Summary
We developed an operational strategy for commingled production with
infinitely variable inflow control valves (ICVs) using sequential linear
programming (SLP). The optimization algorithm requires instantaneous and
derivative information. We propose a workflow in which the production engineer
relies on measurements to determine the flow rate and pressure values and on
models to determine the derivative information (i.e., the changes in flow rates
as a result of a change in an ICV setting). Such a model typically would be a
steady-state wellbore simulator including choke models to represent the ICVs
and inflow models to represent the near-well reservoir flow in the various
zones. The parameters of the model need to be updated regularly using real-time
measurements and production tests, and we discuss the impact of different
smart-well instrumentation levels on the updating process.
We simulated the performance of this production-optimization strategy in a
reservoir simulator. Some numerical aspects of the algorithm and problems
encountered during implementation are discussed. The performance of the
algorithm was tested in two reservoir settings. In both cases, the optimization
resulted in accelerated oil production compared to conventional,
surface-controlled production. However, accelerated production did not always
result in higher ultimate recovery compared to the conventional case. In such
situations, the benefits of either short-term production optimization
(accelerating production) or long-term reservoir management (maximizing
recovery) should be weighed.
Introduction
Smart Wells. The introduction of smart completions in the oil industry has
significantly increased the scope for control of commingled production. ICVs
allow for the adjustment of inflow in each individual zone; see Fig. 1.
Efficient use of ICVs requires the capability to measure the inflow from each
zone. Using downhole instrumentation, this can be done directly, with downhole
flowmeters, or more indirectly, through “soft sensing” (i.e., through
interpretation of pressure and temperature data from surface and downhole
sensors in combination with models for pressure and temperature drop over the
wellbore and the valves). All these measurements require occasional calibration
based on surface production tests, where ideally the flow rates of each
individual layer should be tested.
In addition to measurement and control hardware, smart-well operations
require a control strategy. Present operation of smart wells is based mostly on
a “reactive” control strategy, in which valves are closed in reaction to the
breakthrough of water or gas. The present paper proposes a more “proactive”
strategy to continuously optimize the oil production of a well, using measured
data while honoring constraints on water and gas production.
Optimization Methods. Optimization with the objective to improve the
economics of oil or gas production can, in general, be considered on two
different time scales: (1) reservoir management, which involves the
long-term saturation response of the reservoir (e.g., optimization of sweep
efficiency in waterflooding), and (2) production optimization, which
involves the pressure and short-term saturation responses (such as water
breakthrough) (Rossi et al. 2000). Short-term production optimization can be
performed with simulation models for wellbore flow and near-wellbore reservoir
response. The objective is to maximize production at a specific moment in time,
which leads to the use of optimization techniques such as SLP
(Handley-Schachler et al. 2000) or sequential quadratic programming (SQP) (Wang
et al. 2002; Davidson and Beckner 2003). In reservoir management, however, the
objective is to maximize recovery or present value (PV) over a long time
period. This requires the use of a reservoir simulator in combination with a
gradient-based optimization technique (Brouwer and Jansen 2002; Asheim 1988;
Sudaryanto and Yortsos 2000; Yeten et al. 2002, 2004) or a nonclassical
optimization technique such as a genetic algorithm (Palke and Horne 1997; Yang
et al. 2003) to optimize multiple control variables at multiple points in time.
The present study is restricted to short-term production optimization.
© 2006. Society of Petroleum Engineers
View full textPDF
(
3,469 KB
)
History
- Original manuscript received:
20 January 2005
- Revised manuscript received:
6 September 2005
- Manuscript approved:
9 September 2005
- Version of record:
20 May 2006
View Cart
