This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Dyn am ic Oil an d Gas Produ ction Optimization via Exp licit Rese rvoir Simulation 181
2. Previous Wo rk and Current Chal lenges
A number of scientific publications address modeling and simulation of oil extraction:
they either focus on accurate reservoir simulation, without optimization considerations
(Hepgu ler et al. , 1997; Litvak et al., 1997), or on optimal well planning and operations,
with reduced (Lo, 1992; Fang and Lo, 1996; Stewart et al. , 2001; Wang et al. , 2002) or
absent (Van den Heever and Grossmann, 2000; Saputelli et al. , 2002) reservoir models.
Computational Fluid Dynamics (CFD) is a powerful technology, capable of elucidating
the dyn amic beh avior o f oil reservoirs toward s efficient oilfield operation (Aziz, 1979).
The M INLP formulation for oilfield production optimization o f Kosmidis (2004) uses
detailed well models and serves as a starting point in the case examined in this study.
Therein, the nonlinear reservoir behavior, the multiphase flow in pipelines, and surface
capacity constraints are all considered (multiphase flow is handled by DAE systems,
which in turn comprise ODEs for flow equations and algebraics for phys. properties).
The mo del uses a degrees-of-freedom analysis and well bounding, but mo st importantly
approxim ates each well mo del with piecewise linear functions (via data preprocessing).
Here, explicit reservoir flow simulation via a dynam ic reservoir simulator (EC LIPSE ®)
is combined with an equation-oriented process optimizer (gPROMS®), towards
integrated m odeling and optimization of a l iterature problem (Kosmidis, 2 0 0 5 - Ex. 2a).
An asynchronous fashion is employed: the first step is the calculation of state variable
profiles from a detailed description o f the prod uction system (reservoir) via E CLIP SE ®.
This is possible by rigorously simulating the multiphase flow within the reservoir, with
real-world physical properties (w hose extraction is laborious: E conom ides et al. , 1994).
These dynam ic state variable profiles (pressure, oil, gas and water saturation, flows) are
a lot more accurate than piecewise linear approximations (Kosmidis, 2003), serving as
initial conditions for the higher-level dynamic optimization model (within gPROMS®).
Crucially, these profiles constitute major sources of uncertainty in simplified models.
Considering the oil and gas pressure drop evolution within the reservoir and along the
wells, one can solve single-period or multi-period dynamic optimization problems that
yield superior optima, because piecewise linear pressure underestimation is avoided.
While integrating different levels (sub-surface elements and surface facilities- Fig. 1)
is vital, interfacing CFD simulation with MINLP optimization is here pursued in an
asynchronous fashion (given the computational burden for CFD nested within MINLP).
The concept o f integrated mo deling an d optim ization is illustrated in detail in Figure 2:
...........L / 2 IIII ZL ILI iiii ii I i~ii;...........................................................................................................................
'.,,.. 1. Ca.|culat~on ,of res erv oir stat e vat. pro f i les ( 3D C IFD t s
...........2.. E×t ract io n of acc urate 1D IOs (use i,n 9~R OMS ), . . - '
T O P - L E V E L D Y N . O P T : I M I S A T I O N
( S U R F A C E & ~ ,~ 'E LL S Y S T E M )
B O T T O k T - L E V E L S I M U L A T | O N
( V v EL L & R E S E R V O I R S Y S T E M )
Figure 2 Integrated mode ling and optimization of oil and gas production systems
illustration of the explicit consideration o f multiphase flow within reservoirs and w ells.
Table 1: Oil product ion opt imizat ion using reservoir s imulat ion boundary condi t ions.
2005) work
Oil production (STB /day) 35000 29317.2 30193.7 (+2.9% )
Gas production (MSCF /day) 60000 60000 60000
W ater production (STB /day) ....................................................................................................................................................................................................
5 . Conclus ions and Future Goals
The comb ina t ion of dynamic mul t iphase CFD s imula tion and MINLP opt imiza tion has
the potent ial to yield improved solut ions towards eff icient ly maximizing oi l product ion.
The present paper addresses integrated oi l f ie ld model ing and opt imizat ion, t reat ing the
oi l reservoirs , wel ls and surface faci l i t ies as a combined system: most important ly, i t
s t resses the benefi t o f comp uting accurate s tate var iable prof i les for reservoirs via CFD .
Expl ici t CF D simulat ions via a dynam ic reservoir s imulator (EC LIPS E ®, Schlum berger)
are comb ined with equa t ion-or iented pro cess o pt imizat ion sof tware (g PR OM S ®, PSE)
the key idea is to use reduced-order copies of CFD profi les for dynamic opt imizat ion.
The l i terature prob lem solved show s that explici t use of CFD resul ts in opt imizat ion
yields improved opt ima at addi t ional cost (CPU cost a n d cost for eff icient separat ion of
the addi t ional water ; the percentage difference is due to accurate reservoir s imulat ion) .
These must be evaluated systematical ly for larger case s tudies under var ious condi t ions.
A c k n o w l e d g e m e n t s
The authors acknowledge f inancial support as wel l as a postdoctoral fel lowship from the
European Union (FP6) under the auspices o f a Mar ie Cur ie Research Tra in ing Network:
Towa r ds Knowl e dge - Ba s e d P r oce s si ng Sys t e m s / PRI SM ( M RTN- CT- 2004- 512233) .