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