You are on page 1of 40

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/301453338

Best Practice and Lessons Learned for the Development and Calibration of
Integrated Production Models for the Cooper Basin, Australia

Conference Paper · October 2015


DOI: 10.2118/176131-MS

CITATIONS READS

5 9,400

6 authors, including:

Yang Fei Manouchehr Haghighi


University of Adelaide University of Adelaide
9 PUBLICATIONS   147 CITATIONS    130 PUBLICATIONS   1,512 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Reservoir characterization View project

Rock-Fluid Mechanical interactions View project

All content following this page was uploaded by Yang Fei on 05 July 2016.

The user has requested enhancement of the downloaded file.


SPE-176131-MS

Best Practice and Lessons Learned for the Development and Calibration of
Integrated Production Models for the Cooper Basin, Australia
A. Mantopoulos, D. A. Marques, and S. P. Hunt, Santos Ltd.; S. Ng, Y. Fei, and M. Haghighi, Australian School
of Petroleum - The University of Adelaide

Copyright 2015, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE/IATMI Asia Pacific Oil & Gas Conference and Exhibition held in Nusa Dua, Bali, Indonesia, 20 –22 October 2015.

This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents
of the paper have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect
any position of the Society of Petroleum Engineers, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written
consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may
not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract
This paper investigates the best practices and lessons learned for both the development and calibration
methods of integrated production models. Consequently, this allowed for the successful application of
Integrated Production Modeling (IPM) to large scale and complex petroleum production systems in the
Cooper Basin, Australia. The potential for calibrated integrated production models to aid the assessment
of production system development, forecasting, operational surveillance and optimization opportunities
(flow assurance, integrity considerations, well performance/intervention, incremental approaches to
development and back out/pressures) supported the need for an optimum (quality versus resource
allocated) and innovative development procedure. A structured workflow for developing and calibrating
the integrated production models is presented which allows for an improved ability to create and operate
such models. The key development components, and for which illustrative and detailed workflows are
presented, are: (1) development of reservoir model, (2) development of required well models, (3)
development of required surface facility model and (4) the overall calibration of the integrated production
model. Detailed within this paper is the extensive theoretical analysis which was mandatory in determin-
ing the use of the most suitable parameters which included vertical lift performance (VLP) correlations,
flowline correlations, relative permeability curves (generation and alternatives), inflow performance
relationship (IPR), water influx models, micro-string implications and all related considerations and
assumptions. The paper emphasizes the trade-off between the required outputs from the application of the
integrated production models and the available resources and period of development. The value of
information is also highlighted through the importance of data collection and preparation procedures, both
technical (pressure and fluid) and organizational (procedures). Furthermore, engineering decisions were
found to significantly dictate the accuracy and efficiency of the implemented development process and
workflows, in particular related to average reservoir properties, cyclic well behavior and the history
matching of system pressures and flow rates within the calibration process. The considerations have been
supported by relevant examples throughout the paper.
2 SPE-176131-MS

Introduction
Petroleum production systems can be complex and entail a large quantity of uncertainties. Therefore, it
is of primary importance to possess the ability to determine and forecast the production of any production
system which can be severely affected by uncertainties. Such uncertainties transcend from the complexity
involved in combining multiple petroleum system disciplines as well as business and operational aspects.
More specifically, integrated contributions from engineers, geoscientists and operators are required in
order to effectively understand subsurface and surface components. Business and operational decisions -
the consideration of the economic production limits, operational constraints and overall company
production values or attitudes - are also pivotal. Network constraints or bottlenecks must be quickly
identified and systematically understood in order to be mitigated and allow production targets to be met.
Production can solely control project cash flow and be the primary reason a project is approved or
abandoned. When considering gas production projects, additional complications arise from needing to
honor long term contract requirements (Amudo et al. 2011). As a result, it is of upmost importance for
companies to accurately forecast future production in order to ensure that they can satisfy their contractual
agreements and achieve all economic targets.
IPM provides an effective multi-disciplinary understanding of entire production systems. The reser-
voirs, wells (subsurface) and surface facilities form the components of an integrated production model. All
components are investigated through the integrated analysis provided by IPM, leading to effective system
or project development, production forecasting, surveillance and optimization of production networks
(Aditama et al. 2010; Shrestha et al. 2008; Walrick van Zandvoord et al. 2007). This paper demonstrates
the best practice and lessons learned for the development and calibration of integrated production models
through the extensive applications towards multiple gas and oil production satellites (approximately 1,000
wells and 150 fields) in the Cooper Basin. These satellites were very mature and ranged from 20 to 50
years old and their characteristics varied widely. Due to the age of the basin, data constraints were a
common occurrence for these satellites. Furthermore, the potential for calibrated integrated production
models to assist the assessment of production systems supported the need for an optimum (quality and
resources allocated compared to development period) and innovative development procedure. The
integrated production model development components, for which illustrative workflows are presented,
include:
● Model input data collection and preparation,
● Development of reservoir models,
● Development of required well models,
● Development of required surface facility model, and
● Calibration of the integrated production model.
Also detailed is the extensive theoretical analysis deemed mandatory in determining the use of the most
suitable correlations and applying the most appropriate considerations and assumptions for input param-
eters. Additionally, the assessment of input data quality, constraints, collection and preparation proce-
dures, both technical (pressure and fluid) and organizational (procedures), and their importance are also
outlined for each of the development components. Furthermore, a brief insight is provided into the lessons
learned when developing the models when applying optimization and forecasting scenarios towards the
IPM applications.

Model Background
The developed integrated production models, through which best practice and lessons were learnt, varied
widely in characteristics and features. The models were developed for a particular production system or
satellite within the Cooper Basin and varied as the production systems (satellite) were not solely oil or gas
SPE-176131-MS 3

- rather, they were a combination and, in one case, co-mingling production of oil, gas and gas not
associated to oil. The number of wells within a satellite ranged between 15 to 230, while the producing
formations varied from one to five as the majority of wells in the Cooper Basin had co-mingled
completions. The figure presented in Appendix A outlines the stratigraphy of the Cooper Basin comprised
predominately of the Gidgealpa and Nappamerri groups. The production of the developed integrated
production models originates from the Gidgealpa Group, predominately from the Toolachee, Daralingie,
Epsilon, Patchawarra and Tirrawarra formations. All formations are dedicated gas producers, apart from
the Tirrawarra formation, which also accounted for the oil production of one satellite. Additionally, the
general production was from a low permeability geological environment.

Software
Petroleum Experts’ (PETEXs) PVTP, MBAL, PROSPER and GAP software were used to create the
reservoir, well and surface facility models.

PVTP
The use of PVTP (Pressure, Volume and Temperature Package) allowed comprehensive understanding of
the PVT properties and characteristics. It was possible to match simulated results to laboratory data which
allowed the simulation of several different processes related to both reservoir and production components
of any production system. Furthermore, pressure and temperature parameters and fluid properties were
also able to be matched (PVTP PE 2009).

MBAL
Accurate knowledge of reservoir behavior underpinned efficient development and production forecasting,
especially in complex production systems. The material balance software (MBAL) provided the ability to
define hydrocarbon volumes and drive mechanisms which allowed the generation of reliable production
forecasts through simulation (MBAL PE 2009). The theoretical material balance feature of MBAL was
utilized to model the reservoir component of all production systems (MBAL PE 2009) and is based on the
‘Law of Conservati on of Mass’ as described by Equation 1 (Dake 1978).
(Equation 1)

MBAL adopts a theoretical modeling approach. As the material balance equation is zero dimensional,
the reservoir model is based on a tank model. MBAL does not take into account the geometry of the
reservoir, position and orientation of the wells (MBAL PE 2009). While considered to be important
aspects, these factors are far outweighed by MBAL’s efficiency.

PROSPER
The use of PROSPER (Production and Systems Performance Analysis Software) allows for efficient
modeling of the well component of production systems (Correa Feria 2010; Ozdogan et al. 2008; Shrestha
et al. 2008). PROSPER enables the creation of well models which form the link between subsurface and
surface production system components. As PROSPER utilizes nodal analysis, well performance, design
and optimization were all possible through the software. Well characteristics are accounted for by fluid
characterization (PVT), tubing pressure losses (Vertical Lift Performance (VLP)) and reservoir inflow
(Inflow Performance Relationship (IPR)). PROSPER provides well performance curves by analyzing the
VLP, which account for pressure losses, and the IPR, which represents the reservoir inflow (PROSPER
PE 2009). A summation of production rates, downhole completions, reservoir properties and stimulation
processes assists PROSPER in providing accurate well performance curves which can also be matched to
actual performance data.
4 SPE-176131-MS

GAP
General Allocation Package (GAP) uses non-linear Sequential Quadratic Programming to simulate
multiphase flow, allowing the optimization of production systems. GAP enables the identification of the
optimum control settings that maximize objectives function (for example, production or revenue) (GAP
PE 2009). The IPM software package enables GAP to be linked to reservoir and well models provided by
MBAL and PROSPER respectively. This allows the replication of the entire production and injection
systems (GAP PE 2009). Overall, GAP provides the ability for users to conduct nodal analysis of any
point within a production system.
Integrated Production Model Development Workflow
The workflow used for the development of the Cooper Basin’s integrated production models is outlined
in Fig. 1. The order in which each model was developed remained the same for all production systems and
was typically a front end loaded process. The preliminary development stage, which is dependent on input
data, is also presented. The innovative workflows for each component, along with the considerations and
assumptions which underpinned the best practice and lessons learned, are elaborated upon further in the
paper.

Figure 1—Integrated production model development workflow

Model Input Data


The importance of input data collection and preparation was a critical component of IPM and was noted
by several authors to define the quality (robustness and efficiency) of the integrated production models
(Amudo et al. 2011; Nadar et al. 2006; Shrestha et al. 2006). Perhaps the most noted influence of input
data integrity was the ability to develop an integrated production model whose output was robust and
possessed the upmost engineering confidence. Therefore, the need for thorough and critical evaluation of
input data was required. Prior to the data considerations presented in this paper, Arnondin et al. (1996)
had identified the two data disciplines - technical (pressure and fluid) and organizational (procedures) -
and how they were both equally important. Through the extensive application of IPM to Cooper Basin
production systems, it was evident that such evaluation applied more to the technical input data, although
critical evaluation required efficient organizational procedures in order to be effective. The impact of good
inter-disciplinary communication was most crucial when assessing input data, in particular with complex
reservoirs.
Technical
The technical data required for model development varied for each production system; however, the
fundamental data required remained constant:
● Geological data and structural/topographic maps,
● Established and modernized geological models,
● Core data (Mercury-Injection-Capillary-Pressure, rock compressibility and relative permeability
data),
● Fluid PVT test data on a reservoir (formation) and well scale,
● Both static and flowing pressure and temperature data for reservoirs, wells and flowlines (initial
reservoir, reservoir, shut-in wellhead (WHP), bottom hole (BHP) and compression inlet). Some
SPE-176131-MS 5

examples of test data sources include Flowing Gradient Survey (FGS), Pressure Build Up Test
(PBU), Memory Production Log Test (MPLT) and Static Gradient Survey,
● Gas Analysis (GA) and Water Analysis reports,
● Liquid Evaluation Tests (LET): Gas-Oil, Water-Gas and Condensate-Gas ratios (GOR, WGR and
CGR),
● Well production history (required for all phases: oil, gas and water),
● Deviation surveys and well details (location, name) from well completion reports,
● Downhole equipment (tubing, casing) and completion (perforations) schematics,
● Compressor performance details (performance curves),
● Surface network characteristics (flow line and pipeline specifications and elevations), and
● Well downtime.
Additionally, the Santos Engineering Fluid Flow Calculation Standards were used for all fluid flow
calculation considerations. Pipe roughness, scaling, friction factors, temperatures and heat transfer
co-efficient values for the Cooper Basin region were included.
Organizational
The organizational aspect of input data was reflected by well surveillance procedures and the organization/
preparation of collected data. The analysis of all technical data, when coupled with organizational
procedures, had the ability to evaluate and improve surveillance procedures. In essence, technical data
evaluation required effective organization to avoid complicating the development process. Therefore,
directories and template names, presented in Appendix B, were used to store the required data. These
directories were compiled for specific software and processes. The table also displays the files stored in
each folder and the input data found in each file.
Data Quality and Availability
The quality and availability of data was found to directly affect and govern the development of integrated
production models. Data quality and availability dictated the accuracy of the developed model and
efficiency of the development process and workflows. It was, therefore, critical to understand the
consequences of each development component, the significance of data related engineering decisions
throughout the development process and their overall effect on model accuracy. It was expected and
encountered that once quality checking between simulated compared to historical data had been per-
formed (for example, variance in WHP or wellhead rates (WHR)), any limitations within any systematic
procedure would become evident. Such limitations invoked the implementation of engineering decisions,
where possible, prior to such occurrences. The considerations of what was to be required from the
application of each integrated production model were, therefore, the controlling factor(s). Rigorous
assessment of data made it possible to make the correct engineering decisions. Examples of such decisions
included which IPR models to use and which source of data was the most reliable which, when analyzed
correctly, would depend on how, why and when the data was collected. Additionally, assessment of
individual parameters unique to the reservoir models, well models and surface model would occur at later
stages of the development procedure. This was a major challenge and when considering the complexity
of the Cooper Basin, where several satellites and production systems made up of thousands of wells, had
data constraints on a well scale but there were large amounts of data that were assessed. Wherever
possible, data was collected uniquely for single reservoirs or formations, although, this was heavily reliant
on the production volumes being allocated to individual formations.
Development of Reservoir Model (MBAL)
MBAL Workflow
It was necessary to establish the extent of reservoir compartmentalization in order to allocate wells to
reservoirs and effectively represent this in the reservoir models. Determining the extent of transmissibility
6 SPE-176131-MS

between reservoirs or compartments was achieved by constructing pressure transverse graphs (reservoir
pressure versus date graphs). The transmissibility took into consideration production start date, online/
offline periods, rate of depletion, volumes displaced, drainage radius, formation properties and co-mingled
wells. For co-mingled wells, reservoir compartmentalization could only be analyzed for a formation if
production could be allocated to each formation (MPLT). PVT data was created as required. Three
scenarios were encountered: (1) multiple wells in a single tank, (2) transmissibility between both single
wells and multiple wells in a single tank and (3) one well per tank. The PVT data (gas gravity, separator
pressure, CGR, condensate gravity, water salinity and H2S, CO2 and N2 mole percentages) were not
influenced for scenarios 2 and 3; however, for scenario 1, the weighted averages were used in relation to
permeability, sand thickness and production volumes. Monthly cumulative oil, gas and water volumes
were deemed mandatory for all developed models. The use of relative permeability curves, which were
representative of the developed tank models, was found to be the most important factor in predicting
accurate phase volumes. Simulation and history matching focused on matching reservoir pressure decline
over time with cumulative production to a ⫾ 2 to 5% tolerance range. History matching produced gas and
water volumes were performed to assess the likelihood of gas cap, tight gas and aquifer support. A detailed
MBAL workflow is depicted in Appendix C.
Considerations and Assumptions
Gas Viscosity The application of IPM throughout the Cooper Basin required the use of correlations to
calculate the oil and gas viscosities. For reservoirs where PVT data was available, PVT matching was
performed. Therefore, the most suitable gas and oil viscosity correlation was analyzed. In cases where it
was not measured, the correlation developed by Lee, Gonzalez and Eakin was assumed to be best suited
to calculating the gas viscosity. The Lee, Gonzalez and Eakin correlation is a semi-empirical method that
provided accurate gas viscosity estimates for natural gases (where CO2 is between 0.9 and 3.2 mol%),
while also accounting for contaminants (NO2, CO2 and H2S). The Carr, Kobayashi and Burrows
correlation was suitable for contaminant concentrations of up 15 mol% (Jeje et al. 2006). For oil
reservoirs, the Standing-Katz model estimated the oil formation volume factor while the correlation
presented by Beggs and Robinson estimated oil viscosity. These two correlations best matched PVT data
and, consequently, it was assumed that they were suitable for the Permian oil being produced (Craft et al.
1991).
Water Salinity The water salinity had an impact on the density of the water which affects the com-
pressibility of the formation water (MBAL PE 2009). Ideally, the water salinity parameter was obtained
from PVT reports. However, if the MBAL model had been developed without PVT data, the software
utilized a single deterministic value of water salinity. Where PVT reports were not available, the water
salinity was obtained from a developed correlation between the concentration of NaCl (ppm) and the
resistivity of formation water during the Jurassic and Permian geological time periods. This correlation
had been developed with the use of several hundred formation water samples and was found to accurately
describe water salinity in the relevant Cooper Basin formations. This was crucial where aquifer support
and large volumes of water production were evident as it directly affected the reservoir pressure history
matching procedure.
Reservoir Compartment Analysis: Drainage Radius When undertaking reservoir compartment analy-
sis, it was important to consider all factors which influenced wells producing from the same compartment.
As MBAL did not take into account the geometry of the reservoir, position and orientation of the wells,
the drainage radius parameter was of particular importance as it made it possible to account for no-flow
boundaries. More specifically, it was important to represent this correctly in MBAL as the extent of the
drainage area and the location of the well had a large influence in the performance of the reservoir
(Noaman 1982). If the radius was predicted accurately using either theoretical formulae or software
SPE-176131-MS 7

packages, it would assist the reservoir compartment analysis when analysed in regards to well location and
structure characteristics (lateral extent and continuity of sand layers). The drainage radius of a well is
defined as the distance from the wellbore at which a pressure drop (pressure transient) can be observed.
However, as outlined in Equation 2 (Lee 1982), the drainage radius depends only upon reservoir properties
(k, t, ␮, ␸ and ct) and not pressure. Hence, the drainage radius, used in conjunction with stratigraphic and
topographic maps of reservoirs, could identify compartments. An additional consideration is that the
drainage radius of any well changes with time and this could only be accounted for through ongoing
compartment analysis as new data became available. Appendix D depicts an example of how the
calculation of the drainage radius (for a given point in time) was able to assist in the identification of
compartments within a field. By calculating the distance that the drainage radius extends and plotting this
distance on a topographic map, it provided an indication that there was possible pressure communication
between the two wells due to both radii intersecting with each other.
(Equation 2)

However, a lesson learned through the construction of the integrated production model is that caution
should be exercised when more than one well is allocated into a single compartment (modelled as tanks
within MBAL). As mentioned previously, since Equation 1 is zero dimensional, the reservoir model is
assumed to behave as a tank. This is due to the rapid transmission of pressure changes throughout the
system which allows the tank model to be zero dimensional (Amudo et al. 2011). Dake (1994) cites that
the hydraulic diffusivity constant, k/␸␮ct, influences the speed in which the pressure changes are
propagated throughout a reservoir in order for pressure equilibrium to be reached. However, the software
assumes that the properties of each tank, for example reservoir pressure, permeability, viscosity, are
constant irrespective of the number of wells within the tanks. Realistically, this almost certainly is not
correct due to the heterogeneous nature of reservoirs. Consequently, the hydraulic diffusivity constant
should be unique for each well. Therefore, it is preferable to have single wells represented by single tanks
with transmissibility, as opposed to combining multiple wells into a single tank (reservoir pressure would
not be identical for well).
Average Reservoir Properties One of the most important considerations is the method in which pro-
ducing formations have been represented in the developed reservoir models. Each producing formation
was made up of several producing sand layers and these sands were coupled and modeled as a single
reservoir or tank within MBAL. As each sand layer was unique, the reservoir properties varied for each
layer and, hence, the developed tanks required average reservoir properties in order to be representative
of all sand layers. More specifically, the permeability, porosity and connate water saturation for each sand
layer were weighted by the thickness of each layer. The properties for each layer were obtained from
various sources including core analysis and well logs. Additionally, the classification of net thickness
versus pay thickness was also taken into consideration for any sand layer which had been fracture
stimulated. It is well understood that this technique would not take into account when a single sand had
reached the end of its production life but that this average representation would minimise the effect. In
the case that there was no production split for co-mingled wells, two or more reservoirs would be grouped
together; hence, average formation properties were utilized to obtain average values across more than a
single reservoir. The major disadvantage of such scenarios was when one highly and poorly producing
reservoirs were coupled, the reservoir pressure would not accurately reflect the production behavior of
each reservoir such as end of life or water breakthrough (wellbore behavior compared to reservoir
behavior).
The misrepresentation of sands contributing to flow within a multi-layered well was a common issue.
More specifically, as the MPLTs used to split WHR and, therefore, produced gas volumes were dated, it
8 SPE-176131-MS

was possible to allocate production to formations which actually had no producing sands. This resulted
in a misrepresentation of reservoir and bottom hole pressures which resulted in incorrect WHR as the
pressure drawdown relationship was not represented correctly. In order to have accurate production splits
for wells that are producing from multiple sands, it is necessary to have at least two to three MPLTs during
the entire well’s life - preferably one at the start, middle and present time. This is because sands may be
of high permeability (produce and die off quickly) or low permeability (produce for longer), while other
reservoir or sand layer characteristics may vary quite significantly causing varying effects. Therefore,
previous MPLTs may not be representative of the current production split and, while the well model could
still be matched, future problems could and were faced within GAP. In such situations, these wells were
only built as single layered models in order for a successful history match. This was as a direct result of
misinterpreting which sand was actually still producing and highlights the importance of valid and current
MPLTs. An example which demonstrates this concept is depicted in Fig. 2 with the well initially being
modeled as multi-layered. The most recent MPLT, which had been conducted a few years ago, had
indicated that one of the two sands was still producing at a small production split (30%). This split was
used and, although it was able to obtain a successful history match within PROSPER, the well was not
able to be matched within GAP. However, by analyzing older MPLTs, a steady decline in the production
split for the sand was discovered. Hence, in conjunction with the declining spilt, it was inferred that the
capacity of sand at the current time would be low and, therefore, the sand itself could not be modeled and
correctly represented in a multi-layered well. Consequently, only the second sand (Sand 2) was modeled
when the contribution was extremely low.

Figure 2—Demonstration of a well’s production split over time

Material Balance Equation: Input Parameter Importance It is well understood how MBAL utilizes
the material balance equation to analyze gas and oil reservoirs. However, it is also critical to understand
the influence, sensitivity and importance of each parameter in regards to its ability to accurately describe
tank pressure depletion. Equation 3 represents the general material balance equation for a hydrocarbon
reservoir (Dake 1978). This form of the material balance equation accounts for the expansion of oil and
the originally dissolved gas (liquid expansion, liberated gas expansion), expansion of the gas-cap gas,
changes in hydrocarbon pore volume due to connate water expansion and pore volume reduction, and the
surface production (Np) due to differential pressure (⌬p), is best described as underground withdrawal.
SPE-176131-MS 9

(Equation 3)

The lessons learned relate directly to possessing representative values of these input parameters, in
particular, PVT data, oil, gas and water cumulative production volumes which were representative of the
reservoir’s production life. It was crucial that the PVT data and produced volumes accurately described
the relationship between the surface production and the subsurface physical process which had occurred.
If not, then the model output was poor and reservoir pressure depletion could not be accurately modeled
(for example, erratic pressure behavior was observed per monthly production history input data period).
Therefore, it was essential to consider the changes in hydrocarbon composition over time by performing
several PVT tests, particularly when working with retrograde gas and volatile oil. The influence of
Enhanced Oil Recovery (EOR) schemes on recorded production volumes was found to be the primary
cause of inaccuracy, primarily where miscible oil recovery, such as ethane injection, had been performed.
More specifically, miscible EOR altered the composition of the hydrocarbon being produced and had
resulted in the Bg not aligning with the volumes of oil and gas produced. As described in ‘Average
Reservoir Properties’, it was crucial that each parameter represented the method in which each sand layer
had been modeled in MBAL.
Relative Permeability Curves The theory behind relative permeability is comprehensively understood
within the petroleum industry along with the importance of how this phenomenon affects reservoir
behavior. Accurate reservoir analysis and representation required the full understanding of how oil, gas
and water behaved when flowing simultaneously through porous medium. Hence, accurate relative
permeability curves were required for reservoir model development. For all developed Cooper Basin
models, there was limited relative permeability curve data available. This caused large difficulties in how
reservoir models were developed and, most importantly, it affected the entire integrated production model
development procedure and model accuracy. The ability which relative permeability curves possessed in
predicting the changes in phase saturations over time governed the ability of any completed integrated
production model to do the same. More specifically, it was possible for an integrated production model
to not only describe pressure behavior (⌬p - heavily affected liquid volumes) throughout any production
system, but to also forecast the changes in GOR, WGR, CGR and water cut over time correctly.
Furthermore, as relative permeability curves were dependant on rock and fluid properties, they were
unique for each formation and sand layer. As the reservoir models were developed per formation with
averaged sand properties, the relative permeability curves would represent average saturations, and
minimum and maximum permeabilities (crucial as these permeabilities are a function of the fluid
saturations).
Therefore, a methodology was developed to generate relative permeability curves from available core
analysis, log and well test data. This methodology was only applied to gas reservoirs (gas and water flow)
as the complexity of multiphase flow in oil reservoirs was too high. It was first considered plausible and
then proven that the combination of Mercury-Injection-Capillary-Pressure (MICP) data, permeabilities
and phase saturations derived from wireline logs and liquid ratios from LETs could be combined
successfully. While log and LET data was available for all wells, MCIP data was rare and, therefore,
limited the development of relative permeability curves to a single reservoir in any one field. Appendix
E outlines an example of the numerical values used in successfully developing a set of relative
permeability curves in MBAL. The MICP data provided the maximum (Sgmax) and residual gas (Sgr) and
water saturations (end point saturations). This was possible as the MICP data had been converted from an
air-mercury system (laboratory conditions) to brine-hydrocarbon (reservoir conditions) system and
accounted for the height above the free water level. The capillary pressure entry and maximum pressures
identified the Sgmax and Sgr hydrocarbon saturations. The log analysis provided the krgmax and krwmax
values from the sand layers which had the highest gas and water saturations. The end points were then
10 SPE-176131-MS

fixed and the gas and water exponents were varied until the WGR result from MBAL matched the WGR
for a robust LET. For the example presented in the appendix, the MBAL WGR was matched to the LET
value of 26.54 bbl/Mscf.
A consequence of the absence of relative permeability data was the use of the fractional flow curve
input in MBAL. The fractional flow equation assumed diffuse flow conditions, where at any point in a
linear displacement path, water saturations were uniformly distributed with respect to thickness. There-
fore, as the water saturation was assumed to be uniformly distributed, the same was considered for
hydrocarbon saturations which allowed for the flow to be modeled by thickness averaged permeabilities
(Dake 1978). The assumptions behind the fractional flow equation affected the true representation of
reservoir behavior. However, the use of single deterministic values of WGRs obtained at different stages
of the wells or reservoir’s production life from LETs further affected the accuracy. This provided a
deterministic forecast of WGRs until the most recent production history which could be used as a fixed
value later in the development procedure in GAP. Alternatively, fractional flow matching was also
performed in MBAL using LET results to generate pseudo relative permeability curves. Neither method
was found to be as accurate as the MICP and log method but because of the resources and time required,
depending on the case, it was suitable to apply fractional flow curve analysis.
Another important lesson learned was the use of different relative permeability curves for different
reservoirs. Consider Fig. 3 where A, B, C, D, E and F represent six different perforations from three
different closely located wells that are producing from two reservoirs. Ideally, to get the most accurate
predictions, it would be desirable to have six individual relative permeability curves for each of the
perforations. However, since a static model was to be built and obtaining this information was impractical,
two relative permeability curves were used to represent perforations A, B, C and D, E, F separately. This
was considered the best practice as it took into account the heterogeneity of reservoirs (different sands)
and the well locations.

Figure 3—Representing heterogeneity via relative permeability curves

History Matching Reservoir Pressure in MBAL Using the PVT and tank inputs, reservoir model sim-
ulation was performed with the aim of history matching reservoir pressure history to simulated results.
Tank models which were history matched correctly would provide an accurate representation of reservoir
pressure depletion to date and in the future. While the theory and methodology behind this process was
trivial, the process of matching simulated reservoir pressures should only be performed against historical
reservoir pressure data which represented current reservoir conditions. For Cooper Basin reservoir
models, reservoir pressure data points were only used for history matching if they had been obtained after
any well/reservoir intervention (re-perforations, additional perforations, acidizing, fracture stimulations
and re-completions). However, well test data prior to such interventions could still be used to match
previous pressure trends. The example illustrated in Fig. 4 demonstrates the effect of history matching to
all and only post-intervention reservoir pressure points. Consequently, by matching to all reservoir
SPE-176131-MS 11

pressure points, MBAL will simulate a pressure decline of best fit through all points which results in an
incorrect estimation of the pressure compared to only matching to post-intervention points.

Figure 4 —Reservoir pressure history matching results in MBAL

Water Influx Modeling Water influx modeling was conducted as a result of evidence which suggested
there was an external influence on the reservoir pressure - for example, an additional reservoir pressure
trend or where reservoir pressure support was evident in the mid to late life of the reservoir. Several fields
throughout the Cooper Basin were found to have significant aquifer influence and the Hurst-van
Everdingen-Dake water influx model was discovered to provide the best reservoir pressure and water
volume history match results. The water influx is modeled as a function of a given pressure drop at the
inner boundary of a reservoir system (Dake 1978). The Hurst-van Everdingen-Dake model utilizes various
technical parameters such as permeability, reservoir radius, outer/inner radius and encroachment angle to
calculate time and aquifer constants directly. The water influx is described by Equation 4.
(Equation 4)

where U is the aquifer constant, ⌬P is the pressure difference between initial reservoir pressure with
current reservoir pressure, WD represents the dimensionless aquifer function, tD is dimensionless time and
RD is the outer/inner radius ratio. Aquifer influence was not easily proven as limited water volumes were
recorded over time; therefore, WGRs from LET data were utilized to achieve this. In certain situations,
the reservoir radius, outer/inner radius and encroachment angle were unknown. However, by utilizing the
known original gas in place (OGIP), porosity, permeability and reservoir thickness parameters in
12 SPE-176131-MS

synchronization with MBAL’s regression history matching ability, suitable reservoir pressure matches
were able to be achieved.

Development of Well Model (PROSPER)


PROSPER Workflow
An operating point was defined in order to match the WHP and WHR of the well models. The operating
point was chosen based on the stability of the satellite suction pressures for the most recent date possible.
This date also formed the operating point for the surface facility model. The ‘System Summary’ interface
required the user to define the following information: Fluid Description, Calculation Type, Well Type,
Well Completion, Artificial Lift, Reservoir, User Information and Comments. An aspect of particular
importance was the comment section where the current operating point had to be stated especially when
developing large models. PVT data was required to be consistent with the developed tank model, while
the latest simulated reservoir pressures were obtained from MBAL. The equipment data (well deviation,
geothermal gradient, production tubing and completion diagrams) defined the VLP (the fluid flow
behavior and pressure losses from the perforations to the well-head). Four main types of completions were
modeled: single string, dual, tandem and monobore completions. It was crucial to use the tubing
correlation which was acceptable to the phase behavior of the fluid being produced. This was achieved
through VLP matching where FGS data was used to obtain the most suitable tubing correlation.
The IPR model was then developed and history matched to flowing BHP and WHR. The IPR model
utilized represented a significant engineering decision for gas well models. The Multi-Rate C and n and
Jones IPR models were found to be best suited to the well data available for all gas wells. For oil wells,
the Vogel IPR model was best suited to the data available. The well models were history matched to within
⫾ 2 to 5% of the WHPs and WHRs for the selected operating point (VLP and IPR intersection). The IPR
and VLP intersection of a history matched well model is shown in Appendix F. The well model history
match was very sensitive to the liquid/gas ratios used, which further fueled the need for representative
ratios. A detailed PROSPER workflow is depicted in Appendix G.
Considerations and Assumptions
Black Oil Correlation Model The ‘Black Oil Correlation Model’ was used to describe fluid behavior
for each well model as it is the better option when only limited PVT data is available. This was the case
for most of the wells in the Cooper Basin. The ‘Black Oil Model’ accounted for retrograde condensate
fluids and could predict liquid dropout in the wellbore (PROSPER PE 2009). The ‘Black Oil’ and ‘Dry
and Wet Gas PVT Method’ were utilized for all developed gas well models (no retrograde condensate
models were developed). This PVT process uses input PVT data and adjustment of the ‘Black Oil
Correlation Model’ to match the hydrocarbon composition. The gas phase is, therefore, considered as a
single phase fluid for all pressure drop calculations and multiphase flow is only considered for free water
production (PROSPER PE 2009). However, this was not ideal as condensate behavior in the well was not
accurately described and, consequently, it made it difficult to evaluate wellbore liquid loading issues in
the later stages of a well’s life.
Stable and Cyclic Wells Wells within the Cooper Basin can be classified as either stable or unstable.
Wells which were producing at a constant WHR were considered as stable. Cyclic (unstable) wells
exhibited a constant build-up of BHP under liquid loaded conditions and a short burst of gas production.
Typical cycles lasted four to five hours and could occur more than once a day. Fig. 5 illustrates an example
of a stable and unstable well. Within the Cooper Basin, to assist in well control, a number of cycling wells
have adopted an ‘intelligent’ closed loop control logic code which iteratively monitors and adjusts set
points (targets for minimum shut-in flow rate and WHP) based on historic well performance (Winterfield
et al. 2014). Based on a number of trial operations, the code has been able to respond to field conditions
SPE-176131-MS 13

and operational issues and, in some cases, increased the average flow rate of well in comparison to
parameters manually altered by field engineers.

Figure 5—Example of a stable (left) and unstable well (right)

Within PROSPER, the WHR was required to be entered and, although this was easily determined for
stable wells (due to the constant WHR), there were a number of different methods available to calculate
this value for cyclic wells - (1) the average WHR is taken across a day, (2) the average WHR is taken
across one cycle, and (3) the highest WHR is taken from within one cycle. Since the first method
calculates the average over the entire day, it takes into account the time during the day in which the well
is not producing. However, the other two methods utilises the value of one cycle and, consequently,
assumes that the well is producing at that particular constant WHR throughout the day. Hence, to account
for this overestimation, a downtime was required to be applied in GAP. The concept of downtime will be
discussed later in this paper. The method of which the WHR was determine for cycling wells was based
on an iterative procedure - whichever of the three options yielded the best result when history matching.
Inflow Performance Relationship Model Selection The IPR model adopted represented a significant
engineering decision for gas well models. This choice was due to a trade-off between the required
resources and available development period. Furthermore, the choice could dictate both the accuracy of
the models and also improve the efficiency of model’s development. More specifically, it was important
to balance the accuracy contributed by ad ding more parameters and the time taken to obtain these values,
both of which were related to understanding the sensitivity of the input parameters and the certainty of
each (Bratvold et al. 2010). The size and scale of developing the integrated production model compared
to individual reservoir or well models also had to be considered. These considerations were applied to all
available IPR models in PROSPER, although only the Multi-Rate C and n and the Jones IPR models were
found to be suitable as they provided the most accurate match to WHRs.
The Multi-Rate C and n IPR model is an empirical method derived from the Modified Isochronal well
test. The simplistic formulation and accurate output of the Multi-Rate C and n IPR model were the true
benefits behind its use. The gas rate and average flowing and shut in pressure data are used to calculate
the data to satisfy the Multi-Rate C and n IPR model as described by Equation 5.
Equation 5

where qg represents the gas rate, Pwf2 and Pws2 represent the average flowing bottom hole and shut in
pressures, C is a constant derived from solving Equation 5 using the known pressures and corresponding
14 SPE-176131-MS

gas rates, while n is the slope of the reservoir pressure versus gas rate data as shown in Fig. 6. The Jones
IPR model for stabilized flow is described by Equation 6 (Lee 1982), which illustrates the parameters
required from reservoirs properties and multi-rate tests. The reservoir shape was required to be assumed
in order to calculate the Dietz shape factor, which when considering limited data development conditions
and low permeability environment, was considered to be a very sensitive assumption to make (PROSPER
PE 2009).

Figure 6 —Reservoir pressure vs. gas rate

Equation 6

Table 1 illustrates an example where both the Multi-Rate C and n and Jones IPR models were
employed for a particular well.

Table 1—Comparison between Multi-Rate C and n and Jones


IPR models to metered wellhead gas rate
Gas Rate (1,000m3/day)

Metered (Actual) 18.50


Wellhead Value
Multi-Rate C and n 18.42
IPR Model
Jones IPR Model 19.81

The Jones model required more specific wellbore and reservoir property inputs (for example, drainage
radius, shape factor, permeability and skin) would be better suited for analysing in-wellbore projects
where all required information is known. In addition to this, there is a considerable amount of uncertainty
involved in accurately describing both the permeability and skin parameters. Consequently, as mentioned
previously, it was necessary to weigh up the benefits of adding more parameters to the model to increase
its accuracy and also the uncertainty attached to these parameters. The notion of additional parameters
equating to additional accuracy is a common human bias known as the overconfidence bias. More
specifically, including additional uncertain parameters, but giving them a probability of 1 (probability of
certainty), illustrated the overconfidence bias. Therefore, if additional uncertain parameters were added,
the accuracy itself would not improve, and would rather hinder it further (Bratvold et al. 2010).
VLP Matching VLP matching assists in identifying the most suitable tubing correlation for any well and
any fluid flow phase by utilizing actual performance data (FGS test data). The gravity and friction pressure
loss components are more accurately modeled from this process and allows for the development of highly
accurate well models (PROSPER PE 2009). Scenarios had been observed where tubing correlations,
SPE-176131-MS 15

which were suitable to only horizontal tubing, had been applied to vertical wells to lower BHP below
reservoir pressure and allow well models to function. The following data was obtained from valid FGS
tests: WHP, wellhead temperature, WGR, CGR, WHR, bottom hole depth, BHP and the reservoir pressure
from MBAL for that point in time. These inputs allowed PROSPER to alter correlations, by adjusting
friction and gravity coefficients, to simulate actual performance by providing adjusted correlation values.
FGS data available for the developed well models could not be used as it was only applicable for a
particular date, gas rate and conditions. Hence, VLP matching could only be used to indicate which
correlation would best suit the well production. The friction and gravity coefficients, which had been fitted
to FGS data, could not be used. Due to the large changes in oil composition and volumes, this was only
valid and successfully performed for gas well models. The Gray, PETEX 2 and PETEX 4 VLP correlations
were found to provide the best match for wells flowing at stable rates and unstable rates respectively.
Gray’s correlation provided good results for gas wells with CGRs of up to 50bbls/MMscf and also
accounted for substantial WGRs (Gray 1978). PETEX 2 is a multi-phase correlation which was internally
developed by PETEX. Through comparison with actual measured data, the correlation had been found to
respond well to oil, gas and retrograde condensate phase, for both in-well and pipeline scenarios
(PROSPER PE 2009). The PETEX 4 correlation had been found to provide suitable results for unstable
wells by applying a modified version of Turner’s critical rate equations (PROSPER PE 2009). Appendix
H depicts the output table after VLP matching was conducted in PROSPER. Parameters 1 and 2 represent
the gravity and friction coefficients respectively and, ideally, both should equal to unity. Based on the
parameter values, with respect to the input data entered, the best correlation is able to be selected.
Micro-String Implications It was common for wells that were nearing the end of their production lives
to experience liquid loading conditions (liquid loaded bottom holes did not allow gas production). One
solution which allowed increased production for wells experiencing liquid loading difficulties was the
installation of a micro-string. Adding a micro-string involves installing a 0.25⬙ diameter coil tubing
through which a foaming/surfactant agent is injected. The injection of these agents reduced the interfacial
tension between the water and gas. This mitigated the liquid loading effects which, consequently, would
increase the life of a well. More specifically, the liquid loading phenomenon occurred when the gas
rate/velocity fell below the critical gas velocity known as Turner’s critical velocity, vsl. Turner’s equation
is described by Equation 7 where kv is Turner’s constant (recommended value of 1.3), ␴ represents the
interfacial tension, ␳L is the liquid density, ␳g is the gas density and Cd is the drag co-efficient
(recommended value of 0.44 (Turner et al. 1969)).
Equation 7

The foam/surfactant was found to reduce the liquid density substantially and, hence, reduced the
interfacial tension and ultimately lowered the critical velocity. Interfacial tension describes the surface
energy attraction per unit area at the interface of the water and gas molecules, which is generally uneven
(Dake 1978). Overall, a micro-string lowers the critical velocity required for production. Fig. 7 illustrates
how a micro-string can extend a well’s production life by demonstrating the incremental increase in gas
rate and production. The well mo del development was affected by the inability to model the production
enhancing equipment accurately; therefore, by implementing the effective tubing inside diameter (ac-
counting for the coil tubing diameter) due to the reduction of flow area, this action was able to reduce this
issue (by imitating VLP improvement).
16 SPE-176131-MS

Figure 7—Micro-string well production life implications

Development of Surface Facility Model (GAP)


GAP Workflow
The creation of the surface facility model in GAP required all well details (location, name and type),
flowline and pipeline specifications (length, diameter and roughness) and all elevation data. All active and
inactive wells were included in the surface network as observed in the reservoir and well model
development. Only wells which were online for the chosen operating points were left active in the surface
model. The real trajectories and elevations changes for each pipeline were included, with elevations
changes entered at 100 meter intervals in order to have a better resolution of the pipeline network. For
larger integrated production models, the Flowsheet GAP function was utilized to create sub-models in
order to make it easier to manage. IPR/VLP intersection was found to provide the most accurate results.
Multi-Rate C and n and Vogel IPR models were used to generate the IPR curves in GAP for gas and oil
wells respectfully. In regards to well control, it was essential to use a calculated pressure drop and not the
‘fixed’ option. This was an important best practice which was adopted across all Cooper Basin models.
The calculated option allowed GAP to evaluate the node pressure (in particular WHP) and did not allow
the application of ‘fixed’ ⌬P as a ‘fudge-factor’ to force successful history matching of complete
integrated production models.
All compressors (gas jacks, nodal, low and high pressure) and separators were included which allowed
for an accurate steady-state analysis of the production network. Compressor performance curves, com-
pressor suction and discharge pressures were rigorously analyzed for the chosen operating point as it was
common for these parameters to differ from each data source. This was mainly due to the possibility of
poor communication between field operators and engineers, resulting in undocumented or non-
communicated changes in performance curves, suction and discharge pressures. GAP required that all
production systems end points were separators (as represented by the green and red oval shaped object in
Appendix I). An example of a complete GAP model is also depicted in the same appendix. The GAP
development workflow is documented in Appendix J.
Considerations and Assumptions
Including Offline Wells Wells which were inactive but had scheduled intervention plans were also
included into the GAP model. Once the model was developed, updating models was relatively simple (for
SPE-176131-MS 17

example, if inactive wells were brought back online due to re-perforations or re-completions). For
example, difficult components had been completed and no back-tracking was required for example
collecting input data from scratch. This allowed both reservoir and well changes to be effectively taken
into account.
Multi-layer Gas and Oil Reservoir Well Configuration Multi-layered wells represented wells
which produced from multiple reservoirs. More specifically, the multi-layered reservoir model repre-
sented the co-mingled and tandem well completions which had unique IPR models. Such completions
allowed for production from two or more inflow points (reservoirs/tank models). Multi-layered well
models provided accurate analysis for wells producing from two different gas reservoirs, although if the
one reservoir produced oil, the multi-layered IPR model did not work as PROSPER was limited to
creating solely a gas or oil model. The lesson learnt through a solution proposed by PETEX was the
development of a wellbore in GAP using pipeline to represent the production tubing and casing between
the gas and oil layers (GAP PE 2009). PROSPER models for each well were connected to ‘inflow’
modules and linked to the corresponding reservoir model. This simulated the inflow (only the well IPR
- pipelines which represented the tubing represented the VLP of the well) at the middle of perforation
depth which was driven by the reservoir pressure from the corresponding tank models. The important
consideration was which flowline correlation should be used as it would represent the vertical lift
correlation. The solution was to perform flowline correlation matching using the BHP and WHPs to
simulate the tubing vertical lift correlation of the well. An example of the layout for this consideration is
depicted in Fig. 8.

Figure 8 —Physical multi-layer representation (left) and corresponding GAP multi-layer IPR well model configuration (right)

Well Downtime The downtime history of each well was included in all developed models. The
downtime was a measure of a well’s efficiency - simulated instantaneous rates were multiplied by the
specified downtime. It was crucial that the downtime (entered as a percentage) represented only the time
which the well had been offline as a result of issues or events that were directly related to the well. No
downtime which was a result of surface facility (flowlines, compressor or plant, etc.) issues was
considered. This was found to provide the best results when performing any prediction analysis (future
forecasts). The integrated approach, which was part of the IPM development procedure, assured com-
munication with production support engineers, allowing for the scheduling of future well downtime for
known events (well work-over, plant shut downs, etc.).
18 SPE-176131-MS

As mentioned previously, as cyclic wells exhibited a cyclical behavior of BHP build up and short gas
production, a downtime was required to represent this phenomenon. Failing to do so would overestimate
the duration that wells are online and, consequently, result in an overestimation of production. For
example, if a well was not producing for six hours in a 24 hour period, the downtime was 25%.
Flowline Correlations The primary flowline correlation used for gas networks was the Dukler-Eaton-
Flannigan correlation. Dukler-Eaton-Flannigan was outlined to be best suited for wet and dry gas
production systems by the Santos Design Practice and Fluid Flow Calculations standards. The Dukler-
Eaton-Flannigan pressure drop correlation was designed for two phase flow systems with a primary gas
phase. The correlation originated initially from three individual correlations - the Dukler correlation for
horizontal pipes, the Eaton correlation for liquid loading and the Flannigan correlation for vertical
elevations. The combination of these three provided an overall improved correlation. For production
systems where inline field pressure measurements were available, it was considered mandatory to perform
flowline correlation matching. As with VLP/IPR matching in PROSPER, the gravity and friction
coefficients were adjusted to match measured pressures to simulated pressures. For production systems
where no inline field pressure measurements were available, the assumption was made from past practice
that based on fluid and field environment (conditions and elevation changes) characteristics, the Dukler-
Eaton-Flannigan was the best suited correlation. For oil flowlines, the Vasquez-Beggs correlation was
assumed to best describe the oil flow behavior.
Compressor Development The GAP software makes it possible to model compressors as full model,
fixed ⌬p, fixed power and reciprocating compressors (GAP PE 2009). The most efficient and suitable
method to model compressors was as in-line general objects with constrained Ap at inlet nodes by utilizing
compressor performance curves. The driving factor behind the method used was the availability of
manufacturing data (efficiency, water removed, oil removed, gas removed and cooling data) compared to
easily generated performance curves (up to date suction and discharge pressure). The difference in
accuracy of results by using fixed ⌬p and other options were determined to be negligible. This was further
supplemented by the ease of which compressor performance curves could be generated. This method was,
therefore, applied to all developed models and compressor performance curves were inputted as coded
scripts. In order to guarantee accurate compressor performance, these in-line general objects were aided
by upstream nodes. This was necessary to represent real compressor behavior in terms of flow compared
to suction pressure. The flow and pressure values Qmin, Qmax, Ps1, Ps2 and Psmin were required and were
used to specify an operating range for the compressor unit (essentially, it was the way that GAP
constructed the performance curve). The crucial component in modeling compressors as in-line general
objects was the use of upstream nodes. The nodes were used to constrain the suction pressure and inflow
limits (lower and upper) which corresponded to the real asset operating range. This was required because
the GAP software would extrapolate the input performance data if a flow rate did not lie on the specified
compressor performance curve. However, in reality, the compressor unit would either enter a recycling or
full capacity state or cease to function. If the constraints had not been applied, incorrect results could be
obtained. The setup of the satellite compressor units is illustrated by Appendix K. The satellites in each
model only accounted for low-pressure compression unit. Hence, when analyzing results, high-pressure
compression limits were required to be applied in a deterministic manner and externally from GAP.
Surface Facility Constraints Constraints which were present in the surface facility model included
pipeline design pressures, velocity constraints (which was function of different inhibitor methods - batch
compared to continuous) and compressor constraints (which have been previously detailed). The maxi-
mum velocity (to avoid erosion) and maximum design pressure constraints for all pipes were entered. Both
constraints were dependent on the pipe material, although the maximum velocity was highly influenced
by the inhibitor type. Two types of inhibitor injection methods existed - batch and continual corrosion
inhibitor - which provided velocity restrictions of 20 and 30 m/s respectively. Therefore, any well which
had a high rate (⬎10 MMscf/d) had their pipe spools (chokes & elbows) added to the model. Pipe spools
SPE-176131-MS 19

(10 to 30 meters in length and 4 to 6 inches in diameter) represent the instrumentation from the wellhead
(small pipes, pressure safety valves, chokes, etc.) to the flowline. While the possibility of exceeding the
maximum velocity and pressure constraints existed in such pipe components, the small pipe length meant
that these spools could be neglected from the GAP model as they would only contribute a small amount
to the overall pressure drop. These constraints were identified as being the most important constraints to
include in order to achieve accurate system representation without affecting model performance.

Calibration of Integrated Production Model (GAP)

Calibration Workflow
Once simulation runs could be performed successfully, it was necessary to validate each integrated
production model. The calibration procedure provided the final application of quality control to the
integrated production models as the simultaneous assessment of the entire production system could expose
inconsistencies and errors in the reservoir, well and surface models. Consequently, the reservoir, well and
surface model inputs were all re-assessed. In addition to the quality check, the history matching procedure
required the matching of simulated data to historical data. The simulated WHPs, WHRs, inline flowline
pressures and satellite pressures were history matched to within a ⫾ 2 to 5% range in comparison to the
historical data. This also indicated the importance of using only calculated ⌬P and not fixed AP well
control. The operating point for the complete integrated production model was the date which the well
models had been history matched to. The ‘Prediction’ GAP function was used to simulate the model up
to the operating point using a suitable time-step. The assumption that no additional wells came online
between the prediction start date and model operating point was only applied at this point of the calibration
procedure. For prediction scenarios, wells which had come online after the operating point were scheduled
to come online. Within this function, there is also the option to allow GAP to automatically optimize the
choke settings for any wells, such that it maximizes satellite production from the network.
The pressure behavior of the surface networks was very sensitive to gravity and friction pressure losses.
It was the accuracy to which the models could predict the flowline pressure drop and changes in pressures
that defined the model’s forecasting accuracy. Okoh et al. (2010) had previously outlined the importance
of these parameters and highlighted that matching flowline correlations was mandatory wherever possible.
The gravity and friction co-efficient were not used as ‘fudge-factors’ to increase or decrease the WHPs
and/or WHRs. Past practice only modeled pipelines as being horizontal; however, the inclusion of
elevation changes further improved the model’s representation of the gravity and friction losses, especially
where large elevation changes were evident as shown in Fig. 9. The influence of liquids volumes on the
pressure drop was also noted. It was important to first have all online wells flowing when performing
predictions, followed by matching all stable flowing wells and lastly matching all cycling wells. The
surface model constraints were only added once all stable wells had satisfied the matching criteria. Finally,
to assess the long term forecast accuracy, the predicted integrated production model output was compared
to mainstream company short term production forecasts (for two years). Once calibrated, any new
developments (pipeline, new drills, etc.) were then added.
20 SPE-176131-MS

Figure 9 —Elevation change plot from GAP

Considerations and Assumptions


Pipeline Design Life The Cooper Basin production systems varied in age from approximately 20 to 50
years. Consequently, this meant that although they are still operating sufficiently, it was not uncommon
for pipelines to have exceeded their specified design life. The roughness of such pipelines had increased
over time and, hence, increased friction pressure losses. It was, therefore, adopted as common practice to
adjust the roughness on the basis that it was approved or advised by pipeline integrity engineers as failing
to make this adjustment had been found to potentially make it very difficult to calibrate models.
Predicting Liquid Volumes The use of either deterministic WGR values (fractional flow curves in
MBAL) or relative permeability curve outputs influenced liquid volumes as current liquid volumes were
only obtained using relative permeability curves. Therefore, the use of relative permeability curves was
preferred as the dynamic WGR would accurately describe the changes in pressure drop. The use of a
single fixed WGR value in GAP was found to both underestimate and overestimate the pressure drop. An
under-estimation of pressure was common prior to water breakthrough. Over-estimation of pressure drop
could occur if a well was flowing back fracturing fluid for an extended period of time for an extremely
wet sand or if a formation had been isolated; however, these issues were related to the importance of using
correct and representative input data.
Prediction Timestep It was necessary to select a suitable time step of evaluation and GAP provided a
range of possible time steps from seconds to years. Logically, the smaller the scale, the higher the
accuracy; conversely, the larger the time step, the lower the accuracy. In particular, the use of large a
timestep was found to directly affect reservoir pressures and, subsequently, surface network pressures. For
high rate wells, large volumes of hydrocarbons could be withdrawn in a single timestep (GAP PE 2009).
In addition to this, depending on the date, the time step of evaluation was chosen so that the results on
that day could be analyzed or, if not possible, a mock scheduled event was used to trigger an additional
timestep.
IPR Representation in GAP The use of consistent IPR models in PROSPER and GAP was found to
affect the results and hinder the quality check (QC) process for both gas and oil wells. It had been
documented by Amudo et al. (2011) and determined through the previous development of integrated
production models for the Cooper Basin that inconsistencies between the IPR models in PROSPER and
GAP can occur. The IPR model in PROSPER was required to be consistent with the model selected in
GAP, and this was best created by using Multi-Rate C and n for gas wells and Vogel for oil wells. No
variance was observed in the IPR generated in GAP (prior to prediction) and the PROSPER IPR models.
Consistent water cut, GOR and WGRs were required across both the well and surface models if fractional
SPE-176131-MS 21

flow curves had been used in the reservoir model. For well models where the Jones IPR model was used
in PROSPER, the Forcheimer Psuedo Pressure IPR model was required to be selected in GAP.
VLP Data Tables The VLP curves which GAP generated required ranges of values to be set for the
WHP, WHR, CGR, and WGRs. A general range for each parameter was used at the initial stages of the
QC procedure to simply allow the model to run successfully. The ranges on each of these parameters were
then refined and reduced as the QC procedure progressed. It was evident that the range of values used
influenced the accuracy of result and, therefore, refining of these ranges was crucial. Unlike the WHP and
WHR which decreases over time, the same could not be said for the CGR and WGR as uncertainty was
present with their upper limits. If the integrated production models were fully compositional or analysis
in PVTP had been performed, condensate behavior (condensate drop out or if condensate was being
produced) could be predicted. Relative permeability curves allowed for the forecast of changes in liquid
and gas ratios. Therefore, initial simulation runs could be used to predict the upper limits of these
parameters and allow for optimum ranges to be set. It was also noted that for long term scenario
forecasting (periods greater than three years), VLP data was required to be generated in a geometric
manner. This was due to instabilities which were common at the end of well life or at low rates. Fig. 10
depicts an example of four satellites production of which only three (represented by brown, red and blue)
have utilized geometric VLP data tables for all wells in their respective production networks. Without
employing the geometric function, it is evident that the green satellite’s forecasted production will
experience instability in its gas production later in its life (as demonstrated by the large fluctuations).

Figure 10 —Demonstration of unstable gas rate in later life (green) as a result of not adopting geometric VLP tables

Integrated Production Model Development and Validation Report As each model was developed by
the company’s IPM team, a peer review was required to be conducted by the IPM team manager and the
field’s dedicated reservoir and production engineers. Due to the large quantity of work, input parameters,
considerations and assumptions involved in the development of each model were documented. It was
important to document these in an Integrated Production Model Development and Validation Report along
with the methods used to obtain input data, its values and various data sources used. By adopting this
approach, it ensured that no fundamental theory and necessary constraints were violated during the
22 SPE-176131-MS

development and calibration of the integrated production model. Once approved and signed off, the model
could then be applied to desired scenarios.
Development Procedures for Optimization Scenarios

The extensive application of calibrated integrated production models in the Cooper Basin had also
identified methods which enhanced the application of IPM to optimize, develop and produce forecasting
scenarios. The model’s forecasting accuracy was defined by the accuracy to which the models could
predict the pressure drop. Therefore, if models were calibrated correctly, it was in the best interest of the
company to fully utilize the capabilities of the model. Hence, any optimization, development (new well
drills) and production forecasting scenarios were all allocated resources and development periods which
enabled pressure responsive analysis to take place.
MBAL and PROSPER models were developed for all new development. This practice proved
beneficial for the following scenarios: (1) backout studies for new well drills, (2) new pipeline sizing
(assessing pressure losses) and (3) future compression configuration effects on new well drills. Evaluating
these scenarios in this manner allowed for accurate assessment of reserve addition/loss, incremental rates
and surface network pressure behavior. This practice was beneficial and in the best interest of the
company as it was directly related to making good business and capital investment decisions.
A lesson learnt when conducting optimization scenarios was the avoidance of using fixed source
inflows within GAP as they were not susceptible to backpressures. Table 2 demonstrates the effect with
and without using fixed source inflows. Six development wells come online at different dates and their
initial wellhead gas rate and corresponding backout (at the satellite) are listed. It is evident that there are
differences with the inclusion of fixed sources; however, when compared to the actual rates and backout
in real life, the MBAL and PROSPER measured values are much more accurate. Although the variance
with or without using fixed sources are small, when applied to economics, these differences can have a
large impact on a project’s cash flow, especially with high producing wells. Such examples of improperly
defining project economics can then lead to projects that are deemed to be economical becoming
uneconomical and vice versa. It must be noted, however, that the variance between the values obtained
with fix source inflows in GAP and values from MBAL and PROSPER are unique to each production
network.

Table 2—Comparison between values obtained using with and without fix source inflow
Without Fix Source
With Fix Source Inflow (MBAL &
Inflow in GAP PROSPER only)

Real Qincremental Qincremental Percentage


Development Online of satellite Percentage at Satellite Backout
Well Date (MMscf/day) Qincremental at Satellite (MMscf/day) Backout (%) (MMscf/day) (%)

1 10/11/2014 3.05 3.2 33.8 3.0 30.7


2 17/11/2014 2.30 2.0 30.4 2.2 30.4
3 23/11/2014 3.95 4.2 25.5 3.9 21.1
4 17/02/2015 3.80 4.1 34.5 3.7 27.1
5 21/02/2015 2.45 2.7 31.5 2.5 26.5
6 25/02/2015 2.20 1.9 27.1 2.1 27.8

Furthermore, a best practice was using the calculated ⌬P well control function within GAP that allowed
for choke optimization scenarios to be performed. Appendix L details 42 wells within a satellite in which
the calculated ⌬P method was employed. By utilizing this function, an increase of 1.4 MMscf/day was
SPE-176131-MS 23

seen at the satellite (as opposed to if all the chokes within the satellite were 100% open). The appendix
also documents the wells and the amount required to be choked back in order to provide this rate increase.

Development & Calibration Summary


Engineering decisions were established as the governing factors behind the integrated production model
development and calibration process. These decisions essentially dictated both the accuracy and efficiency
of the implemented development process and workflows. The importance and consequences of these
engineering decisions were well understood, as was their overall effect on the model’s accuracy. This
allowed for the history matching error tolerance of ⫾ 2 to 5% to be satisfied for all Cooper Basin
integrated production models. The general case across all Cooper Basin models were that these decisions
were driven by the relationship between required resources and available developments periods both of
which were directly affected by data constraints. The considerations of what was to be required from the
application of each integrated production model was used to identify the level of detail required for each
model. Fundamentally, it was understood that the developed integrated production model should only be
as detailed as required to successfully fulfil all modeling requirements in an efficient manner. The
workflows and considerations presented are examples of these decisions.

Conclusion
The workflow presented by this paper allowed the successful development and calibration of integrated
production models. The need for an optimum and innovative workflow was achieved, with emphasis
given to resources allocated compared to development period which ultimately determined the model
quality. The best practices and lessons learned can be summarized as:
● The relationship between desired outcomes and the required model complexity must be estab-
lished. This is related to the considerations and assumptions made throughout model development.
● Consequences of considerations and assumptions must be understood. The variance of their effects
for different scenarios must also be understood.
● Comprehensive understanding of fundamental theory of reservoir, well and surface/production
engineering as well as physics behind each software component are required.
● Development and calibration are heavily influenced by quality control. Documentation, workflows
and standards are crucial to not only maintaining model quality but also the way in which model
quality can be assessed. This was under-pinned by good inter-disciplinary communication.
● The workflow, considerations and assumptions presented in this paper were able to improve the
ability to develop, calibrate and operate integrated production models.
● The combination of all conclusions had contributed to an innovative approach for IPM of these
Cooper Basin satellites and highlighted the value of data.

Acknowledgements
The authors would like to thank Santos, Origin and Dehli management for the permission to publish this
paper and all colleagues who have contributed to the development and calibration of these integrated
production models. Special thanks to Geoff Forrest, Vicki Chan, Rebecca Price, Aaron Lawson, Simon
Chipperfield, Brenda Chang, Catherine Karuga and Maria Rondon of Santos Ltd for their assistance with
model development. In addition to this, thanks are extended to Jan Bon Jr of Petrolab for his assistance
with PVTP and Alex Chwetzoff for his assistance with this paper’s development. The authors would also
like to acknowledge the support from the University of Adelaide’s Australian School of Petroleum staff
for their support to the Petroleum Engineering honors project which formed the basis for this work. The
IPM Package (PVTP, MBAL, PROSPER and GAP) are trademarks and under copyright of © Petroleum
Experts Ltd.
24 SPE-176131-MS

Nomenclature
Bg Gas Formation Volume Factor
Bgi Initial Gas Formation Volume Factor
BHP Bottom Hole Pressure
Bo Oil Formation Volume Factor
Boi Initial Oil Formation Volume Factor
Bw Water Formation Volume Factor
C Multi-Rate Coefficient
cd Drag Coefficient
cf Formation Compressibility
CGR Condensate Gas Ratio
Ct Total Compressibility
cw Water Compressibility
D Non-Darcy Flow Constant Appearing in Rate Dependent Skin
EOR Enhanced Oil Recovery
FGS Flowing Gradient Survey
GA Gas Analysis
GAP General Allocation Package
QC Quality Check
GOR Gas Oil Ratio
H Height/Thickness of Formation
IPM Integrated Production Modeling
IPR Inflow Performance Relationship
K Permeability
krgmax Maximum Gas Relative Permeability
krwmax Maximum Water Relative Permeability
kv Turner’s Constant
LET Liquid Evaluation Test
M Ratio of the Initial Hydrocarbon Pore Volume of the Gas-Cap to that of the Oil
MBAL Material Balance
MICP Mercury Injection Capillary Pressure
MPLT Memory Production Log Test
MRTLL Minimum Rate to Lift Liquids
N Slope of Pressure versus Gas Rate
N STOIIP
NaCl Sodium Chloride
Np Cumulative Hydrocarbon Production
OGIP Original Gas in Place
PBU Pressure Build Up Test
PETEX Petroleum Experts
Pres Average Reservoir Pressure
PROSPER Production and Systems Performance Analysis Software
Psmin Absolute minimum suction pressure
Ps1 Minimum Suction Pressure
Ps2 Maximum Suction Pressure
PVT Pressure, Volume and Temperature
PVTP Pressure, Volume and Temperature Package
SPE-176131-MS 25

Pwf Average Flowing Bottom Hole Pressure


Pws Average Shut-In Bottom Hole Pressure
Q Bottom Hole Flow Rate
Qmax Maximum Flow Rate
Qmin Minimum Flow Rate
qg Gas Flow Rate
RD Outer/Inner Radius Ratio
␥e Reservoir Drainage Radius
␥i Radius of Investigation
RP Producing Gas Oil Ratio
Rs Solution Gas Oil Ratio
␥w Well Radius
S Skin
Sgmax Maximum Gas Saturation
Sgr Residual Gas Saturation
STOIIP Stock Tank Original Oil in Place
Swc Connate Water Saturation
T Time
T Temperature
tD Dimensionless Time
U Aquifer Constant
VLP Vertical Lift Profile
vsl Turner’s Critical Velocity
WD Dimensionless Aquifer Function
we Water Influx
WGR Water Gas Ratio
WHP Wellhead Pressure
WHR Wellhead Rates
Wp Cumulative Water Production
Z Compressibility Factor
⌬p Decline Pressure Difference
␮ Viscosity
␳g Gas Density
␳L Liquid Density
␴ Interfacial Tension
␸ Porosity

References
Aditama, P., Sujai, A., Pamungkas, H. W. et al 2010. Construction and Implementation of Integrated
Production Network Model in a Complex Onshore Operation. Presented at the SPE Oil and Gas
India Conference and Exhibition, Mumbai, India, 20-22 January. SPE- 128381. http://dx.doi.org/
10.2118/128381-MS.
Alexander, E.M. 1998. The Petroleum Geology of South Australia. Vol. 4: Cooper Basin - Chapter
6: Lithostratigrahpy and Environments of Deposition. pp. 69 –116. Department of Primary
Industries and Resources, South Australia.
26 SPE-176131-MS

Amudo, C., Walters, M. S., O’Reilly, D. I. et al 2011. Best Practices and Lesson Learned in the
Construction and Maintenance of a Complex Gas Asset Integrated Production Model (IPM).
Presented at the SPE Asia Pacific Oil and Gas Conference and Exhibition, Jakarta, Indonesia,
20-22 September. SPE-146968. http://dx.doi.org/10.2118/146968-MS.
Arnondin, M.C and Jackson, M. A. 1996. A Systems Approach to production Management: Beryl
Field Case Study. Presented at the SPE European Production Operations Conference and Exhi-
bition, Stavanger, Norway, 16-17 April. SPE-35558. http://dx.doi.org/10.2118/35558-MS.
Bratvold, R and Begg, S. 2010, Making Good Decision, Dallas, Texas: Society of Petroleum
Engineers.
Correa Feria, C. 2010. Integrated Production Modeling: Advanced, But Not Always Better. Presented
at the SPE Latin America and Caribbean Petroleum Engineering Conference, Lima, Peru, 1-3
December. SPE-138888. http://dx.doi.org/10.2118/138888-MS.
Craft, B. C., Hawkins M., Terry, R. E. 1991. Applied Petroleum Reservoir Engineering, second
edition. United States: Prentice Hall Inc.
Dake, L. P. 1978. The Fundamentals of Reservoir Engineering. Amsterdam, The Netherlands:
Elsevier Science B.V.
Jeje, O. and Mattar, L. 2006. Comparisons of Correlations for Viscosity of Sour Natural Gas. Journal
of Canadian Petroleum Technology Volume 45, No.7. http://dx.doi.org/10.2118/06-07-TN1.
Lee, J. 1982. Well Testing. Richardson, Texas: Society of Petroleum Engineers.
GAP Petroleum Experts 2009, User Manual, GAP, Edinburgh, Scotland, Petroleum Experts Ltd.
Gray, H.E. Vertical Flow Correlation - Gas Wells. User Manual for API 14B, Subsurface Controlled
Safety Valve Sizing Computer Program, App. B (Jan., 1978) pp. 38 –41.
MBAL Petroleum Experts 2009, User Manual, MBAL, Edinburgh, Scotland, Petroleum Experts Ltd.
Nadar, M. S., Kulkarni, R., Lemanczyk, Z.R. et al 2006. Development of an Integrated Production
Network Model for the Heera Field. Presented at the SPE Asia Pacific Oil and Gas Conference and
Exhibition, Adelaide, Australia, 11-13 September. SPE-101089. http://dx.doi.org/10.2118/
101089-MS.
Noaman A. F, El-Khatib. 1982. The Effect of Drainage Area and Production Rate on the Performance
of Depletion Drive Oil Reservoirs. Presented at the 57th Annual SPE Fall Technical Conference
an Exhibition of the SPE of AIME in New Orleans, LA, 26-29 September. SPE-11019. http://
dx.doi.org/10.2118/11019-MS.
Okoh, E., Sathyamoorthy, S., Olaniyan, E. et al. 2010. Application of Integrated Production System
Modelling in Effective Well and Reservoir Management of the Bonga Field. Presented at the 34th
Annual SPE International Conference and Exhibition, Tinapa-Calabar, Nigeria, July/August.
SPE-140632. http://dx.doi.org/10.2118/140632-MS.
Ozdogan, U., Keating, J. F., Knobles, M. et al. 2008. Recent Advances and Practical Applications of
Integrated Production Modelling At Jack Asset in Deepwater Gulf of Mexico. Presented at the SPE
Europec/EAGE Annual Conference and Exhibition, Rome, Italy, 9-12 June. SPE-113904. http://
dx.doi.org/10.2118/113904-MS.
PROSPER Petroleum Experts 2009, User Manual, PROSPER, Edinburgh, Scotland, Petroleum
Experts Ltd.
PVTP Petroleum Experts 2009, User Manual, PVTP, Edinburgh, Scotland, Petroleum Experts Ltd.
Shrestha, T., Hunt, S. and Lyford, P. 2008. Workflow for Integrated Production Modelling of Gas
Wells in the Northern Cooper Basin. Presented at the SPE Asia Pacific Oil and Gas Conference
and Exhibition, Perth, Australia, 20-22 October. SPE-116936. http://dx.doi.org/10.2118/116936-
MS.
SPE-176131-MS 27

Turner, R. G., M. G. Hubbard, and A. E. Dukler. 1969. Analysis and Prediction of Minimum Flow
Rate for the Continuous Removal of Liquids from Gas Wells. Journal of Petroleum Technology.
Volume 21 pp. 1,475–1,482. http://dx.doi.org/10.2118/2198-PA.
Walrick van Zandvoord, E.J.J. and Hooimeijer, M. 2007. Smart Model Simplifications to Speed Up
Uncertainty Analysis of Integrated Production System Models Used For MT Forecasting. Pre-
sented at the SPE Asia Pacific Oil and Gas Conference and Exhibition, Jakarta, Indonesia,
October/November. SPE-108947. http://dx.doi.org/10.2118/108947-MS.
Winterfield, C.D., Missikos, J., Tio, R. et al. 2014. Application of Closed Loop Control Logic to
Optimise End of Life Cycling Production in a Mature Gas Field. Presented at the SPE Asia Pacific
Oil and Gas Conference and Exhibition, Adelaide, Australia, 14-16 October. SPE- 171523.
http://dx.doi.org/10.2118/171523-MS.
28 SPE-176131-MS

Appendix A

Stratigraphic summary of the Cooper Basin, South Australia (Alexander, 1998)


SPE-176131-MS 29

Appendix B

Input data folder directories


Folder Name Files Input Data

Geological Mapping Data Topographic Maps, –


Reservoir Properties
Field Formation Information – Reservoir Production
Allocation Percentage
Compartment Analysis Pressure Transverse Reservoir Pressure and
Plots Producing Formation Data
MBAL Input Data MBAL Input Data, Field/Reservoir/Well allocation,
OGIP PVT, Reservoir, LET Data
Surface Pipeline Inputs Pipe Characteristics Design Life, Rate, Pressure,
Diameter, Length, Material
Production History Production History Reservoir Division, Production
by Well and Field
PROSPER Input Data PROSPER Modeling Well Status, PVT, Pressures, Rates
Database
Well Testing Data Pressure, Production, Time Plots Well Test, Reservoir Pressure
and Cumulative Production
data
Quality Check Data QC Spreadsheet Historic Pressures
and Rates
30 SPE-176131-MS

Appendix C

General MBAL workflow


SPE-176131-MS 31

Appendix D

Potential pressure communication between wells due to an overlap in drainage radius


32 SPE-176131-MS

Appendix E

Values used in the generation of a relative permeability curve


Residual Saturation End Point
(fraction) (fraction) Exponent

Krw 0.25 0.48 2.115


Krg 0.2 0.77 1.785

Relative permeability curve


SPE-176131-MS 33

Appendix F

Inflow (IPR) and Outflow (VLP) intersection plot


34 SPE-176131-MS

Appendix G

General PROSPER workflow


SPE-176131-MS 35

Appendix H

VLP/IPR matching (tubing correlation) results


Co-efficient 1 Co-efficient 2
Correlation (Gravity) (Friction)

Gray 1.0791 1.60492


Petroleum Experts 1.1312 2.26537
Petroleum Experts 2 1.0826 1.54193
Petroleum Experts 3 1.1961 3.56275
Petroleum Experts 4 1.0852 1.61342
Petroleum Experts 5 0.9123 2.24123
36 SPE-176131-MS

Appendix I

Complete GAP surface network drawing


SPE-176131-MS 37

Appendix J

General GAP workflow


38 SPE-176131-MS

Appendix K

Inline separator, compressor, upstream and downstream nodes


SPE-176131-MS 39

Appendix L

Optimized amount for wells to be choked back using calculated AP well control function within GAP
⌬P Choke ⌬P Choke ⌬P Choke
Well (psi) Well (psi) Well (psi)

A1 0 C3 0 F1 0
A2 0 D1 6.05 F2 0
A3 0 D2 8.23 F3 0
B1 0 E1 27.83 G1 0
B2 0 E2 0 H1 0
B3 0 E3 0 H2 0
B4 0 E4 0 H3 10.48
B5 0 E5 0 H4 0
B6 0 E6 0 H5 0
B7 0 E7 0 I1 0
C1 0 E8 0 I2 0
C2 10.93 E9 0 I3 0

View publication stats

You might also like