equipment like pipeline capacity or downstream
water-handling capacity. Constraints may move from one part of the system to
another part over time. Water production may be low early and increase
dramatically during the decline phase of a reservoir, thereby making
water-handling capacity an issue. Decision variables in RTPO include production
and possibly injection rates, artificial lift inputs like lift-gas rates and
electric-submersible pump (ESP) rates, and routing of well streams.
There are also faster time scales, including supervisory and regulatory
control, such as those presented in several references [e.g., the operational
hierarchy in Fig. 1 in Saputelli et al. (2007)].
This paper focuses on the RTPO problem. The contribution is a method
supported by optimization theory that enables us to decompose an RTPO problem
into parts (i.e., applying a type of divide-and-conquer strategy). This
approach has some intriguing properties highlighted through a simplified field
case. To elaborate, the conquer-and-divide strategy is a commonly used
engineering design principle that has survived since the complex system came
into making. The earlier referenced operational hierarchy in Saputelli et al.
(2007) is one example of the conquer-and-divide strategy.
© 2009. Society of Petroleum Engineers
View full textPDF
(
393 KB
)
History
- Original manuscript received:
25 April 2008
- Revised manuscript received:
6 November 2008
- Manuscript approved:
6 November 2008
- Published online:
23 July 2009
- Version of record:
22 December 2009
View Cart
