17 4 22

Journal of Engineering Science and Technology
Vol. 17, No. 4 (2022) 2567 - 2593

© School of Engineering, Taylor’s University
PLANNING AND IMPLEMENTING THE

REDEVELOPMENT OF VOLVE FIELD, NORWEGIAN NORTH
SEA USING KNOWLEDGE DISCOVERY IN DATABASES DATA
MINING AND NUMERICAL SIMULATION TECHNIQUES
ALEXANDER E. MCALLISTER*, LATEEF AKANJI
School of Engineering, University of Aberdeen,

Fraser Noble Building, King’s College, Aberdeen, AB24 3UE, Scotland
*Corresponding Author: alexander.mcallister@me.com
Abstract
In this study, a brownfield development plan was investigated by mining the data
pack from Equinor Volve field repository, modelling and simulating to establish
the quantity of hydrocarbon that is still recoverable therefrom. This is aimed at
identifying the most appropriate technique for recovering additional oil from the
brown field. Firstly, a data-mining algorithm that is based on knowledge
discovery in databases (KDD) was developed, the reservoir model was built and
utilised in history matching and selection of enhanced oil recovery option. The
preceding development achieved 46% recovery thus a window of 8-12%
additional recovery was set as a goal. The field development plan has two
options; the best technical option which had a double peripheral layout spot
pattern with 7 injectors (2 water 5 WAG) and 10 producers (3 of which were
horizontal and 7 vertical). The incremental increase goal was hit, with recovery
increasing by 10% by year 3 and 12% by year 4. However, it was felt the impact
on the environment of such a project was simply too significant thus the
recommended case took the form of 6 producers and 1 water injector. This was
the improved reactivated well trajectory case.
Keywords: Data mining (DM), Enhanced oil recovery (EOR), Field development
plan (FDP), Knowledge discovery in databases (KDD), Water
alternating gas (WAG).
2567
2568 A. E. McAllister and L. Akanji
1. Introduction
This study was to investigate the benefit of using a KDD data-mining algorithm in
the development of brownfields. The energy landscape has changed significantly
from the past, clearly there has been a shift in public opinion. Unconventional
sources along with drilling in extreme environments are more controversial now.
This means more focus must be placed on conventional sources in non-extreme
environments. As well as this the energy transition is fundamental to the survival
of this industry, drilling in environmentally sensitive areas or conducting practices
which are excessively damaging may take away from this. Redeveloping
brownfields has the potential to facilitate necessary oil demand whilst minimizing
environmental risk. This is because the majority of large new fields are discovered
in environmentally sensitive areas i.e., Artic or are unconventional i.e., oil sands,
whereas conventional brown fields in non-sensitive areas are relatively abundant.
The implementation of data mining in the brownfield development process truly
adds a unique selling point. Data has arisen from the peripheries of the global
economy to centre stage and using data to increase field efficiency would transform
countless projects. This adds a novel dimension on how field redevelopment studies
might be done.
There has been past work on combining data mining in the brownfield
development process and to a large degree this implementation has been met with
success. Possibly the most interesting system researched was Top-Down
Intelligent Reservoir Modelling [1]. This technique combined reservoir
modelling with AI and DM. The main selling feature of this method it only
requires field production data and some well log data i.e. porosity, thickness and
Swi. Other data can be added at the engineer’s prerogative. It is this flexibility in
data requirement that sets this method truly apart, a main disadvantage is that in
order to get reliable results a minimum of 5 years of production history is required
along with a large number of wells.
The authors have proposed to compare the impact of the data mining by
comparing a data-mined case to the production prediction of the original
development. This is a novel way of redeveloping a field and will use data
extracted to improve the recovery of the brownfield. The scientific contribution of
the study is significant, a KDD algorithm was fashioned in the pursuit of an
effective data-mining system whilst using an industry standard reservoir modelling
software to create the model. The study was organised as follows: a literature
review was conducted on brownfields, data mining and reservoir modelling. A
data-mining algorithm was created and used. An investigation into the previous
development plan was undertaken and this information along the information
acquired through data mining was used to create a new development scheme,
enhancement and development strategy.
2. Methods and Approach

The method of this study entails seven steps as shown in Fig. 1. Figure 1 shows the
methodology as i) background study, ii) data mining, iii) reservoir modelling, iv)
development plan, v) development scheme, vi) enhancement, vii) finally
development strategy. This is not strictly chronological, but each new step depends
on the previous and each previous step is supplemented by the subsequent.
Journal of Engineering Science and Technology August 2022, Vol. 17(4)

Planning and Implementing the Redevelopment of Volve field . . . . 2569
Fig. 1. Volve field development methodology

involving 7 sections with a general downward sequence.
2.1. Background study

A brownfield is different from a green field in that it has already been developed.
In an effort to extend economic life more and more operators are returning to such
fields, often involving the use of improved oil recovery techniques. Brownfields
have numerous advantages including but not limited to existing infrastructure and
a treasure trove of available historical data.
The core subject is to calculate the potential brownfields offer. It is seldom that
an operator would return to a field where the situation remains the same; it’s more
common that it is felt that there have been new reservoirs in the field or extensions
in the boundaries of known reservoirs. Using modern technology is a key attribute
in determining the increased potential of brownfields. The Energy Information
Administration (EIA) researched in order to determine the potential of brownfields;

they presented that, of 200 well known fields worldwide in completely different
environments from 1977 to 1993, 44% had increased reservoir by 2 to 3 times, 22%
by 3 to 4 times and 9% by at least 4 times [2]. This is supported by Antariksa et al.
[3] who in their study, presented the undeveloped oil present in multilayer
reservoirs. The reasons for revisiting the fields include the following:
• Remaining not swept oil.
• Even at 90% water cut, a lot of oil can be produced especially in oil wet reservoirs.
• More modern technology such as advances in EOR and reservoir modelling
which have progressed considerably in the past 20 years.
• Oil price dictates the feasibility of brownfields if prices are high enough [3].
Al-Khaledi and Qadri [4] presented a study on a Brownfield project in the
Middle East (Kuwait) from the Kuwait National Oil Company. It too agreed with
Antariksa et al. [3]; putting the case that these projects do have great economic
potential. For this case in the Burgan Fields there was success but only at the
expense of a considerable amount of economic and industrial input. Due to the
complexity, it absorbed the most proficient teams, it can be argued this difficulty
could be an advantage, becoming a specialisation within the industry. Lessons from
brownfield projects in mature regions like the North Sea could be exported and
employed around the world [4].
Kumar et al. [5] presented clear challenges in the execution of offshore
brownfield projects. Chief constraints are available space on existing platform and
minimum production shutdown requirements. Although daunting these challenges
may appear there are ways to mitigate these technical hurdles. Implementing
electrical changeover philosophy minimising production shutdown and optimising
space with innovating machine learning can be solutions to these challenges.
Being an already developed field there are huge caches of data which require
analysing and data mining. Data mining (DM) involves extracting useful
information in the most efficient manner and showing tendencies. It amalgamates
three masteries: statistics, artificial intelligence and machine learning. It uses these
three to commoditise big data; it is through this technique that data has become the
most valuable resource in the world, overtaking oil in 2017 [6].
Traditionally, the oil industry has been sceptical about the use of big data;
preferring the traditional physics-based approaches. However, Balaji et al. [7]
presented the recent advances and applications of data-driven methods in the oil
industry. Still, a strong grasp of the traditional methods is needed in data driven
models, but the process starts off with utilising data to identify issues and their
solutions. It was found that organisation and refinement of the data were the most
important components of an efficient data driven process.
It is put forward by Benjamini and Leshno [8]; that the cause of the mass
utilisation of DM and KDD in particular in the recent years is the accelerated
production of cheaper computers, software improvements, inexpensive data storing
along with the automated collection of data and the amount of data that continues
to be gathered in real time. This opinion is shared in this study, it’s obvious that
DM use will keep expanding as it almost always creates a competitive advantage.
The four key benefits of data mining are [9]:

• Automatic decision making: real time data analysing permits organisations to

see anomalies in data which are then examined.
• Prediction and forecasting: knowing the detailed history permits predicting the
future accurately.
• Reduction in costs: maximising efficient use of resources drives down the
operating expenditure.
• Tailored service: using consumer data, profiles can be formed to understand
the customer and better tailor to their needs [10].
With such a system, there are variations on the methods used to data mine.
Although different, they centre around three steps: pre-processing, data mining and
results validation. Pre-processing is where a target is determined before a specific
algorithm is used. This minimises the data to as low an amount as is possible which
saves time, but also to have enough data set so as to view indications. Any data
which cannot be used is removed and what remains is mined. Data mining then
incorporates 6 steps [11]:
i. Detection of Anomalies.
ii. Association learning: realising correlations between different variables.
iii. Clustering: combining data based on similarity.
iv. Classifying: implementing a known structure to the data.
v. Regression: determining a function which fits the data models best.
vi. Summarisation: a presentation of the data [10].
The final step which involves results validation attributes hypothesis to the
variable. There can be an infinite number of correlations between variables which
are not related to each other (coincidences). Then there will be variables which
make an effect and need to be presented. It is essential this covers the entire data
set and not an artificially created sample [10, 12].
From the data mining inputs can be made into the reservoir model. This type
of modelling is vital in modern development projects. It uses data to model
what we think a reservoir looks like. It then goes further and using the structure
forecasts how fluid may flow through it over many years. It uses a cooperation
of mathematics, science and computer science and its benefit to this study
is huge.
The exciting prospect of this is with increasing development in artificial
intelligence (AI) the DM will become more and more refined, efficient, and
automated. Windarto et al. (2017) discussed the implementation of an AI algorithm
developed by the British Petroleum (BP) in Indonesia. An artificial neutral network
was implemented using backpropagation. Although some alterations were
recommended it was concluded that this system was valuable to BP in further
developments [13].
The Volve field is a brownfield, initially developed by Equinor and then
abandoned. The field was discovered in 1993, in the central North Sea at a relatively
shallow depth of 80m water, 5km north of Sleipner East.

2.2. Data mining

This stage started from the onset and continued until the termination of the study. The
data set had 11 folders containing the information and are detailed as follows: [14]
a) Geological Interpretations.
b) GeoScience.
c) Production Data (monthly and daily on a field and well basis).
d) Reports (discovery report in English and PUD report in Norwegian).
e) Reservoir model (Eclipse)
f) Reservoir model (RMS
g) Seismic (from 2 surveys in 2002 and 2010).
h) Well logs.
i) Well logs per well (same data as above, organised on a well basis).
j) Well technical data.
k) Realtime drilling data.
The method was centred on the KDD format including pre-processing, data
mining and deployment. Pre-processing this data set was vital in setting out a clear
view of what data needed to be gathered (business understanding). This was done
in 5 sections: previous reports, reservoir model, production profile, seismic and
well logs. The data mining algorithm dubbed Volve Oil Field Data mining (VOF-
DM) algorithm was designed based on Decision Tree Data Mining (categorical)
where a series of statements were put forward and only the data which was on the
affirmative side of these statements went forward. There were 5 nodes in the
decision tree: broad selection, focused selection, format selection, current version
and finally evaluation and interpretations. The 5 stages filtered out various data
types which were deemed surplus to requirements. The final stage is deployment
with the next stage to initiate the reservoir modelling section. The decision tree
flowchart algorithm is depicted in Fig. 2.
Figure 2 shows that the pre-processing is in 5 topics. Broad selection takes
folders from the original data pack which fit these five sections, focused selection
goes into these folders and filter key files, format selection filters correct software
or language, current version filters dated, interpretation/evaluation filters through
the remaining to remove surplus data, deployment occurs when the data is ready to
be put into the reservoir model.
2.2.1. Model validation

The model validation took two forms: in-sample validation and out-of-sample
validation. To start with, the in-sample validation, for this we assess the depth of
the nodes. This is standard in assessing whether the branches of the decision tree
extend too much or not far enough. From running both greater and lesser number
of nodes it was assessed that fewer, provided an ineffective system which did not
provide a significant difference than not using the DM at all. Adding more created
a more complex system which did not significantly affect the results.
For the out-of-sample validation, a new data set was sourced courtesy of the Oil
and Gas Authority Open Data. Using the same principles and steps, this data was

input into our model. This method of cross-validation is highly reputable and
considered an essential part of accurate model validation. The results of this were
clear, with minimal modification a functional prediction was created. This clearly
shows the predictive performance of the DM algorithm.
Fig. 2. VOF-DM algorithm decision tree.
2.2.2. Results
Combining the DM with pre-processing and evaluation there were 7 stages. The
method was successful: out of 40,000 files initially in the data pack, only 80 were
remaining. This reduced down to 0.2% in terms of files required. The final volume
was 2GB of storage compared to the original pack size of 5TB.
2.2.3. Efficacy of data mining

In order to determine the efficacy of the data-mining a simple yet effective method
was conducted. Out of the remaining files which were not extracted, 100 were
sampled at random. For each of these files two questions were asked:
i. Do the contents of this file add value to the thesis?
ii. Do the contents of this file affect the final results of the thesis?
Out of the 100 files 4 answered yes to the first question and 0 answered yes to
the second. This leads to the conclusion that based on the quality check conducted,

out of the remaining files 4% would have added value to the thesis and 0% would
have changed the outcome.
2.3. Reservoir modelling

The model has dimensions of 108×100×63, with a cell count of 680,400 and
183,545 active cells. The grid size is 50×50 m with 1-3m of thickness. This model
is shown along with oil producing wells (Fig. 3).
Fig. 3. Well location in reservoir model

as provided in the Volve data pack [11].
The 8 oil producing wells are labelled: F-1C, F-15D, F-5, F-14, F-11B, F-12,
F-15C and F-4 The values from the reservoir model indicated a STOIIP of 22x106
m3 [14]. In order to run models and produce results a number of steps were required.
A history match was conducted on the model and through information provided in
the data mining this history match was improved.
The first simulation took approximately 6 hours; as it was a very detailed run.
To decrease this time, the number of CPU processors used were increased to 6 from
4 and the CPR solver was activated. This reduced the simulation time to 2 hours.
The computer utilised was the HP EliteDesk 800 G3, it has Intel Core i5-7500 3.4
GHz (Quad Core), 8GB DDR4-2400, 256 SSD, Intel HD Graphics 630.
2.4. Development plan investigation

The development plan was an investigation into certain parameters of the previous
development. This stage was vital; running the simulation takes a significant length
of time, thus well placement must be considered thoroughly. Using field properties
and data to best assess the method of redevelopment is pertinent.
2.4.1. Identifying existing producing wells and injector wells

The field had 24 wells, of these only 11 were Volve field development wells, that
was because although all wells were useful in mapping the reservoir and appraising
its value, only 11 were considered economically viable. Wells PIL-N and PIL-NW
were drilled into the Utsira formation for water production. The purpose of the now
11 wells and 2 water wells are detailed in Table 1 [14].

Table 1. Table of Volve development wells [14].

Date Operation
Well Category Comments
Commencing
I-F-4 Injector 20/04/2008 First injector
I-F-5 Injector 18/08/2008 Second injector
I-F-1B Injector 01/10/2013 Final injector
I-F-4G Injector N/A Deviated well
P-F-12 Producer 10/02/2008 First producer
P-F-14 Producer 09/07/2008 Second producer
P-F-15C Producer 06/03/2009 Third producer
P-F-11B Producer 23/07/2013 Fourth producer
P-F-15D Producer 16/01/2014 Fifth producer
P-F-1C Producer 22/04/2014 Sixth producer
P-F-5 Producer 18/04/2016 Converted injector
PIL-N W. Producer 07/09/2008 First water producer
PIL-NW W. Producer 08/07/2008 Second water producer
2.4.2. Recovery factor from primary production

With the first injector well operational from 20/04/2008, about 2 months from first oil
the recovery from primary production was low. 1.731x105 Sm3 of oil was produced in
this time with STOIIP at 2.2x107 Sm3; giving a primary recovery factor of 0.8% [14].
2.4.3. Recovery factor from the secondary production due to water injection
Using the simulation, the recovery factor at the end of the secondary production
was found to be 46%. Note to reader in some older Equinor documents on the Volve
field the recovery factor was stated as 54%, this value was outdated as it did not
take into account the revaluation of the field size due to data acquired from well A-
F-11A. This well showed that the reservoir had an increased boundary with more
favourable conditions and so increased the STOIIP of the reservoir.
2.4.4. Material balance giving the relative component energy of the system
The MBAL values are detailed (Table 2).
Table 2. MBAL values [14].

Input Value Source
GOR (R) 114Sm3/Sm3 Discovery report
Oil gravity (γo) 30° API Discovery report
Gas gravity (γg) 0.883 Discovery report
Water salinity 151,200ppm Print file
CO2 concentration 9.5% Discovery report
N2 concentration 0.46% Discovery report
H2S concentration 3.5ppm Discovery report
Reservoir temperature (T) 230°F Print file
Bubble point pressure (Pb) 273 Bar Discovery report
Initial reservoir pressure (Pi) 329.6 Bar Production results
Porosity (ɸ) 0.225 Discovery report
Connate water saturation 0.203 Log reports
OIIP (N) 21,967,455 Sm3 Print file
PVT data Entire table Ccs results data file
RelPerm tables 3 tables Relperm data files
Production history Excel file Print file
Reservoir thickness (h) 18m Discovery report

The PVT data gives the oil formation volume factor (Bo) and oil viscosity as a
function of pressure for cases. These are increasing GOR values from 27.4 Sm3/Sm3
to 156.9 Sm3/Sm3 [14].
The RelPerm values were given in large, detailed tables for each region in the
model and are shown graphically in Figs. 4(a) and (b) [14].
rg
rog
r
S
(a) Graph of Krg and Krog against Sw.
ro
r
(a) Graph of Krw and Krow against Sw.

Fig. 4. Graphs of Volve RelPerm [14].
This data was input, a model for the aquifer was chosen, and regression was
conducted. This depicted a system where the overwhelming majority of the energy
originates in the water injection (80+%) with a significant minority arising from
fluid expansion (5-15%), a minimal input from PV compressibility (5%) and
virtually no contribution from water influx. The fact that the field is faulted so
intensely has a significant affect, this is confirmed by the variety of fluids in place.

2.4.5. Aquifer strength and strength of the gas cap

As discussed, the natural aquifer strength is minimal, the majority of energy is
through the water injectors. With initial reservoir pressure at a greater value than
the bubble point pressure the gas cap size is also zero.
2.4.6. Characterisation of the fault pattern

There are 29 faults in the reservoir model which have been input [14].
2.4.7. Completion strategy plan

From the former development 11, separate regions were identified within the Volve
field. The associated their fluid contacts are depicted in Table 3.
Table 3. Table of structures WOC, GOC [14].

WOC GOC
Structure
(mTVD) (mTVD)
1- Northwest 3,200.0 500
2- North 3,200.0 500
3- Purged Water 3,025.0 500
4- East 3,200.0 500
5-Prospect N (Eastern Part) 3,000.0 500
6-Prospect Northwest 3,200.0 500
7-Prospect N Upside 2,700.0 500
8- Southwest 3,200.0 500
9-Aquifer West 2,700.0 500
10-Prospect Northwest 3,200.0 500
11-Purged Water 2,910.0 500
12-Outside 2,700.0 500
The GOC depth at 500mTVD means that it is above the top of the structure, this
indicates that there is no gas-cap. From the 11 regions it is possible to assess the
differing fluid properties, OIIP and recovery factor. These are presented (Table 4).
Table 4. 11 Volve regions and total field by OIIP,

production, remaining, transmission and recovery factor [14].
Region OIIP (Sm3) Prod. (Sm3) Rem. (Sm3) Net Input (Sm3)
1 4,350,000 4,254,000 1,956,000 1,859,000
2 4,347,000 4,184,000 2,119,000 1,956,000
3 2,621,000 0 1,094,000 -1,526,000
4 8,113,000 1,291,000 4,937,000 -1,885,000
5 0 0 19,000 +19,000
6 1,404,000 216,000 1,200,000 +13,000
7 0 0 3,000 +3,000
8 175,000 0 134,000 -41,000
9 0 0 0 0
10 329,000 0 187,000 -142,000
11 628,000 36,000 338,000 -254,000
Field 21,967,000 9,981,000 11,987,000 ----
Average

The permeability and porosity of these regions with the average of all the
regions (field) at the bottom is detailed (Table 5).
The averages for the regions were gained from two (2) sources, firstly from
information acquired in the data mining process and secondly from the results of
the reservoir modelling.
Table 5. Average petrophysical properties
by region as provided in the Volve data pack [14].
Region PERM X (mD) PERM Z (mD) Kv/Kh PORO
1 1,798 503 0.280 0.2352
2 1,921 603 0.314 0.2304
3 1,966 588 0.300 0.2288
4 1,697 522 0.308 0.2215
5 2,881 723 0.251 0.2356
6 648 208 0.321 0.2063
7 382 128 0.335 0.2145
8 1,216 275 0.226 0.2504
9 601 200 0.333 0.2096
10 826 242 0.293 0.2017
11 2,278 497 0.218 0.2459
Field Average 1,799 493 0.274 0.2250
2.5. Development scheme

The goal was to increase the field recovery factor from 46% to 54-58%. This 8-
12% incremental increase involved seven (7) components: years of production,
number of wells, regions of interest, well positions and spot patterns, completion
design and model testing as detailed in the logic flow diagram (Fig. 5).
Fig. 5. Logic flow diagram of development scheme

(2 phases containing 5 and 4 stages respectively).

The first phase is determining the years of production, number of producer

wells, number of injector & EOR wells, regions of interest and well positions &
spot patterns. The second phase is the model testing, completion design, optional
reassessment of plan and infrastructure & development implementation. The
different sections of Fig. 5 are described in the following paragraphs. The activities
were conducted in a methodical downward system with the results of the previous
step affecting the next step.
2.5.1. Years of production

The number of years of production is a difficult number to estimate and more often
than not estimated years on oil and gas projects are incorrect. For instance, in the
original development of the Volve field, production lasted 4 years longer than
initially anticipated. Instead of giving an inaccurate estimate as to how long oil
production may last before any redevelopment is actually conducted a minimum
production time was estimated. The minimum timeline, Tm, is estimated as shown:
𝑅𝑓 ×𝑁
𝑇𝑚 = 𝑄𝑜(𝑎𝑣𝑒𝑟𝑎𝑔𝑒)
(1)
where, the recovery factor (Rf) is 10%, the oil initially in place (N) is 22x106m3 and
average flow rate (Qo) is 2000m3/d. This gives approximately 3 years. Note in the
economics section, the payback time from first oil is calculated and thus gives a
more precise minimum production time.
2.5.2. Number of producer wells

The number of producer wells proposed for the redevelopment is estimated as
shown below:
𝑅𝑓 ×𝐻𝑝
𝑁𝑝𝑟𝑜𝑑𝑠 = 𝑡×𝐵 (2)
𝑜𝑖×𝑄𝑜𝑖
where Nprods is the number of producers, Rf is the recovery factor (0.1), Hp is the
hydrocarbon pore volume (26MMm3), t is the time in days which was stated as 3
years, Boi is the initial oil formation volume factor (1.17 resm3/Stm3) and Qoi is the
estimated initial rate of production (200 m3/D). This gives approximately 10
producing wells requirement for the redevelopment which is implemented as the
number of producer wells in this redevelopment plan. Again, in the economics
section this value is checked as to be economically feasible.
2.5.3. Injectors and EOR

The purpose of injectors was two-fold; to maintain pressure and to increase sweep
efficiency. For pressure maintenance, a minimal number of water injector wells are
required. 3 were used in the previous development with one converted to a producer
later on. Thus, for the purpose of pressure maintenance 2 injector wells were
implemented. Given the high recovery of 46% so far, some form of EOR would
have to be implemented to increase recovery.
The importance of selecting the correct EOR method is huge, with around 90%
of all conventional light oil discovered still in the ground and with current
technologies only capable of producing an additional 12%, application of appropriate
new technologies is paramount and Neuro-Fuzzy selection could pose a viable
solution [15]. The Neuro-Fuzzy technique determines the suitability of an EOR

technique to a field, variables are considered based on degree of variance. Some

variables are considerably more sensitive than others, for instance permeability would
not be considered critical for injection of gases whereas gas is [16].
The main considerations discussed for EOR selection were CO2 availability,
infrastructure, reservoir characteristics, aquifer and natural fractures. Miscible CO2
and steam projects account for 76% of EOR projects worldwide, with 21% made
up of miscible hydrocarbon gases, polymers and combustion. The remaining 3% is
made up of surfactants, microbial and miscible acid gas [15].
There were four different EOR methods considered, these were: chemical,
polymer, gas injection and water alternating gas (WAG). Both gas and WAG have
a minimum pressure and temperature requirement (Minimum Miscible Pressure)
which the reservoir exceeds. However, as the reservoir does have a high
temperature and pressure (230oF and 329 bar) this excludes polymer and quite a
number of chemicals. Unfortunately, the chemicals that do operate effectively at
these conditions tend to have a greater potential to be environmentally damaging.
As a result, both polymer and chemical were excluded, with the only options
remaining the aforementioned gas injection and WAG, both were tested in the
model and WAG was considerably more effective, thus it was the method used
going forward.
2.5.4. Regions of interest

Based on investigating the region potential for oil production, five (5) out of the
twelve (12) regions were isolated. These were regions 1, 2, 3, 4, 6. Their positions
are shown (Fig. 6).
Fig. 6. Volve field by FIPNUM region.
Eleven (11) regions colour-coded with the key to the left and the map to the
right. Regions are based on similar fluids in place. (Made in Petrel)
2.5.5. General well position and spot pattern

The spot pattern is a dual periphery as displayed with the orange triangles
demonstrating producer locations and red rectangles demonstrating WAG/injector
well locations (Fig. 7).

Fig. 7. Oil Saturation across the field.
Orange colour triangles represent producer well location and red rectangles
represent WAG/water injector positions. A peripheral injection pattern was used.
(Made in Petrel) The reason for the wells being placed in these general areas is because
these are the regions of interest highlighted in the previous section and in Fig. 6.
2.5.6. Specific well locations and case development

Specific well locations were allocated by running the base case which was the
original development well location. The most effective wells were noted and then,
new wells were made with similar but improved trajectory as to those from the
original development. This is an improved reactivation of the original development.
Then series of redevelopment cases were created with new wells, with new well
locations being optimised by changing locations. This took the form of 10 producer
wells and 7 injectors. The final case (case 7) was an amalgamation of the most
effective reactivated original wells and the effective new wells. The different cases
are explained (Table 6).
Table 6. Case number, name and description.

Case Name Description
1 Base case Production Prediction Based on original
development
2 Reactivated well Similar wells as the base case with improved
trajectory case trajectory and modernised rules
3 Redevelopment Newly created well locations, 10 vertical
case A producers and 7 water injectors
4 Redevelopment Same wells as Case 3 with 10 producers and
case B most effective 2 water injectors
5 Redevelopment Same wells as Case 3 with 10 producers, 1 water
case C injector and 6 WAG wells
6 Redevelopment Same as Case 5 with 3 ineffective producers
case D replaced by more productive well locations
7 Double periphery A merger of Cases 6 and 2, with the peripheral
(final case) spot pattern as Case 6 and the inclusion of the
most effective producers from Case 2

The logic flow diagram (Fig. 8) describes how these cases were formed, starting
with the base case and from the base case the Reactivated Well Trajectory Case
was formed. Then starting from scratch from the development scheme plan a new
case was formed with new wells (Case 3). This was improved creating development
plans B, C and D. From here Case 7; The Double Periphery Final Case was born
from the best parts of Case 2; Reactivated Well Trajectory Case and Case 6; The
Redevelopment Case D.
Fig. 8. Logic flow diagram of well location and case development.
The case development is broken into 2 paths the first highlighted in red draw
inspiration from the original development and the second in blue draw inspiration
from the new development. The trail of thought used a double path to come to the
final case. The first highlighted in the red box (cases 1 and 2) used the reactivated
and improved versions of the original development plan well positions. The second
path (cases 3-6) were based on completely new wells from research into the
geology and properties of the reservoir. With each successive case from 3-6 being
an improvement on the previous. Case 7 then connects cases 2 and 6, in essence
this is the amalgamation of the best parts of cases 2 and 6, using the most effective
wells from case 2 and combining them with the spot pattern and effective wells
from case 6.
2.5.7. Summary of cases

A summary of the 7 cases is illustrated (Table 7).

Table 7. Summary of redevelopment cases.

Vert Horz/Deviated Total Rf
Case
Wells Wells Wells
1 1 6 7 52.0%
2 1 6 7 53.2%
3 17 0 17 49%
4 12 0 12 50.0%
5 17 0 17 51.5%
6 17 0 17 52.6%
7 11 6 17 56.0% (58% year 4)
On examination of Table 7, it may appear that in fact cases 1 or 2 are a lot more
attractive, as they gain a reasonable recovery factor with a significantly smaller
number of wells. However, it must be stressed that both cases 1 and 2 could not
produce economically after the 3-year timeline and would have to be shut down.
Case 7 on the other hand could keep producing economically for 4+ years. The
economic comparison of these is described in the economics section. It may also
appear unusual that cases 1 and 2 both achieve a higher recovery factor than cases
3, 4 and 5 whilst using less wells. The reason for this is due to the positioning and
trajectories of the previous development wells were clearly superior to the
positioning and trajectories in these cases.
2.5.8. Final case

Case 7 – Double Periphery Final Case. The final wells are described (Table 8).
Table 8. Development wells for Case 7.

Well Type Case Origin Horizontal/Vertical
P-F-11B Producer Case 2 Horizontal
P-F-12 Producer Case 2 Vertical
P-F-15C Producer Case 2 Vertical
P-F-15D Producer Case 2 Horizontal
P-F-1C Producer Case 2 Horizontal
P-F-5 Producer Case 2 Vertical
R-2-P-11 Producer Case 6 Vertical
R-6-I-1 WAG Case 6 Vertical
R-1-I-6 Water Injector Case 6 Vertical
R-3-I-7 Water Injector Case 6 Vertical
This final case combines the best parts of Case 6 and the Reactivated Well
Trajectory Case (Case 2). This was the most productive case fielding a recovery
factor of 56% after 3 years and a recovery of 58% after 4.

3. Outputs
Figures 9-12 display the oil production rate, pressure, water injection rate and water
production rate against time respectively. Figure 13 displays the recovery factor
against time for Case 7; the double periphery final case.
e e e ent
rigin e e ent
te
ti n
i r
te
Fig. 9. Oil production rate versus time for Case 7.

r
ir ress re
e e e ent
rigin e e ent
eser
te
Fig. 10. Reservoir pressure versus time for Case 7.

20,000
Redevelopment
Original Development
Water Injection Rate (Sm3/d)

16,000
12,000
8,000
4,000
Date
Fig. 11. Graph of water injection rate against time for Case 7.
From 31/12/2007 to 31/12/2019 water injection rate is displayed with the

orange representing original development and blue redevelopment.
30,000
Redevelopment
Water Production Rate (Sm3/d)
22,500 Original Development
15,000
7,500
Date
Fig. 12. Water production rate against time for Case 7.
From 31/12/2007 to 31/12/2019 water production rate is displayed with the

orange representing original development and blue redevelopment.

e e e ent
rigin e e ent
tenti rt er e er
t r
er
e
te
Fig. 13. Recovery factor against time.
From 31/12/2007 to 31/12/2020 recovery factor is displayed for Redevelopment

(blue), Original Development (orange) and Potential Further Recovery (grey).
Figure 14 illustrates the gas flow rate against time and Fig. 15 illustrates the liquid
flow rate against time.
s n e ti n te
s r ti n te
rte
s
te
Fig. 14. Gas flow rate against time.
From 31/12/2007 to 31/12/2019 gas flow rate is displayed with the orange
representing gas production rate and blue gas injection rate.

ter n e ti n te
ter r ti n te
te
i
i
te
Fig. 15. Liquid flow rate against time.
From 31/12/2007 to 31/12/2019 water flow rate is displayed with the orange
representing water production rate and blue water injection rate.
4. Results and Discussion

4.1. Redevelopment case
The redevelopment case with the best technical success was Case 7 with a recovery
factor at 56% with a year extension bringing the recovery up to 58%. This was the
summation of 47 different simulations and 65 well choices. The main reason for
this choice its high recovery.
An additional advantage of this case it was developed so that it has a surplus
production of gas and water to injected. The reason why this is, is because it used
a more efficient use of water and gas from its efficient WAG regime. The additional
10 wells in case 7 compared to case 2 are significant, Table 9 shows the comparison
between these cases (Table 9).
Table 9. Case 7 summary compared to Case 2 summary.

Case 2 -
Case 7 - Double
Reactivated
Periphery Final
Well Trajectory
Producer Wells 10 6
Water Injector Wells 2 1
WAG Wells 5 0
Horizontal Wells 3 3
Vertical Wells 14 4
Recovery Factor 56% (58% in 4 years) 53.2%
Total Oil Produced (Sm3) 2,298,500 1,700,000
Total Gas Produced (Sm3) 1.8x109 0.269 x109
Total Gas Injected (Sm3) 1.5x109 0
3
Total Water Produced (Sm ) 1.3x107 1.26x107
Total Water Injected (Sm3) 1.3x107 1.4x107

4.2. Redevelopment facilities

Although there was substantial water production how reinjection, at times there is
a deficit to be made up by drilling a well into the nearby aquifer in the Utsira
formation. Systems will be available for the dumping of water if the injection
system is down. Almost all of the injected gas is from produced gas. However again
a pipeline is recommended to import gas when there is a deficit and export when
there is a surplus. It is a complex process to inverse the flow in the pipeline however
there are facilities on both onshore and on the facility which are designed for this.
It has been determined that using a jack-up rig with processing facilities, with
oil exported by shuttle via storage tanks is the best option. Similar to the first
development which was successful to Equinor. The facility will be leased and
capable to be positioned at desired location, drill producer and injector wells,
process reservoir fluid, inject water and gas, export product oil to shuttle,
accommodate all personnel (120 approx.) [14]
The facility will be capable of processing reservoir fluids in the form of
separation and oil stabilization, gas compression, gas export, water treatment and
injection. Power generation is from the gas turbines. [14]
Product quality oil is exported to a storage ship with the load capacity identical
to the production capacity of 9,000 Sm3/d. This vessel is moored to a buoy 2 to 3
km from the facility and has a capacity of 1,000,000 STB with an unloading
capacity of 6,000 Sm3/h. The measurement of oil will take place both from the
facility and the tanker [14]. Several chemicals are to be used to deal with potential
fluid problems. The operation of the facility will be up to the contractor however
the operator has a duty to monitor.
4.3. Economics
CAPEX and fixed OPEX is calculated based on the previous development (Table
10) (2005NOK = 0.1537 2005USD) (2005USD = 1.32 2020USD) [17].
Table 10. CAPEX Estimations [11].
Cost million 2005 Cost million Cost million
NOK 2020 NOK 2020 USD
Platform Rental and 499/y 76.70/y 101/y
Underwater Equipment
Platform Operation and 222/y 34.12/y 45/y
Maintenance
The total facility cost is estimated at being $303 million for the Platform Rental
and Underwater Equipment, $135 million for the Platform Operation and
Maintenance. This gives a total of $438 million
The fuel prices are based on projections from the EIA in May 2020. With the
cost of gas at $3.25 per mmscf and oil price between 2021-2024 at $70. [2] The
effectiveness per well is also illustrated (Table 11).
Each well is viable as the revenue from each e ’s production exceeds the cost
of drilling. An economic overview of the redevelopment is highlighted (Table 12).

Table 11. Cost benefit analysis of producer wells.

Oil Produced (Sm3) Revenue (million USD) Viable
P-F-11B 370,000 163 Yes
P-F-12 270,000 119 Yes
P-F-15C 179,000 79 Yes
P-F-15D 300,000 132 Yes
P-F-1C 470,000 207 Yes
P-F-5 160,000 71 Yes
R-2-P-11 157,500 70 Yes
R-2-P-5 93,500 41 Yes
R-2-P-13 93,500 41 Yes
R-6-P-1 205,000 91 Yes
Total 2,298,500 1014 Yes
Table 12. Economic summary of Volve redevelopment.

Annual 3 Year Total
mill USD mill USD
Drilling ---- 426
Platform Rental and Underwater Equipment 101 303
Platform Operation and Maintenance 45 135
Export Cost 0.12 0.36
Gas Injection --- 0.17
Water Injection --- 17
Additional 4th Year Cost* --- 146*
Decommissioning --- 30
Total Cost --- 911.53+146*
Oil Production 338 1014
Gas Production --- 0.21
Additional 4th Year Production Potential* --- 200*
Total Income --- 1014.21+200*
Net Revenue --- 102.68+54*
*A further 2% recovery in the 4th year could bring in an additional $200
million, bringing an additional $54 million after the $146 million platform costs.
This would be the recommendation as it would increase the net revenue by
approximately 50%.
With a discount rate (r) at 6% this gives an NPV at $106 million, an IRR of
27% and a UTC at $60.8/bbl. Assuming first oil is struck a year after project
conception the payback time would be halfway through the third year (2 and a half
years from first oil). Comparing this approach with the production prediction
approach (case 2) is shown in table 13.
Table 13. Economic comparison of Cases 2 and 7.

Case 2 – Case 7 – Double
Production Peripheral Final
Prediction
Cost 3 Years mill USD 657 912
Income 3 Years mill USD 750 1,014
Net Revenue 3 Years mill USD 93 102
Cost 4 Years mill USD 803 1,058
Income 4 Years mill USD 830 1,214
Net Revenue 4 Years mill USD 17 156

Although there is comparable economic success after 3 years with only $9 million
difference, from four years and further case 7 is by far the most profitable. This
affirms that case 7 is the prudent technical choice over 4 years, however in terms of
the recommended choices by the authors it would have to be Case 2. This was a
difficult recommendation to make, however the facts remain that Case 2 utilizes 10
less wells leading to a much-reduced decommissioning cost and equally vital a far
significant impact on the local environment. The energy landscape is changing, and
it is important to make sensible steps to be on the right side of that, thus it would be
recommended to utilize Case 2 over 3 years and then stop production.
The carbon consideration of this redevelopment is, like with any oil and gas
development something that has to be considered. Assuming a value of 2.71 tons
CO2/m3 of oil, the amount of CO2 released due to the redevelopment is 4.61 million
tons in 3 years. Efforts were made throughout this development to bring about a
reduced emission project with limiting environmental damage. This included the
EOR choice of reinjected reservoir gases in a WAG format, avoidance of any
potentially excessively damaging chemicals and the location itself, as the North Sea
would not be considered an environmentally sensitive area when compared to
others (National Parks, Arctic, etc.).
4.4. Comparison between DM supported field development and

conventional reservoir modelling
Given the clear successes of utilising DM in the field development it is prudent to
compare this to simply conducting a field development without this innovative
technique. This is done by comparing the recommended option (Case 2) with the
base case option. The results are very clear with the exact same amount of CAPEX
and OPEX an additional 1.2% of recovery is achieved. This means that using
conventional reservoir modelling 1.42x106 Sm3 of oil is produced, 180,000 Sm3
less than using DM. This reduces profits by approximately $80,000,000. This
clearly indicates the advantages of DM provided the system developed is at a cost
of less than this figure.
5. Conclusions
This study was a brownfield development of Volve field, a Norwegian oil field.
The data pack was huge and with this considerable size investigating manually
would be unmanageable. Thus, a bespoke DM algorithm was planned and applied.
This data was analysed and used to create the reservoir model, in tandem perhaps
the most effective development scheme was fashioned, and an inclusive strategy
was set out.
The DM successfully reduced the data set to 0.2% of the initial size extracting
only the data instrumental for the study. The 11 steps development plan
investigated the previous development and the reservoir, the successful completion
of which permitted the scheme to be advanced. This culminated in the enhancement
and simulation analysis which presented each of the cases investigated in this study.
The results identified the best scenarios based on 7 options. Resulting in the
best technical case using 2 water injectors, 5 WAG injectors, and 10 oil producers
(3 of which were horizontal). Giving a recovery factor of 56% in 3 years with 58%
if extended for a further year. Due to environmental considerations and impact on

the seabed the recommended choice was the Case 2 utilizing 6 producers and 1
water injector. This gave a recovery of 53.2% over 3 years at which stage the wells
will be shut in.
A jack-up rig with processing facilities is planned for drilling and fluid
processing with a shuttle for export. The project is estimated to cost $657 million
and make $750 million netting a profit of $93 million. The unique selling point of
this case was its minimal footprint on the local environment as well as a lower
carbon and capital intensity. After 3 years the NPV is $93 million, IRR is 57%,
UTC is $52.8/bbl. and payback is 2 and a half years from first oil.
Nomenc
latures
Bo Oil formation volume factor ----

C Concentration kg/m3
(γo) Oil gravity ----
(γg) Gas gravity ----
R Gas-oil ratio scf/stb
J Connections to wells bbl/d/psi
K Permeability mD
Kv Vertical permeability mD
Kh Horizontal permeability mD
Kr Relative permeability ----
L, H, W Length, height, width m
Hp Hydrocarbon pore volume m3
V Volume m3
VStandard Volume at standard Sm3
Conditions
N Original oil in place m3
P Pressure Pa
Q Fluid flow rate m3/s
r Radius m
Rf Recovery factor %
s Saturation ----
S Skin factor ----
Swi Initial water saturation ----
T Transmissibility mD-ft./cp
t Temperature K
TVD True vertical depth m
Greek Symbols
𝜇 Viscosity
𝜑 Porosity
Abbreviations
CAPEX Capital Expenditure
CPU Central Processing Unit
DM Data Mining

EIA Energy Information Administration

ED Microscopic Sweep Efficiency
EOR Enhanced Oil Recovery
Ev Macroscopic Sweep Efficiency
FDP Field Development Plan
FIPNUM Fluid In Place Number
GOR Gas Oil Ratio
HCPV Hydrocarbon Pore Volume
IRR Internal Rate of Return
KDD Knowledge Discovery in Databases
MBAL Mass Balance
Mo Moveable Oil
NOK Norwegian Krone
NPV Net Present Value
OIIP Oil Initially in Place
OPEX Operational Expenditure
PVT Pressure Volume Temperature
STB Stock Tank Barrels
STOIIP Stock Tank Oil Initially in Place
TVD True Vertical Depth
USD United States Dollars
UTC Unit Technical Cost
WAG Water Alternating Gas
References
1. Gomez, Y.; Khazaeni, Y.; Mohaghegh, S.D.; and Gaskari, R. (2009). Top-
down intelligent reservoir modelling (TDIRM). SPE Annual Technical
Conference and Exhibition, New Orleans, Louisiana, SPE-124204-MS.
2. EIA. (2020) Petroleum and other fluids prices. US Energy Information
Administration. Retrieved May 4, 2020, from https://www.eia.gov/
3. Antariksa, Z.N.; Ultra, D.; Malau, A.; and Sukotrihadiyono, T. (2014).
Reviving brown field oil potential: Case study of multilayer reservoir.
International Petroleum Technology Conference. Kuala Lumpur, Malaysia,
32-33.
4. Al-Khaledi, S.M.; and Qadri, F.W. (2010). Execution of brownfield projects.
Abu Dhabi International Petroleum Exhibition and Conference. Abu Dhabi,
UAE, 22-23.
5. Kumar, N.; Mukherjee, D.; Paul, R.; Kamal, F.R. and Takieddine, O. (2020).
Challenges in execution of brownfield offshore oil and gas projects. Abu Dhabi
International Petroleum Exhibition and Conference. Abu Dhabi, UAE, 9-12
6. Lee, S.-J. and Siau, K. (2001). A review of data mining techniques. Industrial
Management and Data Systems, 101(1-2), 41-46.
7. Balaji, K.; Rabiei, M.; Suicmez, V.; Canbaz, C. H.; Agharzeyva, Z.; Tek, S.;
Bulut, U. and Temizel, C. (2018). Status of data-driven methods and their
applications in oil and gas industry. SPE Europe 80th EAGE Conference and
Exhibition. Copenhagen, Denmark 11-14.
8. Benjamini, Y. and Leshno, M. (2005) Statistical methods for data mining. Data
Mining and Knowledge Discovery Handbook (4th ed). Boston: Springer.

9. Li, H.X. and Xu, L.D. (2001). Feature space theory - A mathematical
foundation for data mining. Knowledge-Based Systems, 14 (5-6), 253-257.
10. Azevedo, A. and Santos, M.F. (2000). KDD, SEMMA and CRISP-DM: A
parallel overview. IADIS European Conference Data Mining. Amsterdam, The
Netherlands, 24-26
11. Mukhopadhyay, A.; Maulik, M.A.; Bandyopadhyay, S. and Coello, C.A.C.
(2014). A survey of multiobjective evolutionary algorithms for data mining:
Part I. IEEE Transactions on Evolutionary Computation, 18(1), 4-19.
12. Sahu, H.; Shrma, S. and Gondhalakar, S. (2010). A brief overview on data
mining survey. International Journal of Computer Technology and Electronics
Engineering (IJCTEE), 1(3), 114-128
13. Windarto, A.P.; Dewi, L.S. and Hartama, D. (2017). Implementation of
artificial intelligence in predicting the value of Indonesian oil and gas exports
with BP algorithm. International Journal of Recent Trends in Engineering and
Research, 3(10), 2455-1457.
14. Equinor (2018). Volve field data set. Retrieved May 4, 2020, from:
https://www.equinor.com
15. Akanji, L.; and Sandrea, R. (2016). A neuro-fuzzy approach to screening
reservoir candidates for EOR. Advances in Petroleum Exploration and
Development, 12(1), 1-14.
16. Ramos, G.A.R. and Akanji, L. (2017) Data Analysis and neuro-fuzzy
technique for EOR screening: application in Angolan oilfields. Energies,
10(7), 837.
17. Free currency rates (2020). Online currency converter. Retrieved May 10,
2020, from: https://freecurrencyrates.com/en

17 4 22

Uploaded by

Copyright:

Available Formats

17 4 22

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

17 4 22

Uploaded by

Copyright:

Available Formats

Journal of Engineering Science and Technology

Vol. 17, No. 4 (2022) 2567 - 2593

PLANNING AND IMPLEMENTING THE

ALEXANDER E. MCALLISTER*, LATEEF AKANJI

School of Engineering, University of Aberdeen,

2. Methods and Approach

Journal of Engineering Science and Technology August 2022, Vol. 17(4)

Fig. 1. Volve field development methodology

2.1. Background study

Journal of Engineering Science and Technology August 2022, Vol. 17(4)

Journal of Engineering Science and Technology August 2022, Vol. 17(4)

• Automatic decision making: real time data analysing permits organisations to

Journal of Engineering Science and Technology August 2022, Vol. 17(4)

2.2. Data mining

2.2.1. Model validation

Journal of Engineering Science and Technology August 2022, Vol. 17(4)

Fig. 2. VOF-DM algorithm decision tree.

2.2.3. Efficacy of data mining

Journal of Engineering Science and Technology August 2022, Vol. 17(4)

2.3. Reservoir modelling

Fig. 3. Well location in reservoir model

2.4. Development plan investigation

2.4.1. Identifying existing producing wells and injector wells

Journal of Engineering Science and Technology August 2022, Vol. 17(4)

Table 1. Table of Volve development wells [14].

2.4.2. Recovery factor from primary production

Table 2. MBAL values [14].

Journal of Engineering Science and Technology August 2022, Vol. 17(4)

(a) Graph of Krw and Krow against Sw.

Journal of Engineering Science and Technology August 2022, Vol. 17(4)

2.4.5. Aquifer strength and strength of the gas cap

2.4.6. Characterisation of the fault pattern

2.4.7. Completion strategy plan

Table 3. Table of structures WOC, GOC [14].

Table 4. 11 Volve regions and total field by OIIP,

Journal of Engineering Science and Technology August 2022, Vol. 17(4)

2.5. Development scheme

Fig. 5. Logic flow diagram of development scheme

Journal of Engineering Science and Technology August 2022, Vol. 17(4)

The first phase is determining the years of production, number of producer

2.5.1. Years of production

2.5.2. Number of producer wells

2.5.3. Injectors and EOR

Journal of Engineering Science and Technology August 2022, Vol. 17(4)

technique to a field, variables are considered based on degree of variance. Some

2.5.4. Regions of interest

Fig. 6. Volve field by FIPNUM region.

2.5.5. General well position and spot pattern

Journal of Engineering Science and Technology August 2022, Vol. 17(4)

Fig. 7. Oil Saturation across the field.

2.5.6. Specific well locations and case development

Table 6. Case number, name and description.

Journal of Engineering Science and Technology August 2022, Vol. 17(4)

Fig. 8. Logic flow diagram of well location and case development.

2.5.7. Summary of cases

Journal of Engineering Science and Technology August 2022, Vol. 17(4)

Table 7. Summary of redevelopment cases.

2.5.8. Final case

Table 8. Development wells for Case 7.