See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/311462796

Diluent Injection Optimization for a Heavy Oil

Field

Conference Paper · January 2016

DOI: 10.2118/184119-MS

CITATION READS

1 183

4 authors, including:

Arnaud Hoffmann Curtis Whitson

Petrostreamz Norwegian University of Science and Technology
10 PUBLICATIONS 4 CITATIONS 103 PUBLICATIONS 1,380 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Curtis Whitson on 31 January 2017.

The user has requested enhancement of the downloaded file.

SPE-184119-MS

Diluent Injection Optimization for a Heavy Oil Field

A. Hoffmann and W. Astutik, Petrostreamz AS; Flemming Rasmussen, Statoil; C. H. Whitson, NTNU

Copyright 2016, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Heavy Oil Conference and Exhibition held in Kuwait City, Kuwait, 6-8 December 2016.

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
Past studies have shown that use of diluent injection with ESPs can be an efficient artificial lift method for
heavy oil fields. It consists of injecting a light hydrocarbon liquid to reduce the oil density and viscosity.
This paper describes an integrated modeling solution designed to maximize the reservoir oil production
while minimizing the diluent requirement and keeping the crude oil quality within technical and marketing
specifications. The field studied is an offshore heavy oil asset. It consists of two reservoirs with API gravities
of 14 and 12, and oil viscosities at reservoir conditions of 70 cp and 500 cp. The field includes some 60
production wells.
Diluent can be injected (1) in each individual well at the ESP and (2) in the surface processing facility
prior to the second stage separator. Operating constraints include (1) minimum wellhead pressure, (2) diluent
availability, (3) final crude quality specifications, (4) maximum field oil and liquid production rate. The
difficulty of the production optimization problem lies in the nonlinearity of the well production curves
and viscosity model. In this paper, we develop a Mixed Integer Linear Programming (MILP) formulation
by piecewise linearizing the nonlinear behaviors. For each well at each time step, we adjust the black-oil
rates from a reservoir simulator to create piecewise linear well performance curves giving the reservoir oil
production as a function of diluent injected at the ESP.
The proposed integrated solution is used for the entire production life of the field, which is still in the
development phase. The solution is coupled with a reservoir simulator (1) to determine optimal diluent
requirements over time, (2) forecast field production of reservoir oil, diluent, water and gas, and (3) foresee
eventual bottlenecks in the infrastructure design (e.g. limiting constraints). The proposed solution can easily
be used as a Real Time Production Optimization (RTPO) tool to find the optimal operating point based
on the latest measurements (or real-time data). The optimal solution ensures the highest field reservoir oil
production while meeting all constraints and keeping the diluent consumption at a minimum. The increase
of the field oil production rate due to optimal diluent allocation ranges from 2 to 10 %. Cumulative reservoir
oil production increases by approximately 3 million std m3.
The uniqueness of the solution comes from the integration of all operating constraints into a single
mathematical formulation. The computational time (1s – 10s) of the proposed solution outperforms any
classical nonlinear approach. This allows running many sensitivity analyses of the entire integrated asset
model.
2 SPE-184119-MS

Introduction
Production of heavy and extra-heavy oils is more complicated than production of lighter oils. It requires
more equipments that lead to additional CAPEX and OPEX. Warming techniques like steam injection have
been used successfully in artificial lift methods such as Progressive Cavity Pumping (PCP) and Sucker Rod
Pumping (SRP). There are also non-thermal techniques that can be used for viscosity reduction. Unlike
thermal recovery techniques, cold techniques can be used with more sensitive artificial lift systems like
Electrical Submersible Pumps (ESPs) which cannot support high temperatures.
Typically appropriate for less viscous fluids, ESPs have also been used as artificial method for heavy oils
(Castro et al., 2015; Lopez and Morales, 2014). ESPs have the advantage of being capable to accommodate
a wide range of operating conditions, allowing big variations of reservoir conditions and large range of
production rates (Castro et al., 2015). ESP is therefore a suitable artificial lift technique for offshore heavy oil
fields which require a high production rate to reward the high initial investments (A. Bannwart et al., 2013).
G. Amaral et al. (2009) and A. Bannwart et al. (2013) study the effect of higher viscosity fluids (sometimes
presenting emulsion) on the ESP performances and observed a deterioration of pump performances for high
viscosity oils.
ESPs become a more efficient artificial lift method for heavy oils if the water-cut (WC) is relatively high
(> 80%) or a light oil (a diluent) is injected into the wellbore, see Fig. 1. The diluent is blended with reservoir
oil and therefore reduces (1) the oil viscosity (leading to lower friction pressure drop) and (2) the oil density
(thus lower gravity pressure drop). Both effects will lead to better well and pump performances, therefore to
an increased reservoir oil production. Additional benefits of diluent injection include (1) improved oil/water
separation in the surface processing facility and (2) better final oil product quality (Aponte et al., 2001).

Figure 1—Downhole diluent injection system.

Challenges faced by operators include:

• Optimal diluent allocation among the wells

• Field constraints: liquid processing capacity, power capacity for the ESPs and diluent availability

• Dynamically changing system with reservoir conditions (reservoir pressure, WC and GOR) varying
over time and new wells being constantly drilled
Aponte et al. (2001) successfully applied downhole diluent injection to a heavy oil field in Venezuela.
They show that injecting an adequate rate of diluent below the ESP intake improves the well performances of
the wells. It also reduces drastically the operational costs (less expensive equipments are required) and helps
to extend the ESP run life. They also present an integrated strategy which allows monitoring, supervising,
SPE-184119-MS 3

optimizing and controlling the diluent injection process. However, the optimization and control procedures
are done on a well-by-well basis, without taking into account the field constraints (e.g. diluent availability).
Jose et al. (2010) successfully used diluent injection together with ESPs for extra-heavy oils in Venezuela.
They decided to switch from traditional artificial lift techniques (SRP and PCP) to ESPs to handle higher
production rates. They inject a lower viscosity oil (32° API) at the bottom of the well to dilute the extra
heavy reservoir oil (8° API), resulting in a lighter oil mixture (12-13° API). They show that the incremental
well production obtained by diluent injection and ESP ranges from 200 to 2000 stb/d.
Such a field development requires expensive CAPEX and additional OPEX due to the high price of light
oil (diluent). The impact of diluent on the reservoir oil production depends on many parameters including
(1) producing WC, (2) producing Gas-Oil Ratio (GOR) and (3) diluent type (density and viscosity). The
analysis of the effect of diluent injection is complex as both well performance and PVT aspects have to
be investigated (the crude blend above the injection point having very different PVT properties than the
original reservoir oil). In addition, the performance of individual wells is changing over time (decreasing
reservoir pressure, increasing WC and GOR) resulting in optimum allocation of diluent among the wells to
vary with time. The complexity of the production system and the numerous interdependencies do not allow
this allocation to be manually done solely relying on intuition or on semi-analytical procedures. A reliable
decision-support tool is needed.
This paper presents an integrated model designed to maximize the reservoir oil production while honoring
all field and well constraints. The optimization solution integrates all models of the value chain (reservoir,
IPR, vertical lift, ESP performances, processing facility) and solves the optimization problem using linear
programming. The solution is coupled to a reservoir simulator to (1) determine optimal diluent requirements
over time, (2) forecast field production of reservoir oil, water and gas, and (3) foresee eventual bottlenecks in
the infrastructure design. These forecasts are particularly important as they are used to assess the economic
viability of the field development plan. We also show how the same solution can be used as a Real-Time
Production Optimization (RTPO) tool.

Production System
The field studied is an offshore heavy oil field. It consists of two reservoirs (L for light and H for heavy)
with respective API gravities of 14 and 12 API degrees, and oil viscosities at reservoir conditions of 70
cp and 500 cp. The field includes some 60 production wells. ESPs are installed on all wells. A 2 7/8" slim
tubing with diluent injection sub on the ESP intake is installed to allow downhole diluent injection. Wells
from reservoir L are expected to have a higher production rate than wells from reservoir H.
The processing facility consists of a 2-stage separation train followed by a coalescer aimed to separate
water from oil. Diluent can be injected (1) in each individual well at the ESP and (2) in the surface processing
facility prior to the second stage separator, see Fig. 2.

Figure 2—Diluent injection points.

4 SPE-184119-MS

The field is operated under numerous constraints:

1. Wellhead pressure of individual well has to be greater than 20 bar
2. Two specifications on the final crude blend:
a. A technical specification: the viscosity of the final crude has to be lower than 10 cp at 105° C
to ensure a good water/oil separation in the coalescer (aimed to reduce the WC to 0.5 %).
b. A market specification: the viscosity of the final crude has to be lower than 400 cSt at 40° C
to meet logistical requirements.
3. Maximum liquid and oil load in the coalescer: 15 000 std m3/d for oil and 60000 std m3/d for liquid
4. Diluent availability
5. Maximum liquid and oil well production rate to avoid pump failure and sand production
6. Maximum downhole diluent injection rate (imposed by the size of the injection line)
A major challenge is due to the rapid time variations of the reservoir conditions. Typically for heavy-
oil, reservoirs show rapid water cut increases due to water coning, see Fig. 3. Therefore, well and pump
performances are dramatically changing over time.

Figure 3—Time variation of the field average WC of each reservoir.

Well Performance
Well performance challenges for this field include (1) consistent black-oil and compositional (EOS) PVT
modeling, (2) emulsion modeling and (3) operational constraints. Classical well performance models (IPR/
VLP analysis) usually give the reservoir oil production rate of a given well as a function of the reservoir
pressure (pR), the producing GOR and WC, productivity index (PI), and the wellhead pressure (pwh):

Assuming reservoir pressure, WC, GOR and PI to be constant (at least for a short period of time; e.g. a
reservoir simulator time step or between two periodic well tests), we end up with:

This model does not take into account the performance variations due to diluent injection and therefore
needs to be refined.
SPE-184119-MS 5

Both reservoir oils (from reservoir L and H) have uniform properties (GOR, API, viscosity, etc.) However,
when diluted, these PVT properties can dramatically change. Most of commercial simulator do not allow
more than one fluid model, thus two models are used to model each individual well: (1) a lower well model
(from the reservoir up to the ESP intake) using the reservoir oil PVT model and (2) an upper well model
(from ESP intake to the wellhead) using the diluted PVT model, see Fig. 4.

Figure 4—PVT model for the upper and lower part of an individual well.

Note that the PVT properties of the diluted oil depend on the quantity of diluent being injected. We
introduce the Diluent Cut (DC) defined by:

The PVT properties of the fluid above the injection point are therefore a function of the DC. Consequently,
the well performance depends on the DC:

Note that the dependency on ESP frequency is not considered here.

As often with heavy oil production, a "water-in-oil" emulsion or an "oil-in-water" emulsion may exist
depending on the water-cut. The inversion point is the WC at which the flow changes from oil-continuous
to water-continuous. Close to the inversion point, the viscosity of oil-water emulsion is significantly higher
than the water-free oil viscosity (10 to 100 times higher). This can lead to significant increase of the friction
pressure losses in the tubing, thus to a deterioration of the well performance.
Accurate emulsion modeling is needed to predict the changes of the mixture viscosity. Several emulsion
measurements have been performed on the reservoir oils, with and without diluents. The experimental
results showed that the inversion point increases with DC. Brinkman's correlation (Brinkman, 1952) was
used to calculate the ratio of the emulsion viscosity to the water-free oil viscosity before the inversion point
(oil-continuous flow). After the inversion point (water-continuous flow), experimental data was used for
the ratio of the emulsion's viscosity to the oil-free water viscosity. Note that this ratio does not change
with the amount of diluent injected for water-continuous flow. This implies that after the inversion point,
injecting more diluent does not change the viscosity of the mixture. For some wells of reservoir L producing
high liquid rates (close to the ESP capacity), injecting more diluent will actually deteriorate the well
performances. In these cases, the chosen ESP type-and-capacity limits the possible benefits of diluent
injection, see Fig. 5. After the inversion point and/or when the liquid rates are close to ESP maximum
capacity, there will be no (or negligible) positive effect of diluent injection. Thus, there is an optimum diluent
cut for each well to ensure benefit from the viscosity reduction and optimum ESP operating condition.
6 SPE-184119-MS

Figure 5—Well performance curve with liquid rate close to ESP capacity
3
(3500 std m /d). The wellhead pressure is held constant and equal to 20 bara.

Optimization Problem
Objective, Variables, Operational Constraints
The optimization problem consists of maximizing the reservoir oil production from both reservoirs:

by changing the choke opening and the diluent injection rate ( ) of each individual well. is the reservoir
oil production rate from the lighter oil reservoir (14° API) and is the reservoir oil production rate from
the heavier oil reservoir (12° API). By definition:

and

where is the reservoir oil production rate from well i and WL and WH are the sets of wells producing
respectively in the lighter and heavier oil reservoirs. The reservoir oil rate produced by well i is given by:

The liquid rate coming from well i is computed as follows:

where .
Each individual well is subject to numerous operational constraints such as maximum diluent injection
rate (size of the injection line) and maximum production rates for liquid and oil:
SPE-184119-MS 7

Chokes are not directly modeled, but the pressure drop across the choke can be estimated from:

In addition, the well should not be choked if diluent being injected into the well. This rule-based constraint
is challenging to formulate using linear programming. Its implementation is not discussed in this paper.
Additional operational constraints come from the topside installations. A certain volume of diluent has
to be injected (downhole and topside) to meet (1) technical specification (to ensure a sufficient water/oil
separation in the coalescer) and (2) a market specification (to ensure a minimum marketable quality of the
final blend). The volumes of diluent required to meet these two specifications can be estimated from the
ASTM D7152-11 viscosity model. The ASTM model is a volumetric mixing rule estimating the viscosity
of a mixture of several products (whose individual viscosity is known). Thus, the required diluent quantities
to meet both viscosity specifications (technical and market) can be respectively written as follows:

Note that and are the overall quantities of diluent that have to be injected to meet the
product specifications. They can be injected downhole or topside.
The field oil rate (reservoir oil + diluent injected downhole + diluent injected prior to the second stage
separator) is limited by the coalescer capacity:

In addition, the field liquid rate (reservoir oil + diluent injected downhole + diluent injected prior to the
second stage separator + water) is also limited:

The eventual field limitation for power generation is not included as a field constraint in this paper. It is
assumed that the generator on the platform is capable to generate the required power.
8 SPE-184119-MS

SOS2 Piecewise Linear Approximation

The non-linear problem is piecewise linearized. Non-linear well performance curves and viscosity model
are replaced by piecewise linear tables, see Fig. 6.

Figure 6—Piecewise linearization of a non-linear function.

To allow interpolation between points of the table, we use SOS2 models. A SOS2 (Special Ordered Set
of type 2) is an ordered set of non-negative variables, of which at most two can be non-zero, and if two are
non-zero these must be consecutive in their ordering.
Let f be a continuous function. We know the value of f for a set of breakpoints {Xi, i = 1, …, N}. The
SOS2 piecewise linear approximation of f can be written as follows:

where (λi)1≤i≤N is a set of positive weighting factors. The λi have to verify the following equation:

The SOS2 model imposes the set of weighting coefficients (λi)1≤i≤N to be a SOS2. In addition,

In the case of Fig. 7, we have λ1 = λ2 = 0, λ3 ≠ 0, λ4 ≠ 0 and λ3 + λ4 = 1.

SPE-184119-MS 9

Figure 7—SOS2 models in one dimension (after Hoffmann (2014)).

The SOS2 approximation can be evaluated as follows:

Note that the deviation θ depends primarly on the number of points used to represent function f. The
more points, the lower θ.
SOS2 approximations are used to model (1) the well performances of each individual well
( ) and (2) the technical and market specifications ( and ).

Mixed-Integer Linear Problem

After piecewise linearization using SOS2 models, the resulting optimization problem is a mixed-integer
linear problem. The integer variables (in fact, binary variables) are necessary to model the adjacency of
non-zero coefficients in the SOS2. Efficient algorithms, such as the simplex and the branch & bound, exist
to solve this kind of problem with the assurance of converging towards the global optimum.

Generation of Piecewise Linear Tables

Two types of piecewise linear tables are to be generated: (1) well performance tables and (2) viscosity
specification tables.
Well performance tables provide the reservoir oil rate produced by a given well for a set of operating
conditions (wellhead pressure and DC). Note that these tables are generated at fixed reservoir conditions.
They are generated using two well models and a commercial simulator (in our case, PROSPER). Pipe-It
software (a commercial IAM platform) is utilized to couple the two PROSPER models and to automate
the generation of the tables. For each point of the table, a set of operating conditions is fed to the lower
well model for pressure drop calculation. Reservoir oil is then blended with diluent. Note that the blending
of diluent and reservoir oil is performed on a compositional basis within Pipe-It. Diluent composition is
known from laboratory measurement (the diluent composition is assumed constant here, but it could vary
with time), while the reservoir oil composition is estimated from a black-oil delumping procedure using
oil and gas volumetric rate at perforation level obtained from a black oil reservoir simulator, see Kuntadi
(2013). The black-oil properties of the mixture (oil, gas and water rates, density, etc.) are calculated within
Pipe-It and fed to the upper well model for another pressure drop calculation. This procedure is launched
10 SPE-184119-MS

iteratively until obtaining the desired wellhead pressure. At each iteration, the flowing bottomhole pressure
is adjusted using Brent's algorithm (Brent, 1970).
For a given production rate from reservoir H and L, the viscosity specification tables provide the required
quantity of diluent to obtain a certain mixture viscosity at specified conditions. These tables are obtained by
calling the ASTM viscosity model for different combinations of production rates from reservoir H and L.

Case Study
The field studied being still in its field development phase, the results presented in this section come from
a coupling with a reservoir simulator. After each time step of the reservoir simulator, the well performance
tables are re-generated (using the current values of reservoir pressure, WC, GOR and PI for each well). The
optimization is relaunched with the updated tables.
Table 1 gives the constraints used for the simulation. Synthetic data are used.

Table 1—Data used for the case study.

Maximum field liquid production rate ( ) 60000 std m3/d

Maximum field oil production rate ( ) 15000 std m3/d

Maximum diluent injection rate for wells of reservoir L

( ) 200 std m3/d

Maximum diluent injection rate for wells of reservoir M

( ) 400 std m3/d

Maximum well oil production rate ( ) 3500 std m3/d

Maximum well liquid rate ( ) 3500 std m3/d

Minimum wellhead pressure ( ) 20 bara

Viscosity constraint in the coalescer 10 cp at 105° C

Market viscosity constraint 400 cSt at 40 °C

Diluent API gravity 50° API

Base Case: Diluent vs. no Diluent

This study aims to quantify the impact of downhole diluent on the reservoir oil recovery. The optimization
algorithm presented above is used to determine the optimal allocation among the wells. We consider three
scenarios: (1) no diluent injection, (2) constant allocation of diluent among the wells (5 vol-% DC for wells
producing in reservoir L and 10 vol-% DC for wells producing in reservoir H) and (3) optimal diluent
allocation.
The effect of downhole diluent injection is shown in Table 2. In both cases 2 and 3, the cumulative oil
production increases (4.6% in case 2 and 6.3% in case 3).

Table 2—Impact of downhole diluent injection on the reservoir oil recovery.

Cumulative Oil Production Cumulative

Cases Diluent Demand
3 Deviation from
[million sth m ] [million sth m3]
case 1 [%]

Case 1: No diluent 47.8 - 0

Case 2: Constant diluent allocation 50.0 + 4.6% 4.57

Case 3: Optimal diluent allocation 50.8 + 6.3% 0.53

SPE-184119-MS 11

Figure 8 shows the incremental reservoir oil production (rate and cumulative) obtained by injecting
diluent and Figure 9 shows the quantity of diluent being injected downhole.

Figure 8—Incremental reservoir oil production obtained by diluent injection.

Case 1 (no diluent) is used as base case for increment calculations.

Figure 9—Quantity of diluent injected for the three cases.

In case 2, a constant DC is maintained in all wells. However, reservoir conditions are changing over time
and the impact of diluent on a particular well is also changing over time. In case 3, the optimization solution
injects diluent where its impact is the highest. This is the reason why case 3 can produce additional reservoir
oil (+0.8 MMsm3) while consuming less diluent (9 times less). This shows the benefits of implementing a
consistent and reliable integrated modeling solution for supporting decisions.
Note that no diluent is injected after 2060 in the optimal case because of the high field WC. Diluent
injection is very efficient before the inversion point. After the inversion point, the mixture viscosity is not
affected by the amount of diluent injected. In that case, diluent injection has very limited impact on the
reservoir oil production and can even deteriorate the pump performances by increasing the viscous losses.
On the other hand, the first part of operations (2050-2054) being constrained by the platform limitations
(maximum liquid and oil production), almost no diluent is injected before 2054. Therefore, diluent injection
mainly happens between 2054 and 2060 and the additional oil production is recovered during these years.
12 SPE-184119-MS

Additional benefits of diluent injection include (1) lighter final crude and (2) lower oil viscosity, see Fig.
10. The API gravity increase due to diluent injection (downhole and topside) is ranging from 2 to 6 API
degrees. The viscosity gains are even more significant. The better final oil quality achieved with diluent
injection contributes greatly to a more marketable product and therefore to a higher selling price.

Figure 10—Influence of diluent injection (downhole and topside) on the final crude properties: (a) API gravity and (b) viscosity.

Sensitivity to the Diluent Type

A legitimate question concerns the sensitivity of the diluent injection to diluent properties. Lighter diluents
are more expensive, but do they have greater impact on the oil production? We evaluated the optimal diluent
injection rates in three scenarios: (1) medium density diluent, (2) lighter diluent and (3) heavier diluent,
see Table 3.

Table 3—Properties of the different diluents.

API gravity C7+ (mol-%)

Diluent 1 (used in the base case) 49 72 %

Diluent 2 (lighter) 51 65 %

Diluent 3 (heavier) 45 76 %

Table 4 gives the optimal reservoir oil recovery and diluent demand. Diluent properties (density and
viscosity) have a great impact on the diluent demand. The cumulative oil recovery is essentially the same in
the three cases. However, the lighter the diluent, the lower the diluent demand, see Fig. 11. Injecting diluent
2 instead of diluent 3 allows to save about 0.2 million std m3 while recovering the same quantity of reservoir
oil. This represents eventual significant savings in OPEX. However, this conclusion needs to be balanced
by the fact that market price of diluent 2 may be higher than diluent 3. A detailed study of diluent losses
is also required. The diluent being injected into wellbore, it passes through the surface process. Therefore,
some diluent (usually 4-9 vol-%) is flashed off and ends up in the product gas. This amount of diluent is
effectively lost since it is burn as fuel gas instead of being sold as part of the product oil. The diluent density
can considerably affect the diluent loss (the lighter the diluent, the higher the diluent loss). The lightest
diluent (diluent 2) is the most efficient to increase the reservoir oil recovery (see Tab. 4), but its more volatile
behavior in the surface process may counterbalance its positive effect on the well performance.
SPE-184119-MS 13

Table 4—Influence of the diluent properties on the optimal reservoir oil recovery.

Cumulative reservoir Cumulative downhole

oil prodution diluent demand
[million std m3] [million std m3]

Diluent 1 (used in the base case) 50.80 0.53

Diluent 2 (lighter) 50.84 0.40

Diluent 3 (heavier) 50.80 0.59

Figure 11—Influence of the diluent type on the optimal quantity of diluent injected.

Solution Performance Analysis

Runtime
The optimization solution runtime ranges from 0.1 sec (when diluent injection is no longer beneficial) to 5
sec (during the phase when diluent injection and choking take place). The table generation is the most time
consuming part of the solution as it calls a commercial application and requires iterations. It usually takes
between 1 and 3 minutes to generate new tables for all wells.

Piecewise Linear Models Accuracy

Figure 12 compares the well pressure drop estimated from the piecewise linear tables to the ones obtained
from a non-linear commercial software. This comparison is done for one well of the field L. They show that
the piecewise linear models are capable of modeling the well performance behavior with an acceptable level
of accuracy. The average deviation is smaller than 0.2 % and the maximum deviation observed is around 2
%. The maximum deviation is observed when the WC is approaching the inversion point.
14 SPE-184119-MS

Figure 12—Comparison of the optimization results against the pressure drop calculated
by the commercial simulator for a well producing in reservoir L. Error bars are +/- 2 %.

Real-Time Production Optimization Tool

One of the great benefit of this solution is its flexibility and capability of being easily adapted to short-
term production optimization. Instead of using IPR and reservoir characteristics extracted from a reservoir
simulator, the solution can use the latest well test measurements (rates, GOR, WC, current reservoir
pressure) to update the well performance models of each well. Wellstream compositions can be estimated
from well test data using the well test conversion technology (Hoda and Whitson, 2013). The blending of
reservoir oil and diluent can then be done on a compositional basis. New piecewise linear tables are then
generated using updated well performance models.
The same optimization solution can be used with the new tables. Additional real-time data can also be
included into the solution to model the time variations of the reservoir conditions more accurately. An
additional constraint may need to be added to consider the diluent availability:

where is the remaining volume of diluent in the tank and ΔT delivery the time (in days) until the next
diluent delivery.
A real-time production optimization tool is especially valuable during operations to redefine the optimal
operating point when:

• New test data (rates, WC, GOR and reservoir pressures) are available;

• When critical elements (e.g. gas compressor, diluent injection pump) unexpectedly fail;

• The processing facility is down due to maintenance, leading to a reduced oil and/or liquid handling
capacity;
• Work-over is carried out on one (or several) well(s);

• New wells are drilled and starting production.

It also allows the user to do a lot of "what-if" analyses to compare and assess multiple scenarios.
The approach presented in this paper has the advantage of (1) converging very fast (0.1-5 sec) and (2)
always converging to the global optimum (or very close). It is therefore a good candidate for RTPO.
SPE-184119-MS 15

Further work will include the dependency of the well performance on the ESP frequency and the
maximum power the platform can supply. Additional operating constraints can be added to ensure that ESPs
are operated within a specific envelope which would decrease the risk of pump failure and ultimately lead
to increased pump life time.

Conclusions
ESPs used with diluent injection can be an efficient artificial lift method for heavy oil fields. It is especially
suitable for offshore operations that require high production rates. The time variations of reservoir conditions
make the daily allocation of diluent very challenging. PVT, well performance and pump operation aspects
need to be investigated to optimally distribute diluent among the wells.
We presented an integrated modeling solution to maximize reservoir oil recovery while minimizing the
injection of diluent. The solution was coupled to a reservoir simulation in this paper. Additional reservoir
oil recovered by optimal diluent injection is about 3 million std m3(+ 6 %). Additionally, the final crude
quality (density and viscosity) is improved, leading to a more marketable product (i.e. higher selling price).
We also showed that the diluent type can significantly impact the optimal operating point.
We give guidelines to use the same solution during operations as a RTPO tool to support daily/weekly
decision-making. Light oil being expensive, an optimal injection strategy is consequently required.

Nomenclature
q Volume rate at standard conditions [std m3/d]
p Pressure [bara]
w Set of wells

Acronyms:
WC Water Cut [%]
GOR Gas Oil Ratio [std m3/std m3]
DC Diluent Cut [vol-%]
ESP Electric Submersible Pump
PCP Progressive Cavity Pump
SRP Sucker Rod Pump
MILP Mixed Integer Linear Programming
MINLP Mixed Integer Non-Linear Programming
CAPEX Capital Expenditure
OPEX Operational Expenditure
PVT Pressure, Volume, Temperature
IPR Inflow Performance Relationship
VLP Vertical Lift Performance
PI Productivity Index
RTPO Real-Time Production Optimization
IAM Integrated Asset Modeling

Subscripts:
liq Liquid, considering oil + water
field Relative to the field
R Reservoir
w Water
16 SPE-184119-MS

o Oil
dil Diluent
wh Wellhead

Superscripts:
L Lighter oil field
H Heavier oil field
res Reservoir
max Maximum
min Minimum
tech Technical constraint
market Market constraint

Reference
1. V. Castro, D. Leite, D. Lemo, J. Marins, R. Pessoa, J. Magalhães, ESP Application on Heavy Oil
in Peregrino Field, SPE-173948-MS, 2015.
2. J. E. Lopez, G. E. Jimenez Morales, ESP Optimization in an Extra Heavy Oil Field: Case Study
in Colombian Llanos Basin, SPE-170039-MS, 2014
3. G. Amaral, V. Estevam, Fernando A. Franca, On the Influence of Viscosity on ESP Performance,
SPE-110661-PA, 2009.
4. A. Bannwart, N. Arrifano Sassim, V. Estevam, J. Biazussi, W. Monte Verde, Gas and Viscous
Effects on the ESPs Performance, SPE-165072-MS, 2013
5. H. Aponte, L. Toussaint and M. Ramos, Experiences Using an ESP Application on Heavy-Oil
Cold-Production Automation in Eastern Venezuela Fields, SPE-69708, 2001.
6. F. Jose, L. Garcia, J.C. Brown, Experiences in the Use of ESP's in Orinoco Belt Cerro Negro
Area, Venezuela, SPE-139105-MS, 2010.
7. A. Hoffmann, Application of Piecewise Linear Models to Short-Term Oil Production
Optimization, MSc. Thesis, NTNU, 2014.
8. M. F. Hoda, SPE, C. H. Whitson, Well Test Rate Conversion to Compositional Wellstream,
SPE-164334-MS, 2013.
9. ASTM D7152-11(2016), Standard Practice for Calculating Viscosity of a Blend of Petroleum
Products, ASTM International, West Conshohocken, PA, 2016, www.astm.org
10. Brinkman, H.C., The viscosity of concentrated suspensions and solutions, J. Chem. Phys. 20 (4):
571.
11. A. Kuntadi, Stream Conversion Technology and Gas Condensate Field Development, PhD thesis,
NTNU, 2012.
12. R.P. Brent, An algorithm with guaranteed convergence for finding a zero of a function, Stanford
University, 1970.

View publication stats