Optimal Field Development and
Control Yields Accelerated, More
Reliable, Production: A Case Study
Morteza Haghighat, Khafiz Muradov, David Davies
Institute of Petroleum Engineering, Heriot-Watt University, Edinburgh, UK
SPE Aberdeen 3rd Inwell Flow Surveillance and Control Seminar
3 October 2017
2/18
Outline
Introduction to Intelligent Wells (I-wells)
I-wells control
Reactive
Proactive
Challenges in Proactive Control
Developed framework for proactive optimisation
under reservoir description uncertainties
Intelligent field development case study
Conclusions
Extensions
Acknowledgements
3/18
IntroductionIntelligent Wells (I-wells)
Equipped with down-hole monitoring and
control devices
ICD (Inflow Control Devices) – Single,
fixed position
AICD (Autonomous Inflow Control
Device) – self-adjusting position,
providing a pre-designed, fluid-dependent
flow control
ICV (Interval Control Valve) – Multiple
positions, surface control
ICV provides a flexible production
control, BUT maximum “Added Value”
depends on identifying the optimal ICV
control strategy
From SPE-107676 with
modification
4/18
IntroductionReactive Control Strategy of ICVs
Reactive
• Decisions are based on the
system’s current condition
• Considers Short-term (current)
objectives Production
Improvement
• Fast reaction to recognised
situations
• Potentially can be done using
well intervention
5/18
IntroductionProactive Control Strategy of ICVs
Proactive
• Starts earlier
Mitigates future undesired
problems and/or states.
• Long-term objectives
increased Oil Recovery
• Requires a reservoir model to
forecast production
6/18
Proactive OptimisationProblem Formulation & Challenges
Objective: Find the control scenario of ICVs that
maximises the objective function
Challenges
Large number of control variables
Computationally expensive objective function evaluations (i.e.
reservoir simulator)
Uncertain objective function
𝑁𝑜. 𝑜𝑓 𝑐𝑜𝑛𝑡𝑟𝑜𝑙 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠 = 𝑁𝑜. 𝐼𝐶𝑉𝑠 ×𝑇𝑜𝑡𝑎𝑙 𝐶𝑜𝑛𝑡𝑟𝑜𝑙 𝑝𝑒𝑟𝑖𝑜𝑑
𝑙𝑒𝑛𝑔𝑡ℎ 𝑜𝑓 𝑐𝑜𝑛𝑡𝑟𝑜𝑙 𝑠𝑡𝑒𝑝𝑠
Simulated Reservoir Model
ICVs Control
With Uncertainty
Objective function
(e.g. NPV)
7/18
Objective function (e.g. NPV)
Proactive Optimisation Developed Robust Optimisation Framework
A fast and efficient optimisation algorithm is developed
which can handle large number of control variables
with minimum obj. fun. evaluations SPE-167453, SPE-178918
Accounting for reservoir description uncertainty
Reservoir Model
Optimiser
ICVs
Control
Modified Objective function
- Mean optimisation: Search
for a control scenario which
improve all realisations (to
some extent)
Reservoir ModelReservoir
Model
8/18
Case StudyModel Description & Development Plan
A full-field, consists of two overlaying heterogeneous
reservoirs, each divided into two layers 4 zones,
4 ICVs & 4 Packers to separate zones
Conventional Development Plan:
- 14 producers (single zone)
- 7 injectors
Alternative I-well development plan:
- 3 intelligent producers (commingled)
- 8 conventional producers (single zone)
- 7 injectors
9/18
Case StudyReservoir Description Uncertainty
Formation porosity and permeability, faults (locations
and transmissibility), initial water saturation and
reservoir net-to-gross were the major uncertainties.
3 realisations known as P10 (optimistic), P50 (base)
and P90 (pessimistic) are employed to capture this
uncertainty
Enough to capture the underlying uncertainty?!
0
0.2
0.4
0.6
0.8
1
1.2
01/00 09/02 06/05 03/08 12/10
No
rmal
ize
d F
ield
Oil
Pro
du
ctio
n
Rat
e
Date (mm/yy)
P10
P50
P90
-0.4
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
01/00 09/02 06/05 03/08 12/10
No
rmal
ize
d N
PV
Date (mm/yy)
P10
P50
P90
10/18
Robust Proactive OptimisationEffort
Total Optimisation time using 18 CPUs was lass
than 1.5 days (~ 33 hr).
More than 80% of the improvement was obtained
after 10 iterations requiring only ~3.5 hours
computation time.
-1
0
1
2
3
4
5
0 20 40 60 80 100% In
cre
ase
in N
PV
ab
ove
th
e b
ase
cas
e
Iteration Number
Mean P10 P50 P90
The developed framework is
capable of performing proactive
optimisation of ICVs in a
reasonable time for this
relatively large, full-field model.
11/18
Robust Proactive OptimisationAdded-Value
-14.0%
-12.0%
-10.0%
-8.0%
-6.0%
-4.0%
-2.0%
0.0%
2.0%
4.0%
% C
han
ge in
me
an a
nd
var
ian
ce c
om
par
ed
to
an
I-
we
ll w
ith
fu
lly-o
pe
n I
CV
s
Mean and Variance of all realisations
Mean
Variance
0.0%
0.5%
1.0%
1.5%
2.0%
2.5%
3.0%
3.5%
4.0%
4.5%
5.0%
% in
ceas
e in
NP
V c
om
par
ed
to
an
I-w
ell
wit
h f
ully
-o
pe
n IC
Vs
Reservoir Models
P90
P50
P10
greater
improvement for
the less
favourable
realisations
Improved mean
higher expected
added-value
reduced variance
higher reliability
(lower risk) by
applying the best
ICV control scenario
An extended oil plateau was observed for all realisations.
12/18
Conventional Vs. I-well Development Plan Impact of Robust Proactive Optimisation
-5
0
5
10
15
20
09/01 01/03 05/04 10/05 02/07 07/08 11/09
ΔN
PV
w.r
.t. c
on
ven
tio
nal
we
ll (1
0^7
$)
Date (mm/yy)
I-well (Fully Open)
I-well (Optimised)
End of Plateau Period
End of Drilling
Lower number of wells drilled
Extended plateau by robust proactive opt.
Accelerated field development
I-wells lead to loss mitigated by optimum control
Conventional Development plan
14 conventional producers
I-well development plan
8 conventional & 3 intelligent
producers
-10
-5
0
5
10
15
20
25
NP
V(1
0^8
$)
Date
Conventional-Well
I-well (Fully Open)
I-Well (Optimised)
P90 realisation
13/18 Robust Vs. Single realisation Proactive
Optimisation
Proactive control must be applied early, during the plateau
period, to achieve the highest gain
the reservoir model is most uncertain during the early period.
The importance of robust optimisation is shown by
considering a non-robust optimisation performed using a
single realisation (here P50).
% Change P10 P50 P90 Mean Variance
Robust
optimisation
+0.2 +1.3 +4.3 +1.6 -12
Single-
realisation (P50)
+0.05 +2 +0.3 +0.8 +2.8
Higher Added-value
Reduced uncertainty
Max improvement for P50 but non-optimum performance for other
realisations, lower added value, increased uncertainty
14/18
Conclusions
Reservoir-model-based proactive control should be applied
early, during the plateau period greatest uncertainty in
the model
Single realisation optimisation sub-optimal performance, high
risk.
One of the main reasons why the operators are often unwilling to control
the ICVs/wells proactively. Although a no-control scenario may diminish
the I-well gain!
Robust proactive optimisation is the solution.
Developed robust optimisation framework can efficiently
handle large number of control variables, high computation
time and reservoir description uncertainty
The whole process was performed using a single, high-end PC in a
reasonable time for this relatively large, full-field model
15/18
Conclusions
(Partial) I-well development scenario
Increased, early-time, NPV by reducing the number of wells
to be drilled
May accelerates field development by speeding up the
drilling process
state-of-the-art, proactive optimisation extended the oil
production plateau, ensuring that the early NPV gain was
maintained.
Robust proactive optimisation allows the production
operators to confidently control their I-wells to achieve
maximum expected added-value
lower uncertainty in the operation
16/18
How to select a small ensemble
of realisations as the
representative of all realisations
P10, P50 & P90 are not always
good enough representatives
Developed realisation
selection algorithm: smartly
select an ensemble of
realisations. Tailored to the
subsequent application
A. Visualisation
B. Clustering
ExtensionRealisation selection algorithm
-3 -2 -1 0 1 2 3-3
-2
-1
0
1
2
3
-3 -2 -1 0 1 2 3-3
-2
-1
0
1
2
3
Each circle is one model realisation
Reservoir description uncertainty is quantified by hundreds of
model realisations
Haghighat Sefat, M., Elsheikh, A. H., Muradov, K. M. & Davies, D. R. 2016a. Reservoir uncertainty tolerant, proactive control of
intelligent wells. Computational Geosciences, 20, 655-676.
17/18
ExtensionRobust Completion Design
The developed robust optimisation framework can be
extended to advanced completion design
Reservoir Model
Optimiser
- Type of Flow Control
Devices (FCDs): ICV, ICD,
AICD(V)
- Location, Number (&
strength) of FCDs
- Autonomous, fluid
dependent performance of
AFCDs
(e.g. NPV,
cumulative oil)Reservoir
ModelReservoir Model
Objective function
Control Variables(Completion design parameters)
To be presented in Inflow Control Technology (ICT) Forum, 12th & 13th October 2017, San Antonio, USA.
18/18 New Phase of “Value from Advanced
Wells” JIP (2018-2021)
Theme A: Maximum “Added value” from
downhole flow control completions
Theme B: In-well monitoring and data
interpretation in advanced wells
Modelling Design Control Analysis InterpretationData
mining
AFCDs
AFCD
completions
TIFs
Robust Prod./Inj.
with AFCDs
considering
- Uncertainties
- TIF
Robust ICV
control
- Large fields
- Uncertainties
PTTADTS oil and
gas wells
Test design
Missing
data
Other topics….
Sponsor steered
19/18
Thanks For Your Attention.Morteza Haghighat
AcknowledgementsThe authors are grateful to the sponsors of the “Value from
Advanced WElls” (VAWE) Joint Industry Project at Heriot-
Watt University for funding
Top Related