Summary
Gas/condensate reservoirs have been the subject of intensive research
throughout the years because they represent an important class of the world’s
hydrocarbon reserves. Their exploitation for maximum hydrocarbon recovery
involves additional complexities that cast them as a different class of
reservoirs, apart from dry-gas, wet-gas, and oil reservoirs. Gas/condensate
reservoirs are good candidates for compositional-simulation studies because
their depletion performance is highly influenced by changes in fluid
composition. Often, highly sophisticated and computationally intensive
compositional simulations are needed for the accurate modeling of their
performance, phase behavior, and fluid-flow characteristics. The desired
outcome of a simulation study for gas/condensate reservoirs is the
identification and development of the best operational production schemes that
maximize hydrocarbon recovery with a minimum loss of condensate at reservoir
conditions. However, compositional simulations are demanding in terms of
computational overhead, manpower, and software and hardware requirements.
Artificial-neural-network (ANN) technology (soft-computing) has proved
instrumental in establishing expert systems capable of learning the existing
vaguely understood relationships between the input parameters and output
responses of highly sophisticated hard-computing protocols such as
compositional simulation of gas/condensate reservoirs. In this study, we
conduct parametric studies that identify the most influential reservoir and
fluid characteristics in the establishment of optimum production protocols for
the exploitation of gas/condensate reservoirs. During the training phase of the
ANN, an internal mapping is created that accurately estimates the corresponding
output for a range of input parameters. In this paper, a powerful screening and
optimization tool for the production of gas/condensate reservoirs is presented.
This tool is capable of screening the eligibility of different gas/condensate
reservoirs for exploitation as well as assisting in designing the optimized
exploitation scheme for a particular reservoir under consideration for
development.
Introduction
In a gas/condensate reservoir, initial reservoir conditions are located
between the critical point and cricondentherm of the reservoir fluid, as shown
in Fig. 1. In general, in a gas/condensate reservoir, the initial fluid is
found in an all-vapor condition. A liquid phase is later developed once the
dewpoint of the system has been reached as a result of isothermal depletion.
Appearance of a liquid phase upon vapor expansion (under isothermal conditions)
is not possible for pure substances; thus, this behavior is categorized as
“retrograde” for this particular type of mixture. The retrograde liquid may
even revaporize if depletion continues. The major concerns while producing a
gas/condensate reservoir have to do with the loss of this valuable liquid to
the reservoir and the associated impairment in gas productivity. Accordingly,
the study of a gas/condensate reservoir usually entails the use of the most
sophisticated phase-behavior and fluid-flow tools at the disposal of the
reservoir engineer.
The type of application of neural-network technology used in this work is
described as neurosimulation. In neurosimulation, hard-computing techniques
(such as numerical computations) are coupled with soft-computing techniques
(such as ANNs) for the development of a powerful expert system. Numerical
models provide a precise and formal “expertise”—at a significant computational
expense—that can be taught to a soft-computing tool that, once trained, can
exploit and apply the learned expertise through much-less-intense computational
work.
© 2007. Society of Petroleum Engineers
View full textPDF
(
2,805 KB
)
History
- Original manuscript received:
15 December 2004
- Meeting paper published:
18 October 2004
- Revised manuscript received:
28 November 2006
- Manuscript approved:
28 November 2006
- Version of record:
20 April 2007
View Cart
