You are on page 1of 20

SPE 158497

Integrated Asset Modeling for Reservoir Management of a Miscible WAG


Development on Alaska's Western North Slope
Roadifer, R.D., ConocoPhillips Alaska, Inc., Sauvé, R., Schlumberger, Torrens, R., Schlumberger Middle East
SA, Mead, H.W., Pysz, N.P., Uldrich, D.O., ConocoPhillips Alaska, Inc., and Eiben, T., ConocoPhillips Canada

Copyright 2012, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Annual Technical Conference and Exhibition held in San Antonio, Texas, USA, 8-10 October 2012.

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
An integrated asset modeling (IAM) approach has been implemented for the Alpine Field and eight associated satellite fields
on the Western Alaskan North Slope (WNS) to maximize asset value and recovery. The IAM approach enables the
investigation of reservoir and facilities management options under existing and future operating constraints. Oil, gas and
water production from these fields are processed at the Alpine Central Facility (ACF). A number of local constraints exist
for the asset, such as the requirement that all associated gas be used for facilities power generation, gas lift or re-injection.
All produced water must be re-injected and, for pipeline integrity reasons, must be segregated from imported make-up sea
water used for injection. Additionally, surface gas and water handling capacity is limited at the ACF. To further complicate
matters, gas injected for EOR purposes is enriched such that it is miscible or near-miscible at reservoir conditions. These
conditions create a unique and changing relationship between the oil, gas and water production, gas lift, miscible water
alternating gas (MWAG) injection, lean gas injection, facilities constraints and injection availability.

The scope of the current IAM project has been multi-fold. Optimization of oil production across all WNS fields requires the
placement of injection fluids be simultaneously optimized. The optimization procedure begins by allocation of oil production
targets based on current operating conditions, the potentials of the wells in each field to deliver fluids, and total gas lift
availability. Excess gas compression capacity is utilized for gas lift and is allocated via an incremental gas-oil ratio sort on
the production wells. Given the constraints on water injection noted above, optimization of injection fluids begins by
determining pump requirements for produced water and the optimal field or injection manifold placement of the produced
water. Following this, optimized placement of the miscible injectant (MI) and lean gas injectant (LGI) is determined based
on a dynamic MWAG scheduling methodology developed to maximize oil recovery and ensure the number of gas injection
wells have sufficient capacity to inject the required volume of gas in each reservoir. The volumetric split of gas into MI and
LGI streams falls out directly from the specification of a target minimum miscibility pressure (MMP) constraint for the MI
and the volume of condensates driven off the top of the condensate stabilizer column at the process facility. Finally, the
volume of the make-up fluid (sea water) is determined based on the minimum of the remaining pump capacity or potential of
the remaining wells to inject the water and allocated to each field based on a fractional oil voidage replacement scheme.
Maximizing production across multiple fields necessarily requires that the best player (well) plays, regardless of the field to
which it belongs. This requirement relates to both instantaneous production as would be considered under a gas lift
optimization scenario as well as the longer term MWAG performance and recovery of each individual well pattern across all
the fields.

The IAM technology utilized for managing the WNS fields consists of full-field compositional reservoir simulation models
for each reservoir integrated with a pipeline surface network model and a process facility model. Spreadsheet based
allocation routines and advanced mathematical coupling algorithms complete the IAM model enabling not only the
prediction of the assets’ performance under the aforementioned constraints, capacities and operating conditions, but to
optimize overall performance and analyze the impact of decisions. To the authors’ knowledge, this is the first time integrated
asset modeling has been applied to bring the entire production stream including reservoir, wellbore, surface network and
process simulation together for planning and managing MWAG injection to optimize recovery from an existing development.
2 SPE 158497

Introduction

Previous Works
Integrated asset models of various forms have appeared in the literature for a number of years. Probably the largest advance
in these models over the last ten years or so has been in the development of open interface software for integrating the
disparate software components comprising the reservoir model, the surface network or pipeline model, and the processes or
facility models. Prior to reviewing what we consider to be the state-of-the-art in an integrated asset model, it will be
worthwhile to review some of the most recent (last 15 years) models presented in the literature.

Litvak et al. (1997) incorporated the network pipeline model directly into the reservoir simulation program to enable the
simultaneous solution of that entire system. Zapata et al. (2001) on the other hand, simply coupled a reservoir simulation
model to a network pipeline model. They used implicit coupling of the surface network model to the reservoir simulation
model at the bottomhole and suggested that implicit coupling was necessary in their case. Chow and Anondin (2000) used
material balance methods for the reservoir coupled with Monte Carlo methods in the IAM process to establish the uncertainty
in their model results. Liao and Stein (2002) chose to use material balance approximations for the reservoir portion of their
integrated model.

In 2003, Ghorayeb et al. described an integration tool in which reservoir simulation models were loosely coupled to the
surface network system. In this scenario, the reservoir models and the surface network models were “reconciled” only at
specified synchronization steps. They also discussed component splitting and lumping methodologies implemented in the
controller for the blending of fluid streams from the simulator into the surface network and the reconciliation of injection
streams back down to each reservoir simulation model.

Liao et al. (2007) followed their previous work with a new IAM model wherein their model was directly tied to the SCADA
system. The main purpose of the IAM model was to reconcile well test data with the SCADA system via adjustment of well
productivities or skin factors to match the well performance obtained from the well tests. Once the well information had
been updated, the model could then be used for production and gas lift allocation.

Gutierrez et al. (2007) used reservoir proxy models coupled to a surface network program and process facility program to
conduct a gas lift optimization study. Tesaker et al. (2008) used simplified reservoir models (SRM) coupled with a surface
network program to conduct optimization studies on WAG injection wells.

In 2009, Serbini et al. presented an IAM integrating 7 reservoir simulation models with a surface network model for a brown-
field re-development study. An immiscible water alternating gas EOR recovery scheme was being considered. In their work,
the surface network model was explicitly coupled to the reservoir models at the wellhead. Though explicit, the IAM did
iterate between the reservoir and surface network models to within a specified tolerance. The application of the IAM in this
paper resulted in significant modifications to the pipelines and facilities to optimize future production.

Gonzalez et al. (2010) presented work on a fully-compositional integrated asset model build for a giant gas-condensate field.
Their IAM integrated reservoir simulation models, production and injection surface network models, and a process model.
All reservoir simulation models were fully compositional. The model was built to address facility configuration under ever
changing service of wells and the re-injection of gas streams under various H2S and CO2 constraints.

Developing an Integrated Asset Model – Reservoir Management Needs


As with the construction of any reservoir or surface model, the first step in developing an Integrated Asset Model is to define
with as much clarity as possible the objectives in undertaking such an endeavor. Simply stating better reservoir management
or maximizing profit from the field as goals, while noteworthy, is not sufficiently detailed to define the scope of the project
and the level of detail that will be required in the model, nor the amount of manpower that will be required to build the
model. Satter et al. (1994) stated that sound reservoir management practice relies on utilization of available resources
(human, technological and financial) to maximize profits from a reservoir by optimizing recovery while minimizing capital
investments and operating expenses. Reservoir management then is about making choices under the aforementioned
constraints. In essence, one can choose to let reservoir management happen, a reactionary philosophy typically resulting in
extensive and continuous “fire fighting” efforts to recover production affected by some unforeseen event. Or, one can
choose to make reservoir management happen, a proactive, sometimes preventive, philosophy resulting in proper planning,
design, and development driven to maximize profits over the longer term.

What became clear to the WNS group was that the tools in current use were unable to capture the complex interaction
between the fields, surface network, and process facility. Multiple models existed with little integration; e.g., reservoir
simulation models for individual fields, limited wellbore hydraulics models, surface network models for individual pipeline
segments, a full blown process facility model, and a variety of overly simplified Excel™ spreadsheets with Visual Basic
SPE 158497 3

Figure 1 – Location and Identification of Western North Slope Reservoirs.

programming behind them for trying to combine the production and injection streams into some coherent predictive profile
for production and injection. In fact, there was no direct interaction between fields sharing the common facilities. Material
balance and optimization on injection fluids was difficult and a rough estimate at best. The impact of changes in field and
facility operating conditions on total production was difficult to predict with any certainty. Long term forecasting was a
multi-step, time-consuming manual process done once a year for internal planning. Lastly, rigorous evaluation and
comparison of opportunities was not possible.

A detailed list of all possible objectives for creating an IAM covering all possible development and operational situations that
might occur would be quite extensive. As is often the case, however, simply identifying the “big rocks” will allow the finer
details of those leveraging aspects to be identified and planned for in the development of the IAM.

The Alpine field anchors the western-most oil production and processing facility on Alaska’s North Slope (see Fig. 1), thus
the moniker Western North Slope (WNS). Discovered in 1994, the Alpine field is located in the Colville River Delta six
miles south of the Arctic Ocean and approximately 70 miles west of the Trans Alaska Pipeline. The Alpine field began
production in November 2000 and continues development today. Subsequently, a number of satellite fields including the
Fiord-Nechelik, Fiord-Kuparuk, Nanuq-Kuparuk, Nanuq-Nanuq, Qannik and Alpine-Kuparuk have been brought online and
continue to be developed. Additionally, there are a number of fields in the National Petroleum Reserve with the potential to
be developed in the future. The common theme across all these developments is that they are or will be produced through the
Alpine Central Facility (ACF).

API oil gravities for these fields range from 27-47 degrees and the initial solution gas-oil-ratios vary from 400-1700
scf/STBO. Some fields are initially at the saturation pressure while others are substantially undersaturated. Most reservoirs
in the WNS are developed using horizontal line-drive miscible-water alternating gas (MWAG) recovery schemes. In recent
years, the MWAG scheme has been altered to sub-miscible lean-gas alternating water (LWAG) recovery schemes. Though
called an LWAG process, the injected gas is still enriched to a degree with some mid-component constituents like propane
and butane. The switch to an LWAG process was driven by the installation of a condensate stabilizer column at the Alpine
Central Facility used to extract some of the condensates historically used to blend into a miscible injectant for shipment and
sale through the sales line. One of the fields undergoes waterflood only. Several gas condensate fields also exist in the area
and may have potential for gas cycling projects in the future. The oil and gas fields lying within WNS had an original-oil-in-
place volume in excess of 2.5 billion barrels of oil and 2.5 Tcf of gas.

The Alpine Central Facility is a single-train processing facility. Originally designed to handle 80,000 STBOD, the ACF has
undergone a number of expansion projects over the years to increase total oil and water handling capacity. Expansion and re-
design efforts have increased gas compression capability significantly over the years. Because gas compression efficiency is
significantly impacted by temperature and seasonal temperature variations on the North Slope of Alaska can by nearly 150○
F, seasonal variation in compression capacity may be significant as well.
4 SPE 158497

The only significant fluid which leaves the Alpine Central Facility is the sales oil. All other gas not used for fuel or lift must
be either blended for injection as MI (enriched LGI) or lean gas injected into two black-start gas injection wells. All
produced water must be re-injected and, for pipeline integrity reasons, must be segregated from imported make-up sea water
used for injection. Since the water injection manifolds at each drill site are limited to a single water source at any one time,
all wells at any particular drill site must inject the same water type; i.e., all wells inject produced water or all wells inject sea
water (except the CD1 drill site has 2 water injection manifolds and each may be on a different service at any one time).

With the ever-growing mix of fluid types, GORs, and water production, the first question an IAM must answer is how to
optimize production given the constraints on the surface network and production facilities and the large disparity in the
produced fluids. While this question may seem obvious at first, there are many details that come into play. First and
foremost, daily optimization of oil production can typically be achieved via standard gas lift optimization procedures under a
constrained volume of lift-gas. While this is true, it only considers the instantaneous optimization of production. If one were
to shut down all water and gas injection to one drill site, it is not hard to imagine how the production would rapidly decline as
the reservoir becomes depleted in that area, yet the daily oil production could surely still be optimized as before, maximizing
the current daily production. To fully optimize production then necessarily requires the injection fluid type and placement
also be considered; the second major question to be answered via an IAM.

Current development in the WNS is entirely within the environmentally sensitive area of the Colville River delta. This has
required the development to meet the strictest of environmental challenges. The Colville River Unit encompasses a little
more than 127,000 acres with the existing gravel footprint limited to 128 acres. No permanent roads connect Alpine back
across the Colville River to the east. Instead, ice roads are constructed during the winter across the Colville River to allow
the transportation of equipment and drilling supplies to the site. The ice roads are then simply allowed to melt in the spring,
minimizing the environmental impact. Additionally, the CD3 (Fiord) development 6 miles to the north of Alpine is a
roadless remote drill site with an ice road built each winter to allow the drill rig to travel to CD3, drill a few wells, then return
to Alpine before the ice road melts. What does all of this have to do with an integrated asset model? What this means is that
in some regards, planning is even more critical at Alpine than for an offshore platform where supplies can be delivered year
round. In addition to bringing all drilling supplies required for the year to Alpine during the ice-road season, all facility
equipment must be delivered during this 100-day window. Most construction, especially if it requires access to the tundra
must occur during the winter. In these regards, the IAM model becomes very important in identifying the maximum needs
for the drill sites significantly in advance of the winter transportation season. This becomes especially true when considering
the significant future development remaining in the existing producing fields and the potential development of new satellite
fields. For these reasons, the IAM model constructed must be capable of incorporating those expansions for development
scenario planning; the third critical element of the IAM.

So, the major needs for the WNS IAM include the capability to (i) optimize both daily and long term production, (ii)
optimize gas and water injection, and (iii) include a planning capability for all future reservoir development and facility
expansion.

Developing an Integrated Asset Model – Model Requirements and Tool Selection

Having identified the major process needs required in an IAM, other criteria determined important in selecting a software tool
to enable building an IAM were identified. While integration of the individual reservoir models with a pipeline surface
network model and facility model was deemed critical, it was recognized that being able to run the individual reservoir
models outside the IAM was still valuable. A single monolithic model would be cumbersome with different reservoir
engineers responsible for different fields and the fact that the individual fields are not always in-sync. Individual field
models, however, would still allow the individual reservoir engineers to optimize the development within each field.
Certainly, the individual models runs would be much quicker than within the IAM. Lastly, each field is still unique with its
own set of issues that may be better resolved and understood through more mechanistic type studies without the complexity
of a multifield run. So, the first criterion for selecting a tool was that it would need the ability to run the individual models
outside the IAM.

A number of criteria arise within the IAM tool itself. First, global facility limits on both production phases and injection
phases can only be honored with a truly integrated model. The IAM would need the ability to easily understand and visualize
model results. This means that the IAM tool must be capable of easily interrogating information from all models both
numerically and graphically. Ease of use or ease of construction was deemed critical if the IAM tool was to survive and
become a perpetual tool within WNS. In this regard, integration of updated individual field models back into the IAM tool
needed to be fairly straight-forward and require minimum intervention. Having the capability to incorporate operating
strategies unique to WNS and write our own control and optimization logic was also considered critical to realize the true
potential at WNS. Lastly, the IAM tool was viewed as being valuable only if run times were low enough to allow different
scenarios to be run quickly to enable management decisions in a timely manner. What this boils down to is that the reservoir
SPE 158497 5

simulator itself, being the most time consuming piece of the IAM, must be a parallel simulator to minimize the overall run-
time of the IAM.

As noted in the previous section, having the ability to do development scenario planning and timing was critical. The tool,
therefore, must have the capability to easily adjust drill schedules and field startup timing. Additionally, all controls on
production, injection, operational items, facility expansions, etc. must be able to be tied to a date within the IAM.

Though having pre-canned optimization routines available in the tool were considered a plus, the more important
consideration was that the tool provide the capability of incorporating our own optimization routines or building them within
the IAM tool for all aspects of the production and injection.

Since the composition of injection gas was to continually change either by design or simply due to the physical system, it was
important that the IAM tool be capable of passing compositional information between the models.

Cost of course must be considered in any decision. Items that must be considered include the cost of the integration software
itself, the cost of additional computing hardware capable of running all models simultaneously, the cost of the manpower
required to build the IAM as well as to maintain the IAM, the cost of training, and the impacts to productivity resulting from
the construction of the IAM.

The WNS team evaluated a number of options for an Integrated Asset Model. Both internal spreadsheet-based tools and
external commercial spreadsheet-based tools and reservoir simulation-based tools were evaluated against the criteria
established above. The spreadsheet tools evaluated were essentially post-simulation rollup tools based on production rate
versus cumulative production curves. The main issue with these tools was the very limited tie to injection. Since the
simulation models used to generate the production curves assume a priori injection volumes and schedules, this results in a
disconnect between the production and the actual injection profile that would occur during an integrated model run. These
tools were eliminated from consideration as simply not being rigorous enough.

The WNS reservoir simulation models vary in size from 6000-1,000,000 grid blocks. When the evaluation was conducted,
run times varied from 15 minutes to 2 days for a full 30 year predictive runs on a single-CPU for these models. A two-plus
day turnaround was considered unacceptable for an integrated model. Thus our in-house sequential simulator was eliminated
from consideration. Of the two remaining reservoir simulation-based options, one of the reservoir simulators exhibited
unimpressive parallel performance, the upfront cost was substantially higher than the other simulator being considered plus
there was a lack of a user base within the company for that particular piece of software, and the model was missing a number
of the tool criteria.

In late 2006, the decision was made to go with an IAM solution consisting of a commercial parallel reservoir simulator
(Eclipse1, 2011), surface network simulator (GAP2, 2010), facility simulator (Hysys3, 2007) and integration tool (Avocet
IAM1, 2011) that satisfied nearly all of our criteria for assembling an integrated asset model. For IAM performance reasons,
a decision has recently been made to swap out the current surface network for a different one (PIPESIM1, 2011). This
combination of tools / products met almost all of our preset selection criteria.

One of the most powerful aspects of the selected integration tool is its open communication design and its links to Excel™
base spreadsheets. This is perhaps one of the most powerful utilities within the software in that it enables one to control
almost anything and everything one has a wish to control within the modeling process.

This integration tool is an end-to-end software solution that integrates reservoir, wells, surface infrastructure, and process
facilities as well as the asset's operating parameters, financial metrics, and economic conditions into a single production and
reservoir management environment. The application's synergies deliver end-to-end improvements in asset management.
Production systems can be designed and scaled properly for dynamic reservoir conditions. Field development strategies can
be formulated with consideration for constraints in the process facilities. The application allows domain specialists,
regardless of location, to contribute models specific to their disciplines with the results propagated throughout the end-to-end
integrated model. They work concurrently to identify and resolve conflicts that could add cost or impair productivity.
Collaboratively, they exploit synergies that add efficiency, boost output, and extend the economic life of the asset. With
integrated models, it is possible to access information that reduces cross-disciplinary uncertainties and mitigates the need for
corrective reactions. The tool is truly a solutions enabler.

1
Avocet, ECLIPSE and PIPESIM are trademarks of Schlumberger
2
GAP is a trademark of Petroleum Experts Ltd
3
HYSYS is a trademark of Aspen Technology
6 SPE 158497

Lastly, a 16-node, 32-CPU (dual-core Opteron) Linux computing cluster was purchased to enable optimum performance of
the IAM. The computing cluster has since been upgraded to quad-core Intel Nehalem CPUs providing 128 compute nodes
for reservoir simulation processes. During an IAM run, the IAM controller, the surface pipeline model, and the process
facility model programs are typically run on the user’s PC and the reservoir simulation models are run on the Linux cluster.
This is completely arbitrary, however, as the IAM controller has the capability to submit any model to any computer the user
has access to within his/her network; e.g., the user could choose to run the IAM control on his/her own PC, run the pipeline
model on another PC, the facility model on yet another PC, and each of the reservoir simulation models could be submitted to
different computers. That wouldn’t necessarily be an efficient way to do things, but it is possible within the IAM.

Building and Assembling the Integrated Asset Model


Having identified reservoir management and modeling needs for an IAM and having selected the tools to be used in
assembling the IAM, now comes the task of actually designing and building the model. Several key pieces of advice here
will make the task much easier and reduce potential problems in the future as the tool is built out. First, as the common
saying goes, “a chain is only as strong as it’s weakest link,” each piece of the integrated model should be built and tested to
ensure it is capable of accurately modeling the particular link in the model. As most reservoir engineers are aware, having a
model fail after 20 years of a predictive run because of an oversight or poorly tuned model is no fun and may be difficult to
find the issue. With an integrated asset model, multiply that difficulty by at least an order of magnitude simply due to the
number of moving parts. It is much easier to do the due diligence on the individual models up front, rather than once they
have been brought into the IAM.

Reservoir simulation models should be history matched and up to date. While sounding like a reasonable goal and
assumption, with many reservoirs models to be integrated into the IAM and each reservoir with their own reservoir engineer
and development schedule, it is more than likely that the simulation models will be out of sync with one another at any one
time in terms of the level of and completeness of the simulation history match. While the assembled tool may still be useful
for identifying certain potential bottlenecks or the potential to partially optimize performance, incorporating unmatched
simulation models into the IAM will lower the confidence level in the IAM and the model’s utility as a trusted and timely
reservoir management tool. Wellbore hydraulics tables used within the reservoir simulation models must be tuned and
interrogated as much as possible to not only accurately model the historical performance of the wells, but must be designed to
cover the entire expected production and performance range during the predictive phase. Surface network models similarly
must be tuned and matched to historical production and performance. Ties to future satellite production must be thoroughly
interrogated to ensure results make sense. The complexity of process facility models can be daunting for even the most
seasoned engineer to understand the interaction of all the moving parts in the model. Full-blown facility models can also be
the most difficult models in the entire IAM to achieve stable convergence over the entire range of production scenarios that
may occur over the life of the field.

Having tried to emphasize the need to be diligent in creating, tuning, and history matching each of the disparate models being
assembled in the IAM, emphasis also needs to be placed on the selection of the simplest tool or model available to answer the
questions to be asked. We will discuss this further as we outline the IAM developed for WNS.

The Western North Slope Integrated Asset Model

Figure 2 is a generic schematic of the entire production and injection process at WNS. A number of reservoirs are produced
from one or more drill pads. Production (oil, water and gas) from each field is sent through multi-phase pipelines back to the
Alpine Central Facility (ACF) for processing. Note that the gas from each drill site includes lift gas produced back up the
well. At the ACF, the phases are processed and separated. The gas phase is further processed separating condensates out of
the gas stream. The condensates stream is then passed through a condensate stabilization column and a fraction of the
stabilized condensates are pulled and blended with the crude oil stream to be sent down the sales line. Fuel gas and a minor
amount of sales gas are pulled from the lean gas stream. A small amount of the lean gas is diverted to two black-start gas
injection wells. An optimization is then performed on the remaining lean gas to determine the optimum amount of lift gas for
the current operating conditions. The remaining lean gas is then blended with the remaining gas condensates to form an
enriched-gas for re-injection. Optionally, a volume of lean gas to be blended with the condensate stream is determined such
that the resultant gas will be miscible at reservoir conditions for most of the reservoirs in the development. Any left over lean
gas is then distributed to a number of lean gas injection wells either for gas storage or as part of a lean-gas alternating water
recovery scheme. Produced water is treated and necessary injection pump capacity assigned. The remaining injection pump
capacity is assigned to imported sea water used as the makeup or swing fluid. Additional details on various aspects of these
processes will be covered in the discussion below.

Reservoir Simulation Models


All reservoir simulation models for the WNS fields are fully-compositional. To keep run times reasonable while capturing
all important aspects of the recovery scheme, fluid descriptions within the models are captured with tuned 8-component
SPE 158497 7

equations-of-state (EOS). All models employ the Peneloux (1982) modification to the Peng-Robinson (1976) equation-of-
state. Without exception, the eight component description for each field is represented using 4 pseudo-components for the
C7+ fractions. Note that where the 4 pseudo-components are split may differ for each field. The four “light” pseudo-
components for each field are comprised of the same pseudo-components in that the same pure species are combined to
create the pseudo-components; i.e., C1-N2, C2-CO2, C3-C4, and C5-C6. Note here, however, that the critical properties and
binary interaction coefficients for those components will be different for each field. Also note that there are only minor
amounts of nitrogen and carbon dioxide in any of the WNS fields.

Figure 2 – Schematic of production and injection process. Figure 3 – Horizontal line-drive patterns in WNS.

This commonality to the EOS description provides a significant benefit and simplification when it comes to integrated
modeling. The question naturally arises as to how to model the combined fluids within the process facility model. Because
of the commonality in the light pseudo-components, it becomes straight-forward to simply split the pseudo-components into
their pure components when passed into the facility model. Since the “heavy” pseudo-components for each field may differ
significantly in both their properties and split, they must be carried forward as they are. The net effect of this split philosophy
is a fluid description comprised of eight pure “light” components common to all fields and four “heavy” pseudo-components
for each field. The sum of all this is a 37-component (including water) fluid description for the facility model (soon to
expand to a 45-component description).

A secondary but also extremely beneficial aspect of this component splitting strategy arises when considering the injected
gas. The gas-condensate stabilizer is generally operated to split the condensate stream nominally between normal- and iso-
butane. This results in a condensate stream, and ultimately a blended gas injection stream, which is comprised entirely of the
“light” components (mostly butanes and lighter with just a trace of pentane). Because of the commonality of the light
components, this then allows one to simply re-lump the pure components back into the light pseudo-components to define the
injection gas composition for each reservoir simulation model. The composition of the lift gas (lean gas) is also traced such
that the proper gas composition is applied to the lift gas of each reservoir model when it comes back into the plant.

The individual reservoir simulation models vary in size from approx. 6000 gridblocks to over 1,000,000 grid blocks with as
few as 8 wells in one model and a maximum of 175 wells in another. All reservoirs are developed predominately with
horizontal wells undergoing a line-drive MWAG (or LWAG) EOR process with one shallow reservoir being waterflooded
only (See Fig. 3). To be efficient in the numerical solution of the reservoirs and minimize the run time of the simulation
model solves, every attempt has been made to equalize the run times of each model. This results in some models being
solved on a single CPU, whereas others are parallelized to run on as many as 8 CPUs. The parallel models have had their
domains optimized to obtain as close to equal loads as possible on each CPU to optimize the run time of that model.
Combined, 27 CPUs are utilized by the simulation models on a Linux cluster. To further minimize run times for the IAM,
restarts are used in the reservoir simulation models 2 years prior to the predictive phase. This 2-year window is used to
verify historical matching of the IAM prior to moving into the predictive phase.

Depending on the overall production and injection scheme set at the start of the run, the simulation models run either under
rate control with all of the common constraints on bottomhole pressures, economic limits, etc., or the wells are all run under
tubing head pressure control with constraints on field-level production controls and gas lift availability. Within the
simulation model itself, gas lift optimization is performed once at the start of each month when the wells are under tubing-
head pressure control. Gas lift is assumed to be constant for each well during the rest of the month. Typical constraints on
8 SPE 158497

the gas lift include minimum lift per well, maximum lift per well and minimum economic gradient (bbls oil/mscf lift gas).
Gas lift optimization will be discussed a little more in the optimization / allocation section.

Each month, each simulation model passes its well, group and field production and injection rates, operating pressures, gas
lift usage, gas lift gradient, and well status to the IAM controller.

Because no reservoir simulation model can possibly contain every potential future well and because some potential future
developments may not have a simulation model built yet, spreadsheets have been incorporated within the IAM to contain
complete rate profiles (both production and injection) for those items not captured within the reservoir simulation models.
The spreadsheet rate streams are treated within the IAM model just as if they had come from a simulation model and must
include compositional information for feed into the facility model. Note also that this design provides the capability, should
one choose to use it that way, to incorporate surrogate reservoir models (material balance, decline curves, response surfaces,
neural nets, etc.) directly within the IAM and could, again if so chosen, completely bypass the reservoir simulation models.

Surface Pipeline Network


To account for the pressure drop through the surface pipeline network and its impact on the production at each drill pad, a
surface pipeline network model has been integrated into the WNS integrated asset model (see Fig. 4). This surface network
model represents a simplified model as compared to the detailed model tied into the WNS SCADA system used for daily
production and gas lift optimization. Because the goal of the WNS IAM is the long-term optimization of rate and recovery,
we are not so concerned with the detailed prediction of pressure losses down to very tight pressure tolerances. Of more
importance is to capture longer term (monthly) pressure trends at each drill site or each production manifold. Figure 5 shows
example manifold pressures over time for several drill sites. Of note are two items: First, the manifold pressures may vary by
as much as 200 psi over the life of the field. Second, the month-to-month variations are generally small, increasing
consistently during heavy drilling activity, followed by a long fairly stable pressure history, and lastly, a decreasing manifold
pressure as field production declines.

Figure 4 – Simplified WNS Pipeline Network. Figure 5 – Calculated Manifold Pressures.

Because all production wells at a drill site produce into a common manifold, they share the same wellhead pressure
(assuming they are not choked upstream of the manifold). The surface network is designed with a common delivery pressure
at the Alpine Central Facility. Supplying the network with the drill site overall produced oil, gas and water rates as well as
the gas lift rates, the network is back solved to determine the manifold pressure at each drill site. These pressures are then
sent down to the reservoir simulation models as wellhead pressure constraints.

Process Facility Model – Alpine Central Facility


The primary purpose of including a process facility model in the WNS IAM is not to impose the fluid processing constraints
upon the total fluid and phase production; these are easily handled elsewhere within the WNS IAM. Rather, it is primarily to
capture the impacts of extracting condensates out of the gas stream for shipment down the sales line and the resulting impact
of having less condensate to blend in the gas stream for re-injection into the reservoir for EOR benefits.

Figure 6 is a schematic of the Alpine Central Facility process model. What should be obvious to any process engineer is the
simplicity of the facility model being used. This is by design. Again, use the simplest model you can to answer the questions
you need answered. The full-blown Alpine facility model is an extremely complex and comprehensive model incorporating
every aspect of the facility including re-cycle loops, heat exchangers, chokes, valves, chillers, heaters, etc. and solves
comprehensive energy and material balance equations. The comprehensive ACF model also requires 45 minutes to solve one
SPE 158497 9

steady-state system. That was considered a non-starter for a tool to be used to make timely reservoir management decisions.
Instead, a greatly simplified model which captures the essence of the facility process was created.

As shown in Fig. 6, the main aspects of the model are the feed, the separation and stabilization, and the outlet streams. Feed
streams for the process model are the composition (split into the facility 37-component description) and surface volume of oil
output by each reservoir model. The process model begins by back-calculating through a simplified 3-stage separation flash
process the total molar feed rate from each model. The entire feed stream (sum of all feed streams) including the molar rate
of gas lift gas used by each model are then fed into the conventional separation train which includes conventional separators
and gas condensate knock-out drums. Though it shouldn’t require stating, we do state that all process equipment is operated
at exactly the same conditions across all models; e.g., stage-1 separator temperature is the same in the reservoir simulation
models as in the process model. Condensates knocked out of the gas stream become the feed-stock for the condensate
stabilizer column. Though numerous controls exist for specifying the stabilizer operating conditions, the primary ones used
to match historical performance are the reboiler-temperature, reflux-rate, and the temperature of tray-25.

To understand the importance of capturing the correct condensate volumes and composition of the blended miscible-injectant
(MI) or enriched-gas injectant, we refer the reader to Figures 7 and 8. First, Fig. 7 shows the impact of the condensate
stabilizer temperature on the calculated minimum-miscibility pressure (MMP) of the blend gas injected over time. Two items
are of particular importance here. As the stabilizer temperature is increased, driving more condensates off the top of the
stabilizer to be blended with the lean gas stream to form the injection gas, the MMP of the injection drops. The second item
of note is that no matter what temperature the stabilizer is operated at, the MMP of the blended MI stream can vary by
hundreds of psi. Reasons for this variation include seasonal variations in the gas handling capacity (compressor efficiency)
discussed previously, the optimal use of the available gas compression for gas lift and produced gas handling (which field or
well that gas comes from under optimal production), or simply due to increased production at one field or another due to
bringing on-line newly drilled wells. Figure 7 would suggest that the MMP of the injection gas may vary by close to 1000
psi over the course of the WNS development.

Figure 6 –Simplified Alpine Central Facility Model in WNS IAM.

The importance of the MMP of the injection gas to WNS should be apparent when one refers to Figure 8. Here, the oil
recovery, normalized to current operating conditions at WNS, is plotted against the MMP for one of the fields in the WNS
IAM. For this particular field, one sees that the recovery could drop by as much as 6% if the gas injection stream is leaned
out too much by stripping the condensates out of the gas stream and shipping them down the sales line. Recall earlier that we
stated that the OOIP in WNS was estimated to be over 2.5 billion barrels. For any reasonable recovery factor, if that 6%
applied across the board, it would add up to more than an insignificant amount of oil. Of course, the best use of the
condensates (sales or enhanced recovery) is driven by economics of both options and is continually evaluated.

An additional constraint on the volume of condensates that can be stripped out of the gas stream for sale is the vapor-pressure
10 SPE 158497

specification for the sales line. While this is actually a constraint on the vapor pressure of the Trans-Alaska Pipeline (TAPS),
that vapor pressure specification propagates all the way back to the major production centers on the North Slope of Alaska.

IAM Controller, Production and Injection Optimization, and Reservoir Management Controls
The primary goal of this section is to cover the production and injection allocation calculations, the optimization techniques
utilized within those calculations, and the operational and reservoir management constraints and considerations incorporated
into the model.

Figure 7 – Impact of Stabilizer Temperature on MMP. Figure 8 – Impact of MMP on Recovery.

The heart and soul of the WNS IAM must be the IAM controller (AVOCET IAM®, 2011). It is this IAM controller that
enables the seamless integration of so many models and allows the WNS group to design and build such a comprehensive
and flexible reservoir management tool. Figure 9 displays the complete integrated model and some of the flow connections
between the various pieces of the IAM. Other than the reservoir simulation models, pipeline model, and facility model, the
reader should note a variety of Excel™ spreadsheets incorporated within the model. These spreadsheets contain much of the
control logic for optimizing production and injection as well as the operational and reservoir management controls. In
keeping with our philosophy of building a model that will be easily maintainable and not require a programming expert to
enhance, the vast majority of allocation methods have been incorporated directly using Excel™ cell computations. Only two
rather small Visual Basic routines have been incorporated into the model for the purpose of gas lift optimization and water
injection optimization. In so doing, most any reservoir engineer with a familiarity with Excel™ should be able to; (i)
understand the computations and allocation methods being used, and (ii) feel confident that he/she can easily modify the IAM
as the need arises in the future.

Figure 9 – WNS Integrated Asset Model and Component Connections.


SPE 158497 11

The first step in setting up the IAM is to establish the constraints to be incorporated in any allocation scheme. All drill pad
limitations on production and injection are specified. Maximum facility limits are established and any fixed additions or
subtractions to those limits established; e.g., facility fuel gas requirements, gas sales volume, minimum black-start gas
injection volumes, and maximum allowed lift gas. Monthly gas compression efficiencies are input; recall that due to seasonal
variations in temperature, the compressor efficiencies can vary by as much as 15% over the course of the year. Water
injection pump minimum water requirements and maximum capacities are set. Minimum and maximum gas lift usage per
well are set for each reservoir / drill pad. Water injection efficiencies are established and set for each reservoir; i.e., the
injectivity of wells injecting produced water has been seen to be approximately 80% that of wells injecting sea water. This
loss of injectivity is recovered when wells are converted back to sea water injection. Optimization of water injection
considers the impact of produced water on well injectivity and will be covered in more detail shortly. Facility expansion
options are established and dates for their completion set; e.g., add compression or water handling at some future date.
Planned and unplanned facility and export line downtimes as well as downtime for pigging are established. These result in a
facility uptime for each month. WNS imports sea water for injection from the Kuparuk River Field Seawater Treatment
Plant (STP) and so must incorporate uptime for the STP. Facility water separation efficiency is established for the plant and
allowed to vary by date. Well uptimes are established and set by drill pad. Generally, different drill pads represent wells of
different ages and so carry different well uptime estimates. Certain situations such as the Fiord wells at drill pad CD3 require
additional treatment due to the remote access nature of the site. Recall that the CD3 is a remote drill site accessible by road
only in the winter after an ice road has been built. This means that if a well goes down during the remaining 8 months of the
year and the well requires a drill rig or coiled tubing unit to bring it back on-line, it may be down for quite some time. Start-
up dates for future satellite fields are entered. A drill schedule incorporating all future wells to be drilled is incorporated.
This drill schedule can be varied as necessary to model differing future development scenarios. If a well is not to be drilled,
one merely sets the drill date beyond the life of the model run. Initial WAG injection schedules are set for all present and
future WNS injection wells and includes a specification of miscible (or enriched) gas or lean gas injection. A target MMP for
the miscible injectant or a blend ratio between condensate and lean gas is specified as a function of date. Most of these
constraints are utilized during any one IAM run.

In developing the WNS IAM, a design philosophy of starting simple and building in complexity with time was employed to
minimize potential problems and allow for incremental testing and evaluation of the model. This has resulted in a number of
different allocation schemes for production and injection as well as simpler incorporation and testing of each of the
constraints and operational and planning items noted above. A second design philosophy incorporated into the WNS IAM
was that more involved allocation methods should be as close to identical as those used within the reservoir simulation
models. If done properly, the optimal group level or field level production and injection rate allocations calculated at the
IAM level (whether using field level or rolled up well level allocations) will be all that need be passed back down to the
reservoir simulation models for the next time step. Having used similar allocation schemes as the reservoir simulation
models will hopefully result in the simulation models allocating well level production and injection matching that estimated
at the IAM level. We should note that in addition to the group or field level production and injection rate targets, injection
gas compositions (both lean and enriched), gas lift volumes and the composition of the gas lift gas, and well head pressures
(if wells are under tubing-head pressure control) are also passed down to each reservoir simulator.

Variables obtained from the reservoir simulators for conducting the rate allocation calculations within the IAM at the end of
each synchronization time step include: field surface production and injection phase rates, group surface production and
injection phase rates, well surface production and injection phase rates, reservoir conditions field, group and well voidage
rates, surface compositional information for each produced hydrocarbon phase, bottomhole pressures for all wells, well-grid
reservoir pressures, gas lift volumes for all wells and the field, current operating gas lift gradients for all wells, current well
status (open or shut), current well type (producer, water injector, gas injector), well-level oil production potential, well-level
water and gas injection potentials, and cumulative production and injection phase volumes. The Avocet IAM control
seamlessly handles the transfer of these variables from the simulators to the IAM model.

Once drill pad level allocated phase rates have been estimated, they are passed through the IAM controller to the pipeline
model to back calculate the production manifold pressures. These manifold pressures are then passed back to the IAM
controller to be passed down to the individual reservoir simulation models as well head pressure targets.

As previously mentioned, the field-level surface oil rates and oil compositions from each simulation model (or spreadsheet
model) are passed through the IAM controller to the facility model. The facility model then passes the lean-gas composition
and the stabilizer tops (blend-gas) composition and volume back to the IAM controller for computing the gas balance and the
composition of the gas injection streams for the next synchronization time step.

Production and Injection Allocation – Constant Rate


The simplest of the allocation methods incorporated in the WNS IAM for both production and injection are fixed constant
rate targets with no facility constraints. This method is very useful for testing options across the spectrum within the IAM
12 SPE 158497

without the burden of the facility constraints or other optimization routines which may cloud whether or not the option has
been implemented properly.

Qβ ,i = Ci , β = o, g , w, gi, wi ……………………………………. (1)

The second allocation method implemented across both production and injection is a constant rate scaled to meet facility
limits. This option is very useful for the test of the facility constraints and facility and well up-times.

N
Qβ ,i = Qβ , FacLim * FU * Ci / ∑ Ci , β = o, g , w, gi, wi ..…………………… (2)
i =1

Production Allocation - Oil Rate Constrained


Three additional oil rate allocation methods have been incorporated in the WNS IAM and include from simple to complex:

Method 3:”Field” Level Guide Rates Allocated to meet Facility Oil Limits
Guide rate calculations for production allocation are well known and this particular form is covered in detail in the user’s
manual for the reservoir simulators (ECLIPSE™, 2011). The coefficients a, b, c, d, e, and f are arbitrary input coefficients
used by the user to penalize production of high GOR or WOR wells or provide additional priority to high rate oil producers.
The same coefficients are used within the reservoir simulation models as are used within the WNS IAM. Here, the Ci
represents the maximum production constraint for the well group and Poti is the oil potential of the well-group. Once the
guide factor has been calculated for each group, it is used to allocate the well-group target production rate (Eq. 4).

G o ,i =
[min(Ci , Pot i )]a ………………………….…………………. (3)
b + c * GORid + e * WORi f

n
Qo ,i = Qo , FacLim * Go ,i / ∑ Go ,i ……………………………………………….. (4)
i =1

Method 4:”Well” Level Guide Rates Allocated to meet Facility Oil Limits
Method 4 is directly analogous to Method 3, with the exception that the guide rate calculation is taken down to the well level;
i.e., all production parameters in Eqn. 5 represent the individual well values. The guide-rate factors are then summed for
each well-group and applied to determine the oil production allocation target for that well-group (Eq. 7).

go, j =
[min(C , Pot )]
j j
a

…………………………………………….. (5)
b + c * GOR + e * WOR jf
d
j
n
Go ,i = ∑ g o , j ………………………………………………………….. (6)
j =1
n
Qo ,i = Qo , FacLim * Go ,i / ∑ Go ,i …………………………..……………………….. (7)
i =1

Method 5: Gas Lift Optimization Across all Wells to optimize Oil Rate
Within the IAM allocation routines, gas-lift optimization calculations have been constructed to mimic the gas-lift
optimization method used with the reservoir simulator. With one exception, the application of the gas lift optimization
calculation method is identical to that described in detail within the simulator manual (ECLIPSE, 2011) and is not repeated
here. The exception to the methodology concerns how the gas lift gradient is calculated. Referring to Figure 10, suppose
point C is the current operating condition for a well. The gas lift gradient (slope of the curve) is calculated within the
reservoir simulator at point C by interpolating the wellbore hydraulics curves at that operating point. The simulator then adds
an increment of gas lift to the well and calculates at new gradient at point P, again, by interpolating the wellbore hydraulics
table. Lastly, the simulator will take away an increment of gas lift from the well and calculate the lift gradient at point M.
The gas lift gradient is calculated as
SPE 158497 13

ΔQo
GGL = …………………………………………………………. (8)
ΔQGL + ΔQ g

The gradients for all wells are calculated at each point, C, P and M. The well (call it Well A) with the highest gradient at
point P has its gradient from point C to P compared to the gradient from point M to C for the well with the lowest gradient at
point M (call it Well B). If the gradient at P for Well A is greater than the gradient at point M for Well B, an increment of
gas lift is taken from Well B and given to Well A. The gradients for both wells are then recomputed and resorted.

Figure 10 – Gas Lift Optimization Gradient Calculation Figure 11 – Injection Optimization Gradient Calculation

The gas lift optimization procedure within the WNS integrated asset model is identical. The only difference comes in how
the gradients are calculated at the points M and P. In order to compute the gas lift optimization identically as the reservoir
simulators, the WNS IAM would need to incorporate the wellbore hydraulics tables within its calculation routines. We have
chosen instead to use a first-order approximation under the assumption that if we accept that the reservoir simulator does a
pretty good job of optimizing the gas lift allocation, then the current operating condition of the well should be pretty close to
optimum. The reservoir simulator does export the gas lift gradient at the current operating conditions of each well (point C
on Fig. 10) and those gradients are imported into the IAM controller each synchronization time step. From there, given an
incremental volume of gas lift, ΔQGL, the incremental volume of oil and gas that will be produced is calculated as

ΔQo = GGL ∗ ΔQGL ……………………………………………………. (9)


and
ΔQ g = ΔQo * GOR …………………..……………………………… (10)
An incremental lift gradient is then calculated as

ΔQo
dGGL = ………………………………………………. (11)
ΔQGL + ΔQ g
Lift gradients at points P and M are subsequently estimated as, respectively:

GGL (P ) = GGL (C ) − dGGL ……………………………………………. (11)


and
GGL (M ) = GGL (C ) + dGGL ……………………………………………. (12)

Utilizing Method 5 while insuring all well, group, and field constraints are not violated will insure that the oil production rate
is maximized at any instant in time.

Production Allocation - Gas Rate Constrained Oil Target


Here, the maximum oil rate achievable when operating under a gas constraint is determined. First, the gas rate to use needs
to be determined and two options are provided.
14 SPE 158497

The first method simply uses the uptime scaled current operating group producing GOR:

Q g ,i
GORi = PGOR = ……………………………………………. (13)
Qo , i
The second method uses a rolled up group total GOR calculated as

∑q
j =1
gT , j

GORi = TGORi = n
……………………………………………. (14)
∑q
j =1
o, j

where

q gT , j = q g , j + qGL , j ……………………………………………. (15)


and

qGL, j = min[qo, j * max(GLRj ,t arget − GLRj ,0), qg , LiftMax] ………..…………. (16)

An allocation factor based on either an oil target rate, Ci, or oil potential rate, Poti, for each well-group can then be calculated
as

af i
AFi = n
…………………………………………………. (17)
∑ af
i =1
i

where

min (C i , Pot i )
af i = ………………………………………………. (18)
GORi
The facility gas handling constrained oil target for each well-group then falls out directly as

Q g , FacLim * AF
Qo , i = …………………………………………………. (19)
GORi
The well-group produced gas limit based on the current producing GOR follows as

Q g ,i = Qo ,i * PGORi …………………………………………………. (20)

Given the produced gas limits, the required, excess and total available gas lift limits can be determined as follows. The total
gas lift gas available is

N
Q g , LiftTotal = Q g , FacLim − ∑ Q g ,i ………………………………………. (21)
i =1
The required gas lift follows as

N
Q g , Lift Re quired = ∑ Qo ,i * (TGORi − PGORi ) ……………………………. (22)
i =1
and the excess gas lift available (actually, excess gas compression available) is

Q g , LiftExcess = Q g , LiftTotal − Q g , Lift Re quired …………………………………. (23)


SPE 158497 15

This excess gas handling is then reallocated by passing back through Eqns. 14-23 until the excess goes to zero.

Production Allocation - Water Rate Constrained Oil Target


Computation of a water constrained oil target rate for each well-group follows a similar calculation methodology as for the
gas. In this case, a partial guide rate calculation is used where

G o ,i =
[min(Ci , Pot i )] ………………………………………………. (23)
b + e * WORi f
and the allocation factor is

G o ,i
AFi = n
…………………………………………………. (24)
∑G
i =1
o ,i

The allocated oil rate under a water constraint is then calculated as

Qw, FacLim * AF
Qo , i = …………………………………………………. (25)
WORi

with a water limit of

Qw,i = Qo ,i *WORi …………………………………………………. (26)

Water Injection Allocation


As noted previously, all produced water within the Colville River Unit on the Western North Slope must be re-injected. For
pipeline integrity reasons, the produced water must be segregated from sea water imported for injection. On top of that, well
injectivity shows a marked decline when under produced water injection, though the injectivity recovers when switched back
to sea water. At any one time, optimization of water injection then requires a decision on which drill sites (or individual
water injection manifolds) will be on produced water injection or on sea water injection. To further complicate matters, there
are a fixed number of water injection pumps to handle both the sea water and produced water.

Since all produced water must be re-injected then requires that injection pump priority first go to produced water. This
results in some inefficiency as the pump capacity may exceed the volume of produced water to be re-injected. The capacity
of the remaining pumps establishes the maximum of sea water that can be imported and injected.

As noted above, two simple constant injection allocation methods (Eqns. 1-2) have been implemented to enable testing. The
constant allocation methods for injection apply at the facility, or to each water injection manifold. Note that Eqn. 2 may
further be applied independently for those manifolds on produced water injection and those on sea water injection.

Method 3: Allocate Pad Injection Targets Proportional to Pad Reservoir Voidage Rates
For this method, water injection is allocated to each water injection manifold proportional to the reservoir voidage of the
production wells offsetting the injection wells for that water injection manifold. Again, this calculation is done separately
across those manifolds on sea water and those on produced water. This method is commonly referred to as voidage
replacement.

Qwi ,i = Qwi , FacLim, β * Qvoid ,i / ∑ Qvoid ,i ……………………..…………………… (27)


i =1
Where the reservoir voidage is defined as

Qvoid ,i = Qo ,i * Bo + Qw,i * Bw + Qo,i * ( PGORi − Rs ,i ) * Bg ………………… (28)


16 SPE 158497

Method 4: Allocate Pad Injection Targets Proportional to Pad Reservoir Voidage Efficiency Rates
Here, the oil cut of the offset production is considered in the allocation of the water injection. We begin by defining voidage
efficiency as

Qo,i
FVE ,i = ………………………………………………… (29)
Qvoid ,i
where the reservoir voidage is given by Eqn. 28. An allocation factor for each pad is then defined as

FVE ,i
AFi ,α = Nβ , α = pw, sw …………………………………… (30)
∑F
i =1
VE ,i

and the pad injection allocation is

Qwi ,i = Qwi , FacLim,α * AFi ,α ……………………..…………………… (31)

Here, Qwi,FacLim,α is the available produced or sea water. Note that if the calculated water injection for a particular pad
exceeds its maximum, it is set equal to its maximum, and the excess is reallocated to the remaining injection manifolds.
Equations 29-31, then are iteratively solved until all constraints are satisfied.

The use of the voidage efficiency defined by Eqn. 29 is a quasi-optimization of the benefit of the injection on oil production.
Inherent in the assumption is that a reservoir barrel of injection will result in a reservoir barrel of production. One may also
draw an analogy to the gas lift optimization method presented earlier. Figure 11 is an example of how injection (whether it
be straight water injection, gas injection or MWAG injection) may impact oil production across wells, groups or fields. For
the moment, let us assume that each curve represents a single well, regardless of which field the well belongs. The different
wells may be at different levels of maturity as marked by the solid-line outlined stars on the plot. With production and
injection, a particular well will move from the solid outline star to the dotted outline star; i.e., its efficiency will change. The
efficiency of the injection for any well (the production benefit) is defined as the slope of the curve at the current time, or

ΔN po,i
FVE ,i = ……………………..…………………… (32)
ΔN inj ,i

For any short period of injection and production, we have

ΔN po,i = Qo,i * Δt ……………………..…………………… (33)


and

ΔN inj ,i = Qinj ,i * Δt ……………………..…………………… (34)

Substituting Eqns. 33 and 34 into 32 we obtain

Qo,i
FVE ,i = ……………………..…………………… (35)
Qinj ,i

Note in Equation 35 that Qo,i is in STBO and Qinj,i is in reservoir barrels. If one further assumes that a reservoir barrel of
injection will result in a reservoir barrel of production (100% pure voidage replacement), then we have

Qinj ,i = Qvoid ,i = Qo,i * Bo + Qw,i * Bw + Qo,i * ( PGORi − Rs ,i ) * Bg ………… (36)


SPE 158497 17

Substituting Eqn. 36 into Eqn. 35, one obtains Eq. 29. In essence, the methodology here is akin to the gas lift optimization in
that the injection is first allocated to the offset production with the highest gradient of produced oil per barrel of injectant.

So far, allocation of the water injection once the manifold water service (produced or sea water) has been determined has
been covered. How that decision is made, however, has not.

Injection Manifold Water Service

The decision on water service to each water injection manifold is a simple optimization problem where we wish to maximize
the benefit of placing each type of water to each pad subject to the constraints that all produced water must be injected, yet
the maximum water at each manifold cannot exceed the capacity of the wells at that pad to inject the water. Fundamentally,
the optimization problem is

N man

Maximize: Qo,T = ∑F
n =1
o,n [ N pw,n * E wi ,n * Q pw,n + (1 − N pw,n ) * Qsw,n ] …………… (37)

subject to
N pad

∑N
n =1
pw,n * E wi ,n * Q pw,n = Qwp,T …………………………………… (38)

and
Qα ,i ≤ Qmax,i , α = pw, sw …………………………………… (39)

Where the water service Npw,i is in the subset of {0,1}. The injection efficiency, Ewi,n, is 1.0 for sea water injection, and the
empirically determine injectivity reduction for produced water injection.

The optimization problem posed by Eqns. 37-39, is often termed a combinatorial optimization or Knapsack problem
(Mathews, 1897). The “Traveling Salesman” problem also falls into this category of optimization problems. In essence,
combinatorial optimization is a simple method of finding an optimal object from a finite set of objects. For the case at hand,
there is a finite set or combinations of injection manifolds that may be on produced water service or on sea water service
while still satisfying the constraints. The optimization piece is simply finding the combination that results in the maximum
oil benefit. For the WNS problem, the number of possible combinations of water service is limited by the various
combinations of a limited number of water injection manifolds. The optimization problem is then solved merely by looping
over all possible combinations, discarding any case that does not satisfy the constraints (non-feasible solutions), and selecting
the set of Npw,n that result in the maximum oil rate calculated from Eqn. 37.

Gas Injection Allocation


Allocation of gas injection follows closely to that for water. As noted above, the first two methods of allocating gas injection
are a fixed constant injection rate without facility constraints and a constant rate scaled to satisfy facility constraints; the
facility constraint here being the total available compression.

Once fuel gas requirements, minor sales gas, flare, lift gas, black-start injection gas and stabilizer condensates withdrawn for
sales have been subtracted from the total compression, one is left with the total volume of gas available for injection (both
lean gas and enriched gas). The volume of lean gas and condensates available for blending are known. To determine the
volume of enriched gas and lean gas available for injection, the WNS IAM currently uses a target minimum miscibility
pressure (MMP) for the injected gas. Given the MMP of the lean gas (MMPLG) and the MMP of the enriching fluid
(stabilizer tops, MMPEF) estimated from an empirical equation of the form

MMP = a + b ⋅ yC1 + c ⋅ y N 2 + d ⋅ yC 2 + e ⋅ yCO 2 + f ⋅ yC 3 + " + z ⋅ yCN ……… (40)


the target blend ratio of the lean gas can be determined directly as

( MMPT arg et − MMPEF )


RLG = ………………………………..…… (41)
( MMPLG − MMPEF )
18 SPE 158497

Currently, any remaining lean gas is injected into the black start wells. Given the ratio of the lean gas to blend and the
amount of enriching fluids, one then knows the volume of enriched gas available for injection; i.e.,

Qgi ,T = RLG * QLG + QEF …………………………………..…… (41)

Method 3: Allocate Pad Injection Targets Proportional to Pad Reservoir Voidage Rates
The third gas injection method then follows that of the water injection; i.e., pure voidage replacement allocation for each gas
injection manifold then follows as:


Qgi ,i = Qgi ,T * Qvoid ,i / ∑ Qvoid ,i ……………………..…………………… (42)
i =1
where the voidage, Qvoid,i, is calculated from Eqn. 28.

Method 4: Allocate Pad Injection Targets Proportional to Pad Reservoir Voidage Efficiency Rates
This method also follows that of the water injection with the oil efficiency and the pad or manifold allocation factors being
defined by Eqns. 29 and 30, respectively. The gas injection allocation for each injection manifold follows as

Qgi ,i = Qgi ,T * AFi …………………………..…………………… (43)

This completes the description of the WNS Integrated Asset Model. The well-group allocations for both the producers and
injectors are now sent back down to the simulation models. The compositions of all gas injection streams as well as the lift
gas are also passed back down to the simulation models.

If any new wells are to be brought on-line for the next synchronization step, commands to open the well and initial well
operating conditions (rate target or tubing head pressure target) are passed down to the reservoir simulator. At this point, the
reservoir simulators should have all the information they need for the next time step and they are allowed to proceed.

Run Time
Current run times for the WNS integrated asset model enable multiple runs to be made in a day. The goal is for complete
predictive runs to run in 5 hours or less.

Future Work
Future work planned for the WNS IAM centers primarily around incorporation of an additional gas injection optimization
method. The method under consideration centers on a concept similar to the oil-efficiency discussion above. The main
thrust of the methodology is that any particular production well would have two different curves on a plot of cumulative oil
versus hydrocarbon pore volumes injected, one for enriched gas and one for lean gas. Since all injection wells connected to
any gas injection manifold must be on the same gas service, there would become an optimal determination of gas service to
each gas injection manifold. This would be directly akin to the water injection optimization for produced water versus sea
water. So, once again, a simple combinatorial optimization should suffice to optimize the gas injection at any point in time.

Summary
An integrated asset modeling (IAM) approach has been implemented for the Alpine Field and eight associated satellite fields
on the Western Alaskan North Slope (WNS) to maximize asset value and recovery. The operating conditions in WNS create
a unique and changing relationship between the oil, gas and water production, gas lift, miscible water alternating gas
(MWAG) injection, lean gas injection, facilities constraints and injection availability.

Within this work, we have:


• Reviewed previous Integrated Asset Model works in the literature
• Discussed the reservoir management needs for developing an IAM
• Discussed the model requirements and tool selection criteria
• Discussed our IAM development philosophy in building and assembling the IAM
• Provided details of the WNS IAM including
a. An overview of the entire system being modeled
b. The reservoir simulation models incorporated within the IAM
c. The surface pipeline network model
d. The process facility model
SPE 158497 19

e. The IAM controller used to integrate all models and control logic and allocate production and injection
across all wells and fields within the WNS IAM
f. Production optimization methods employed within the IAM including gas lift optimization
g. Injection optimization methods employed with the IAM including
i. Injection water service for each injection manifold and optimized allocation of injection fluids
ii. Lean gas and enriched gas splitting methodology and gas injection allocation.

This paper has tried to be reasonably comprehensive in its coverage of the intimate details in justifying, selecting, and
building an integrated asset model. We realize that every development will be different, yet hope that we have supplied
plenty of food for thought in the approach, selection and utilization of integrated asset models.

Acknowledgements
The authors thank the management of ConocoPhillips, Anadarko Petroleum Corporation, and Schlumberger for permission
to publish this paper. The paper reflects the views of the authors, and does not necessarily reflect the views of the owner
companies.

Nomenclature
a,b,c,d,e,f,z empirical coefficients or constants in equations
af well allocation factor
AF well group allocation factor
B formation volume factor
C input constant rate
Ewi injectivity efficiency
Fu uptime
FVE oil efficiency
g well guide rate
G group guide rate, or
Gradient
GOR gas-oil ratio
GLR gas lift rate
MMP minimum miscibility pressure
Ninj cumulative injection (water, gas or total)
Npo cumulative oil production
Npw,n injection water service(1=produced water;0=sea water)
PGOR producing GOR
Pot potential rate
RLG fraction of lean gas used for blending MI or enriched gas
Rs solution gas-oil ratio
t time
TGOR total gas-oil ratio
WOR water oil ratio
y mole fraction

Subscripts
EF enriching fluid
g gas
gi gas injection
GL gas lift
i well, well group, or drill pad
j well, well group, or drill pad
LG lean gas
o oil
pw produced water
sw sea water
T total
void voidage
w water
wi water injection
20 SPE 158497

Greek
α water type (produced water or sea water)
Β phase and service (o=oil, g=gas, w=water, gi=gas injection, wi=water injection)

References
Avocet Integrated Asset Modeler User’s Manual (Avocet IAM), Schlumberger Technology Corporation (2011).

Chow, C.V. and Arnodin, M.C.: “Managing Risks Using Integrated Production Models: Process Description,” Journal of Petroleum
Technology, 52 (3): pp.54-57 (2000).

ECLIPSE Simulation Software Manuals 2011.2, Schlumberger Technology Corporation (2011)

GAP User Guide, Petroleum Experts IPM (2011)

Ghorayeb, K., Holmes, J., Torrens, R. And Grewal, B.: “A General Purpose Controller for Coupling Multiple Reservoir Simulations and
Surface Facility Networks,” paper SPE 79702 presented at the SPE Reservoir Simulation Symposium held in Houston, TX, 3-5
February 2003.

Gonzalez, F., Bertoldi, L., Lucas, L., Paterson, G., Shah, K., Grewal, B., Okafor, C. and Rodriquez, N.: “A Fully Compositional Integrated
Asset Model for a Gas-Condensate Field,” paper SPE 124131 presented at the 2010 SPE Annual Technical Conference and Exhibition
held in Florence, Italy, 19-22 Sept. 2010.

Gutierrez, F., Hallquist, A., Shippen, M. and Rashid, K.: “A New Approach to Gas Lift Optimization Using an Integrated Asset Model,”
International Petroleum Technology Conference held in Dubai, UAE 4-6 December 2007.

Hysys v7.1 User’s Guide, AspenTech (2009)

Liao, T.T., Lazaro, G.E., Vergari, A.M., Schmohr, D.R., Wallgura, N.J. and Stein, M.H.: “Development and Application of the Sustaining
Integrated-Asset-Modeling Tool,” SPE Production and Operations, pp. 13-19, Feb. 2007.

Liao, T.T. and Stein, M.H.: “Evaluation of Operating Strategies via Integrated Asset Modeling,” paper SPE 75525 presented at the SPE
Gas Technology Symposium held in Calgary, Alberta, April 30 – May 2, 2002.

Litvak, M.L., Clark, A.J., Fairchild, J.W., Fossum, M.P., MacDonald, C.J. and Wood, A.R.O.: “Integration of Prudhoe Bay Surface
Pipeline Network and Full Field Reservoir Models, “ paper SPE 38885 presented at the SPE Annual Technical Conference and
Exhibition held in San Antonio, TX, 5-8 Oct., 1997.

Mathews, G.B.: “On the Partition of Numbers,” Proceedings of the London Mathematical Society, 28, pp. 486-490, (25 June 1897).

Peneloux, A., Rauzy, E. and Freze, R.: “A Consistent Correlation for Redlich-Kwong-Soave Volumes,” Fluid Phase Equilibria 8 (1): pp.
7-23 (1982).

Peng, D-Y., and Robinson, D.B.: “A New Two-Constant Equation of State,” Ind. Eng. Chem. Fund., 15 (1), pp. 59-64, (1976).

PIPESIM User’s Manual 2011.1, Schlumberger Technology Corporation (2011)

Satter, A., Varnon, J.E. and Hoang, M.T., “Integrated Reservoir Management,” Journal of Petroleum Technology, pp. 1057-1064, Dec.
1994.

Serbini, F., Wee, L.K., Wong, L.H. and Gomez, N.: “Integrated Field Development – Improved Field Planning and Operation
Optimization,” paper IPTC 14010 presented at the International Petroleum Technology conference held in Doha, Qatar, 7-9 Dec.
2009.

Tesaker, O., Overland, A.M., Arnesen, D., Zangl, G., Al-Kiani, A., Torrens, R., Bailey, W., Couet, B., Pecher, R. and Rodriquez, N.:
“Breaking the Barriers – The Integrated Asset Model,” paper SPE 112223 presented at the 2008 Intelligent Energy Conference and
Exhibition held in Amsterdam, The Netherlands, 25-27 Feb. 2008.

Zapata, V.J., Brummett, W.M., Osborne, M.E. and Van Nispen, D.J.: “Advances in Tightly Coupled Reservoir / Wellbore / Surface
Network Simulation,” SPEREE 4 (2) pp. 114-120 (2001).

You might also like