You are on page 1of 9

Modeling of Intelligent Wells

Forecasting Life-Cycle Costs

AUTHORS:
Geir-Ove Strand, geir-ove.strand@exprosoft.com is a research scientist at the
Department of Subsea and Well Technology at SINTEF Petroleum Research, Trondheim,
Norway.

Jake Ansell, The University of Edinburgh, Scotland

Marvin Rausand, marvin.rausand@ipk.ntnu.no, is professor of reliability engineering at


the Norwegian University of Science and Technology

Operators need comprehensive methods to evaluate reliability and forecast life-


cycle costs before they will embrace intelligent well technology fully. This modeling
project aims to provide the tools to make those judgments and make the most of
upcoming subsea development opportunities. Prototypes should be available by the
end of 2001.

Comprehensive methods to evaluate reliability and forecast life-cycle costs are needed
for intelligent well technology to gain operator conifdence. This modeling project aims
to provide those tools.
Since the birth of “smart”, or “intelligent” well concept in the early 1990s, there has been
intensive development of the technology. The goal has been to realize the potential of
advanced well completions.
The first intelligent well system, the so-called SCRAMS system1, was brought to
market in 1996, and in August 1997 was first installed at Saga’s Snorre TLP in the North
Sea. A sketch of this completion is shown in Fig. 1. Other installations have followed.
Generally, these completions include remotely controlled downhole valves with zonal
isolation and monitoring functionality. The remote control coupled with observability of
several zones allows performing continuous on-line reservoir management. Potential
benefits of intelligent systems can be numeous:
- Minimize or eliminate the need for well interventions
- Accelerate production
- Increase/optimize reservoir recovery
- Provide production trend data
- Provide quality reservoir data for support of total field development or field
extension scenario assessments
- Improve performance of other well completion equipment through more favorable
well-stream flow conditions, e.g., sand control devices, safety valves, submersible
pumps/separators and gas lift valves.

But, as with most new technologies, there have been teething problems. While some
installations have been relatively successful, others have been less flourishing. A number
of system failures have been encountered during or shortly after completion of most of
these wells as may be seen in the status overview of installations as of May 2000 (Table
1).
Manufacturers are still struggling to develop and prove “sufficient” reliability of
their most sophisticated intelligent well systems; operators are therefore very reluctant to
fully embrace the new technology although some of the potential benefits clearly have
been demonstrated. Operators are concerned with the higher initial investment and the
poor reliability that has been reported from the first few installations. It has, however,
been demonstrated that the higher initial investment actually has paid off despite rather
poor system reliability4.
The intelligent well concept should not only optimize hydrocarbon recovery
process, but also greatly reduce field development and operating costs. As Table 2 shows,
the trend towards more subsea field developments is rapidly taking off. The level of
activity is right for considerable savings from using intelligent well technology.
The intelligent well concept brings new issues to tackle. Linking reservoir zones
provides inherent difficulties, and using valves and sensors to manage production
remotely adds complications. How should the hydrocarbon reserves efficiently and
reliably be depleted? How will the reservoir dynamics behave? What will be the optimal
well configuration? Given the recent changes in oil prices, cost consciousness must come
to the fore. The criteria for judgement should be a sound balance of potential benefits and
costs. It is therefore necessary to assess the life cycle costs (LCC) for intelligent wells7,8 .
With the expected lifetime of an intelligent well running 15 years, is it possible to
establish a figure for the well’s LCC during this time? A joint research project between
SINTEF, the Norwegian University of Science and Technology (NTNU) and the
University of Edinburgh has been initiated to provide methods to assess this question.
The research strategy calls for a model together with relevant techniques which, while
initially simple, can be elaborated over time guide the development and application of
intelligent well completions.

Modeling aspects
One objective of the joint research project is to quantify the LCC for all types of
intelligent well completions in order to:
• Optimize well completion design and architecture by comparing options.
• Guide the development of field and well operational strategies and requirements, e.g.,
how to effectively handle unwanted equipment and reservoir events.
• Guide minimal acceptable component/system reliability measures (e.g., mean time to
failure) as input to design specifications, etc.
• Assess redundancy and other back-up features.

2
Using the common approach to LCC analysis7,8, we aim to quantify a set of defined
cost categories for intelligent well completions based on well configuration and
operational philosophy. Examples are listed in Table 3.
Prediction of intervention and deferred production costs gets special attention. These
costs are typically where the potential gain is conspicuous when applying intelligent well
technology. A necessary input to the cost estimation is a reliability assessment..
Thorough system knowledge is vital in any reliability assessment. The term system
knowledge covers not only the basic understanding of the components and their
configuration, but also how and why system failures occur and the options available to
restore function. An intelligent well has a relatively simple structure and the reliability of
the well should therefore be easy to model. For LCC modeling, however, three factors
inherent to the nature of the wells pose significant challenges:
1. The reliability of the completion is not “static” in the sense that all component
functions are not necessarily required for the whole design life.
2. The different possible repair and restore options have very distinct effects on the
residual lifetime of the completion.
3. The actions taken and consequences of failure in terms of direct costs and/or well
performance will vary depending on the time the failure occurs, which component
fails, the failure mode, and the state of the system and the reservoir.
These factors are very difficult to assess with existing analytical methods and
techniques for system reliability (see, e.g., Ref. 9). The standard option in such cases is
Monte Carlo simulation. Simulations are, however, time-consuming and seldom provide
sufficient comprehension of the system’s for design and operational planning. In
addition, simulation models require verification from analytical or empirical reasoning or
both.

Modeling approach
A distinct feature of intelligent wells is that the design life is divided into a set of planned
life cycle phases. These are defined in terms of the operational strategies for the well.
During each phase, the system will have specific configurations allowing either selective
or commingled production from (or injection into) different zones.
Our modeling approach is based on the various system components and their
minimal (or critical) functional requirements in each life cycle phase. The effects of each
component failure on direct costs and the performance of the well are assessed.
Components are grouped with respect to required action and associated consequences.
Typical groups with associated failures, actions and consequences are shown in Table 4.
From these groups it is possible to establish potential failure-consequence chains that can
be modeled by a time-dependent event and decision tree. The modeling approach is
illustrated in Fig. 2.
In the simple example in Fig. 2 it is assumed that the system is functioning when
it is installed, i.e., at time t = 0. A first failure may occur some time during Phase I of the
system’s life cycle. The failure may be attributed to one of the groups mentioned above.
The time from start-up at t = 0 until the first failure will be a random variable T =
minimum{T1, T2, …, Tk }, where Ti denotes the time to first failure of “group no. i” for i =
1,2, … , k, and k is the number of different groups. The first failure may be from different
categories, with a probability distribution that is easy to determine when we know the

3
distribution of the Ti-s. Once the failure has occurred and been diagnosed, we have to
decide what to do: ignore the failure or make an intervention. Our decision will depend
on the time of the failure, the reservoir conditions, the time until the planned transition to
Phase II, etc. If we decide to make an intervention we must select the type of
intervention. We may be successful or not. In the example in Figure 2, only one failure
and one decision are indicated in Phase I. In a real situation, several failures might occur
with repeated decisions. The dotted vertical line in Fig 2. illustrates that the failure-
consequence chains in Phase I are not pursued. The transition from Phase I to Phase II
will normally involve the operation of one or more valves which may be successful or
not. If a phase transition failure occurs, we have to make a decision about what to do, and
so on.
For practical systems there is a danger that the complexity of the models will
become overwhelming. In such cases the calculations become cumbersome, and the
degree of uncertainty increases rapidly. In many cases there is a need to simplify the
model by adding restrictions or making assumptions. For example, by limiting the
number of failures in each phase or along each path in the consequence chain and by
disregarding components with high reliability.
Modeling also requires a thorough system knowledge and a wide range of input
data. Reliability data for oil-field equipment are available from databases such as
Wellmaster3 and SubseaMaster10. The reliability data may initially be established from
these databases for similar equipment types, and modified through expert judgement.
If the input data are made available, we may use the time-dependent event and
decision model to establish the Ownership Cost part of the LCC, which may be
calculated based on three different methods::
• Approximation formulas deduced from the model
• Dynamic programming11
• Monte Carlo simulation
We are still working with simple approximation formulas at the current stage of the
joint research project. However, we strongly believe that this approach will increase our
comprehension of the intelligent well system and its operational strategies and form a
sound basis for more detailed assessment by dynamic programming and Monte Carlo
simulation.
The LCC analysis work within the research program will be completed early next
year. A series of case studies are planned with inputs and support from interested
operators. These case studies will allow verification of the models and help establish
specifications to develop software modules. A prototype tool should be available for
testing and further refinements by the end of 2001.

4
FIGURES

Safety valve

SCRAMS Packer

Dual control line

SCRAMS inflow control


valve with sensor/control
module

Sliding sleeve

Polished bore receptacle

Fig. 1. Schematic of the first intelligent well completion (adapted from reference 4).

5
Phase
Decision
First transition Failure Decision
failure Decision failure?

Yes

No

Phase I Phase II Time


Start

Fig. 2 Modeling of system performance by a time-dependent event and decision tree.

TABLE 1 STATUS OVERVIEW OF INTELLIGENT WELL INSTALLATIONS AS


OF MAY 2000*

System Experience
SCRAMS: - Three installations lost all, or most of
Ten installations with a total of 26 inflow the zonal control during, or shortly
control valves (ICV). The number of after installation.
ICVs per well is ranging from 1 to 4.
- Four installations have experienced
loss of some monitoring and/or cable
redundancy.
- Three installations have been 100%
successful to date.

Other (hydraulic operated valves with No known failures.


separate instrumentation):
One installation with 3 valve units.
* based on references 2 and 3.

6
TABLE 2. WORLDWIDE SUBSEA FORECASTS*

• 1999: 287 new subsea wells completed

• 2003: 405 new subsea wells completed

• 95% of all new deepwater (water depths > 300 meters) wells
will be subsea completed

• Subsea production systems 2003: 12 billion US$ worth of


equipment investments
* based on references 5 and 6

TABLE 3. LIFE CYCLE COST CATEGORIES FOR INTELLIGENT WELLS

1. Acquisition costs
1.1 Design, development and testing (with service company)
1.2 Equipment purchase cost
1.3 Installation costs
1.4 Commissioning cost
1.5 Administration and training cost
1.6 Insurance spares cost
1.7 Reinvestment cost
2. Ownership cost
2.1 Operating, (top-side) servicing and consumables cost
2.2 Scheduled intervention cost
2.2.1 Rig, other equipment, spare parts, consumables and man-hour
2.2.2 Deferred production
2.3 Unscheduled intervention cost
2.3.1 Rig, other equipment, spare parts, consumables and man-hour
2.3.2 Deferred production
2.4 Reduced production cost (due to equipment failure)
2.5 Logistics support cost
2.6 Cost of risk (cost of possible rebuilding and clean-up)
2.7 Abandonment and disposal cost

7
TABLE 4. EXAMPLES OF GROUPS WITH ASSOCIATED FAILURE
CONSEQUENCES

Group Failure example Actions Consequences

Requires Failure causes loss of a safety Full workover 1. Well performance


immediate barrier or greatly reduces the permanently
actions. production/injection ability. affected.
E.g., premature/spurious Partial
closure of valves. workover2.
Well performance
Will require Failure causes loss of function temporarily
actions. that is required to enter into a Light affected.
new phase. E.g., fail to open an intervention3.
inflow control valve.
Well performance
No action Failure causes loss of a system None. ‘not’ affected.
performed. monitoring function, or has
minor impact on the
production/injection ability.

1. Remedial actions are performed by retrieving the entire completion string.


2. Remedial actions are performed by retrieving the x-mas tree and/or a part of the
completion string (e.g., completion separated in a seal assembly)
3. Remedial actions are performed without retrieving the x-mas tree or any part of the
completion string (wireline, coiled tubing, etc.)

References
1. Petroleum Engineering Services (International) Limited (now fully acquired by the
Halliburton company). Internet address: http://www.pes.co.uk/

2. Jensen, H.P., Intelligent well summary, Internal memo at SINTEF Petroleum


Research, Trondheim, May 2000

3. Wellmaster: Reliability of Well Completion Equipment. Internet address:


http://www.exprosoft.com/

4. Terdre, N., Long-term reliability goal of next smart well research phase, article in
Offshore, PennWell, Aug. 1999

8
5. Douglas-Westwood Associates: The World Subsea Report 1999-2003, Internet
address: http://www.douglas-westwood.co.uk/

6. Douglas-Westwood Associates: The World Deepwater Report 1998-2002, Internet


address: http://www.douglas-westwood.co.uk/

7. NORSOK: LCC standard O-CR-002, Life cycle cost for production facility,
Download available at: http://www.nts.no/norsok/

8. Kawauchi, Y. and Rausand, M., Life Cycle Cost (LCC) analysis in Oil and Chemical
Process Industries, NTNU, Trondheim, 1999

9. Høyland, A. and M. Rausand, System Reliability Theory; Models and Statistical


Methods, J. Wiley & Sons, New York, 1994

10. SubseaMaster: Experience Database for Subsea Production Systems. Internet address:
http://www.exprosoft.com/

11. Dynamic programming. See Internet address:


http://www.deakin.edu.au/~gecole/bibliography.html

12. SINTEF strategic research program on intelligent wells. Internet address:


http://www.iku.sintef.no/Borebronn/SIPintwell/

You might also like