Computers and Chemical Engineering 128 (2019) 216–227

Contents lists available at ScienceDirect

Computers and Chemical Engineering

journal homepage: www.elsevier.com/locate/compchemeng

A multi-objective optimization model for tactical planning of

upstream oil & gas supply chains
Ahmed M. Attia a,∗, Ahmed M. Ghaithan b, Salih O. Duffuaa a
a
Department of Systems Engineering, King Fahd University of Petroleum & Minerals, Dhahran, Saudi Arabia
b
Department of Construction Engineering & Management, King Fahd University of Petroleum & Minerals, Dhahran, Saudi Arabia

a r t i c l e i n f o a b s t r a c t

Article history: The supply-chain of oil and gas is known as the hydrocarbon supply-chain (HCSC). HCSC constitutes a sig-
Received 13 December 2018 nificant part of the world economy of the energy sector. Consequently, planning the activities of the HCSC
Revised 30 May 2019
is an important and challenging research area. Here we present a multi-objective optimization model for
Accepted 13 June 2019
intermediate-term planning of HCSC. The model brings forward sustainability and the environment aspect
Available online 14 June 2019
of planning HCSC operations. Sustainability is considered as an objective by minimizing the depletion
Keywords: rate. The environment is considered as a constraint in the model by restricting CO2 release. In addition,
Hydrocarbon supply chain the model includes minimizing the total cost and maximizing revenue as two objectives. The proposed
Tactical planning model assists decision makers to study trade-offs among alternative decisions. Decisions have an impact
Multi-objective optimization on meeting demand, depletion rate of natural resources, minimizing costs and/or maximizing revenue. A
case study based on real data is provided to demonstrate model practicability and a sensitivity analysis is
conducted to derive managerial insights. For instance, oil and gas reserves can be sustained for a longer
period without the reduction of market share and simultaneously, return covers the total cost.
© 2019 Elsevier Ltd. All rights reserved.

1. Introduction product framework. In addition to considering the sustainability of

Energy generating products are some of the world’s essential
commodities. Consequently, countries that have high levels of trad-
ing and reserves of energy resources, especially crude oil and/or
natural gas, represent a significant amount of power in the world’s
economy. The supply of these products made available to the world
energy market has an impact on energy prices. Over the past years,
crude oil prices have declined sharply, leading to a considerable
revenue shortfall in producing countries. Besides, if these countries
were to reduce their oil production, they would expect to lose a
market share and cut-off natural gas (i.e., associated gas) supply
to their industrial plants. Thus, the production of crude oil and
natural gas should be planned simultaneously. HCSC produces oil
and gas, in which, they are usually processed into several products
such as gasoline, benzene, jet oil, natural gas, methane, and other
products. In addition, many objectives can be thought in such a
complex supply chain. Crude oil and natural gas are depletable
resources, and their production and processing hurt the environ-
ment since they release gases such as CO2 . A such, modeling HCSC
requires a multi-objective, multi-dimension, multi-echelon, multi-
∗
Corresponding author.
E-mail addresses: amattia@kfupm.edu.sa (A.M. Attia), ahmedgh@kfupm.edu.sa
(A.M. Ghaithan), duffuaa@kfupm.edu.sa (S.O. Duffuaa).
natural resources and the impact on the environment.
The activities within the HCSC begin with exploration and
production, followed by processing and refining, and finally, end
with distribution to the end customer. These activities usually
are segmented as upstream, midstream, and downstream sectors.
Sahebi et al. (2014) list the entities and activities in each segment
for the crude oil supply chain, as listed in Table 1.
Based on the literature survey in Section 2, considerable work
has been conducted in the area of HCSC deterministic optimiza-
tion; there are two papers using MOO. The first paper is by
Iakovou (2001) who formulated a multi-objective model for the
logistics of the downstream segment. The second paper is by
Ghaithan et al. (2017) which proposed a multi-objective model
for the downstream segment of the HCSC. It is worth noting that,
the first paper only addressed logistics. While the second paper
focused on downstream HCSC segment and did not consider the
control of the depletion rate of natural resources as a measure of
sustainability.
To the best of the authors’ knowledge, no multi-objective
model has been formulated for the upstream segment of the
HCSC. Moreover, the environmental impact of such a system has
been considered by bounding CO2 emissions to a certain limit. The
model produces feasible plans in terms of returns and expenses, at
the same time, mitigates environmental impact and depletion of

https://doi.org/10.1016/j.compchemeng.2019.06.016
0098-1354/© 2019 Elsevier Ltd. All rights reserved.
 A.M. Attia, A.M. Ghaithan and S.O. Duffuaa / Computers and Chemical Engineering 128 (2019) 216–227 217

produces an efficient Pareto set that allows the decision maker to

Acronyms compromise between economic and non-economic objectives.
The remainder of this paper is organized as follows: In
AH Arabian heavy crude Section 2, a literature review of the upstream HCSC is provided.
AL Arabian light crude The HCSC network description and the MOO formulation are dis-
AM Arabian medium crude cussed in Section 3. A case study and sensitivity analysis are pre-
AXL Arabian extra light crude sented in Sections 4 and 4.2, respectively. The paper is closed by
Bbbl Billion barrels conclusion and recommendation for future research in Section 5.
CO2 Carbon dioxide
GOR Gas-oil ratio 2. Literature review
GOSP Gas-oil separator plants
HCSC Hydrocarbon supply chain Several research papers have modeled the different segments
H2 S Hydrogen sulfide of the HCSC either in an integrated form or studied each segment
LP Linear programming separately. The literature can be classified by product as oil- or gas-
MILP Mixed integer linear programming oriented or by segment as upstream, downstream, or integrated;
MINLP Mixed integer nonlinear programming where the midstream segment has not been studied separately. In
MMbbld Million barrels per day this review, only deterministic models were considered.
MMcft Million cubic feet The first set of papers discusses the upstream oil-oriented
MOO Multi-objective optimization HCSC. Nygreen et al. (1998) reported a MILP that has been used
NGL Natural gas liquid by the Norwegian Petroleum Directorate for production and trans-
NPV Net present value portation planning. Iyer et al. (1998) formulated a MILP model for
OPEC Organization of the Petroleum Exporting Countries facility allocation, production planning, and scheduling of offshore
PSC Petroleum supply chain oil fields. For simplicity, they linearized the reservoir behavior
SCN Supply chain network utilizing a piecewise linear function.
van den Heever and Grossmann (20 0 0) discussed the same
problem in Iyer et al. (1998) considering non-linear reservoir
behavior. The goal of their research was not to formulate a model
the natural resources. The model proposed in this paper addresses rather than to develop an algorithm to solve the MINLP models
filling the gap identified. utilizing decomposition and aggregation/disaggregation techniques.
This paper proposes a multi-objective mathematical program- Later, van den Heever et al. (20 0 0) extended the model by includ-
ming model that can be used for solving real HCSC problems. ing more complex economic objectives. The model objective was
The multi-objective framework (i.e., study the trade-off among to maximize the NPV of sales revenues, capital costs, operating
conflicting objectives) and tactical planning decision level (i.e., costs, taxes, tariffs, and royalties. Consideration of the complex
weekly or monthly planning period) are embraced to model the economic rules was found to be more profitable, and yielding a
HCSC. The first objective function aims to minimize the total completely different solution. van den Heever et al. (2001) solved
costs of production, processing, transformation, transportation, the same model in van den Heever et al. (20 0 0) using the La-
distribution, and production above or below the demand. The sec- grangean Decomposition and a heuristic to reduce the complexity
ond objective function ensures covering total costs, paying other of the model solution.
expenses, and sustaining development, through maximizing the Carvalho and Pinto (2006a) formulated a MILP model for the
total revenue. The third objective function minimizes the rate of assignment of well-platforms to well-heads in offshore oil fields
depletion of both oil and gas reserves to secure sufficient reserves as discrete decisions. They utilized the Bi-level decomposition
for the coming generations. approach to facilitate the solution of the large scale model with
The model can be used to address tactical decisions related discrete and continuous variables. Carvalho and Pinto (2006b) ex-
to HCSC activities, for instance: allocation of oil production to tended (Carvalho and Pinto, 2006a) model by assuming that, the
reservoirs with a high GOR to ensure the supply of the crude oil pressure inside the reservoir changes globally with the extraction
and natural gas by-products to the vital industries. The considered of oil or gas, independent of the pressure of other reservoirs in
activities in this research work include upstream and midstream the same field; the pressure of all wells belonging to the same
activities for crude oil and the three segments for gas. The related reservoir is, therefore, the same. Although their work focused on
products are crude oil, natural gas, and natural gas byproducts. reservoir behavior, they ignored the change in pressure between
The proposed model has been applied to a real-world case the wells and platforms.
study from the Saudi Arabia HCSC. The obtained MOO model has Ulstein et al. (2007) constructed a MILP model to maximize
been solved by implementing the augmented ε -constraint method the net income from the offshore oil fields in Norway. Although
proposed by Mavrotas and Florios (2013). The proposed model is the model was simple, it is generic and effective in production
a multi-objective, multi-dimension, multi-echelon, multi-period planning for medium terms. Rocha et al. (2009) developed a model
and multi-product. Other features of the model are oil and gas to generate a daily plan for shipping crude oil from the production
reserves sustainability and CO2 emissions. The proposed model site to the refineries.

Table 1
HCSC entities and activities.

Segment Entities Activities

Upstream Wellhead, well platform, production platform, and crude Exploration, production (i.e., recovery and separation), and
oil terminal. transportation to refineries.
Midstream Refinery plant and petrochemical plant. Transformation and production through refineries and
petrochemical plants
Downstream Distribution center/depot, market, customer. Transformation, including storage and distribution to
customers
218 A.M. Attia, A.M. Ghaithan and S.O. Duffuaa / Computers and Chemical Engineering 128 (2019) 216–227
Gupta and Grossmann (2012) formulated the nonlinearity of
the reservoir behavior as a third and a higher order polynomial.
Aizemberg et al. (2014) tackled transportation planning problem
of crude oil from offshore facilities to the next processing units.
They solved problems using a commercial software based on a
Branch-and-Bound algorithm and to make the problem tractable
they used a column generation based heuristic. Moradi Nasab
and Amin-Naseri (2016) considered the installation and capacity
expansion of pipeline routes and facilities for crude oil production,
where they ignored dependency between crude oil and natural
gas. Rocha et al. (2017) proposed a decomposition algorithm based
on a cascading knapsack structure to solve large scale models of
the petroleum supply chain.
The second set of papers discusses downstream oil-oriented
modeling. Sear (1993) addressed the problem of transportation
cost minimization originating from refineries ending at customers.
Iakovou (2001) considered the logistics of petroleum products as a
multi-objective problem. The main objective was to minimize the
transportation costs, while the second objective was to minimize
the costs associated with the risks of spilled oil.
Persson and Göthe-Lundgren (2005) increased the complexity
of Sear (1993) problem by considering refinery scheduling opti-
mization. Elkamel et al. (2008) focused on reducing CO2 emissions
from refineries. Kuo and Chang (2008a, b) modeled the operations
inside the refinery as a detailed SCN. The model was able to
coordinate the planning and scheduling decisions of the refining
segment.
Al-Othman et al. (2008) formulated single objective determin-
istic and stochastic models. The deterministic model objective was
to minimize the difference between the total costs and revenue.
Al-Qahtani and Elkamel (20 08, 20 09) formulated a MILP model
to coordinate the operation of multi-refinery plants. The objective
was to minimize the annualized operating and capital costs based
on decisions regarding capacity expansion, production levels, and
blending levels. Kim et al. (2008) tackled the same problem with
an extra decision which is facility relocation. The sensitivity of
the proposed models, therefore, needs to be examined against the
variation in demand.
Guyonnet et al. (2009) compared the effect of formulating a
fully integrated model for crude oil unloading operation, produc-
tion planning, and distribution process versus a non-integrated
Fig. 1. Upstream HCSC network.
model for each operation. They concluded that the integrated
model achieved a significantly higher profit because of lower
penalties of lost demand, safety stock, and unsatisfied demand.
They tested the model against small problems with unreal (i.e.,
estimated) data. Fernandes et al. (2011, 2013) developed a MILP
for strategic decisions related to depot locations, capacities (e.g.,
refinery, depot, retailer), transportation modes, and transportation
routes. Ghaithan et al. (2017) proposed a complementary model
for the proposed work starting from refining operation considering
crude oil products.
The third set of papers discusses the integrated oil-oriented
network. Neiro and Pinto (2004) considered decisions of the
amount of production of each entity, refinery operational vari-
ables, and inventory and entities assignment. Jiao et al. (2010) pro-
posed a MILP model for Chinese PSC to decide how much to
produce from each entity. They assumed, entities and routes
having unlimited capacity and the shortage is allowed and it is
completely satisfied during the next period before the demand.
Chen et al. (2010) focused on minimizing transportation cost of
imported crude oil within Chinese PSC. Cost elements are trans-
portation costs, operation cost in logistics centers, handling costs
and domestic transportation cost.
The fourth set of papers discusses gas-oriented network.
Duffuaa et al. (1992) formulated an LP model to study the ability
of Saudi Arabia to satisfy the industrial demand of methane and
ethane from associated gas supplies. The model considered a
ceiling of 4.5 Mbbld as an OPEC quota, and, at that time the
associated gas production could satisfy the industrial demand. The
proposed model did not consider the effect of increasing the pro-
duction levels on CO2 emissions. Hamedi et al. (2009) presented
a case study considering transmission and distribution planning
of natural gas. A MINLP model was developed to minimize the
total costs of transportation and processing. The model contains a
constraint to linearize the objective function.
For further reading, Sahebi et al. (2014) conducted a recent
review of the literature on the area of applying mathematical
programming to PSC. Their recommendations for future research
include: (i) examining both strategic and tactical decisions in
an integrated form; (ii) examining nonlinearity of the refin-
ery operations; (iii) exploring environmental impact of the PSC
problems; (iv) modeling uncertainty features with multi-stage
 A.M. Attia, A.M. Ghaithan and S.O. Duffuaa / Computers and Chemical Engineering 128 (2019) 216–227 219

stochastic models, and (v) finding efficient solution techniques for  

multi-objective function problems. GORoi jt xoi jt = ygjit ∀o, ∀ j ∈ n (2)
i∈ro i∈gg
In summary, the literature review indicates that there is a re-
search gap needing contributions for HCSC optimization. Such as,  
xoi jt + xo+
jt−1
= xojit + xo+
jt
∀o, ∀ j ∈ go (3)
managing the HCSC in a MOO framework, planning the production i∈n i∈ po
of oil and gas (associated and non-associated gas) simultane-
 
ously, considering the depletion rate of natural resources, and Piojt xoi jt = xojit ∀o, ∀ j ∈ po (4)
considering environmental impact. i∈go i∈do

 
3. Mathematical formulation Piojt xoi jt = ygjit ∀o, ∀ j ∈ po (5)
i∈go g∈H2S;i∈dg

The considered entities and activities of HCSC, in addition to Regarding the natural gas network, associated gas from GOSPs
model development, are described in the following subsections. (n) and non-associated gas from reservoirs (rg) are collected at gas
gathering centers (gg), Eq. (6). At (gg), the incoming streams from
3.1. Network description
(rg and n) plus the end inventory from the previous period (t−1)
should be equal to the outgoing stream plus the end inventory at
Crude oil and natural gas supply chains constitute the HCSC.
period (t). Next, the outgoing stream sent to gas processing plants
The two supply chains are formed by three echelons: production
(pg), Eq. (7), to produce different gas byproducts is based on the
areas, processing plants, and demand terminals. An overlap exists
stream yield (p).
between the two networks as the fact that crude oil contains   
associated gas. Fig. 1 depicts a schematic representation of the ygi jt + ygi jt + ygjt−1 = ygjit + ygjt ∀g, ∀ j ∈ gg (6)
i∈rg i∈n i∈ pg
network.
SCN of crude oil starts from oil reservoirs, as production areas,  
Pigjt ygi jt = ygi jt ∀g, ∀ j ∈ pg (7)
subsequently the produced oil is transported to GOSPs to separate i∈gg i∈dg
the associated gas from oil. After that, oil streams from GOSPs are
collected at the gathering centers, and then sent to oil processing Plant capacity constraints: the formulation of plant capacity
plants for stabilization and sweetening (i.e., removal of H2 S and constraints depends on the purpose of the plant either processing
other gases). Produced gas from gas reservoirs (i.e., non-associated or gathering and storing. Eqs. (8), (9), and (12) represents the
gas) and associated gas from GOSPs are collected at the gas maximum processing capacity for oil (n and po) and gas (pg), re-
gathering centers and then fed into gas processing plants. At gas spectively. While Eqs. (10) and (11), and Eqs. (13) and (14) account
processing plants, H2 S (used for sulfur production) and CO2 are for storing plants (gathering centers and demand terminals) for
removed, and methane and NGL are produced. After that, NGL is both oil and gas. Route capacity constraints: for all products and all
fractionated to its gas components (e.g., ethane, butane, propane, routes in the proposed network are represented in Eqs. (15) and
and natural gasoline). (16).

Finally, sweetened crude oil and gas-byproducts are distributed poi jt xoi jt ≤ C oj ∀o, ∀ j ∈ n (8)
to satisfy customer demand at different terminals (e.g., local, i∈ro
industrial, and international). 
poi jt xoi jt ≤ C oj ∀o, ∀ j ∈ po (9)
i∈go
3.2. Model notations

xoi jt + xo+
jt−1
≤ C oj ∀o, ∀ j ∈ go (10)
Table 2 summarizes the notations used in the mathematical i∈n
model.

xoi jt + xo+
jt−1
≤ C oj ∀o, ∀ j ∈ do (11)
3.3. Model constraints i∈ po


A set of linear constraints have proposed to determine the pgi jt ygi jt ≤ C gj ∀g, ∀ j ∈ pg (12)
feasible region of the model. They are grouped into eight blocks: i∈gg

material balance of the plants’ 2nd echelon, capacity of the plants’  

2nd and 3rd echelon, capacity of the routes connecting all the ygi jt + ygi jt + yg+
jt−1
≤ C gj ∀g, ∀ j ∈ gg (13)
i∈rg i∈n
plants’ demand at 3rd echelon plants, OPEC quota at international
terminals, CO2 emissions at gas processing plants, reserves’ sus- 
ygi jt + yg+
jt−1
≤ C gj ∀g, ∀ j ∈ dg (14)
tainability of reservoirs at 1st echelon plants, and non-negativity i∈ pg
constraints.
Material balance constraints: using the fact that the sum of xoi jt ≤ Cioj ∀ o, ∀i, ∀ j (15)
incoming and outgoing streams at any plant must be equal; con-
servation of mass through the network. Eqs. (1) and (2) represents
the mass balance for crude oil and associated gas separated at ygi jt ≤ Cigj ∀ g, ∀i, ∀ j (16)
GOSPs (n), respectively, based on yield (p) and GOR. The output Demand constraints: the produced quantities of oil and gas
streams transported to the gathering centers (go, gg). Eq. (3) rep- byproducts from the processing plants are used to satisfy demand
resents the mass balance at oil gathering centers (go), where the at the terminals, Eqs. (17) and (18). To avoid infeasibility, above
incoming stream plus inventory from the previous period (t-1) production and below production decision variables are added to
equals to the outgoing stream plus the end inventory at period (t). the constraints and the end inventory of the previous period is
The outgoing stream from (go) sent to oil processing plants (po), subtracted from the demand. Below production allows for satisfy-
where processed oil and H2 S are produced based on their yields ing part of the demand from the outside market. Whereas, OPEC
(p), Eqs. (4) and (5). quota constraint (19) specifies that the total amount of crude oil of
 
Piojt xoi jt = xojit ∀o, ∀ j ∈ n (1) all types at international terminals should not exceed the OPEC’s
i∈ro i∈go
220 A.M. Attia, A.M. Ghaithan and S.O. Duffuaa / Computers and Chemical Engineering 128 (2019) 216–227

Table 2
Notations.

1. Sets/Indices:

i, j : all nodes
ro, rg: set of (oil, gas) reservoirs; i.e., production areas
n: set of GOSPs
go, gg: set of (oil, gas) gathering centers
po, pg: set of (oil, gas) processing plants
do, dg: set of (oil, gas) demand terminals
loco, locg: a subset of (do, dg) represents (oil, gas) local depots
indo, indg: a subset of (do, dg) represents (oil, gas) industrial complexes
into, intg: a subset of (do, dg) represents (oil, gas) international terminals
o: set of crude oil types; e.g., AH Arabian Heavy Crude, AM Arabian Medium Crude, AL Arabian Light Crude, and AXL Arabian Extra Light Crude.
g: set of natural gas byproducts; includes subsets: gn natural gas, gp gas byproducts produced at processing plants, H2 S and CO2 .
t: set of time periods

2. Decision Variables:

xoi jt : amount of crude oil of type o produced in time period t transported from node i to node j, Mbarrel/period; where (i, j) ∈ (ro, n), (n, go), (go,
po), (po, do).
ygi jt : amount of natural gas of type g produced in time period t transported from node i to node j, Mcft/period; where (i, j) ∈ (rg, gg), (n, gg), (gg,
pg), (pg, dg).
xo+
jt
: crude oil production of type o in time period t above the demand at node j, Mbarrel/period; where j ∈ go, do.
xo−
jt
: crude oil production of type o in time period t below the demand at node j (i.e., satisfied from the outside market), Mbarrel/period; where
j ∈ go, do.
yg+
jt
: natural gas production of byproduct g in time period t above the demand at node j, Mcft/period; where j ∈ gg, dg.
yg−
jt
: natural gas production of byproduct g in time period t below the demand at node j (i.e., satisfied from the outside market), Mcft/period;
where j ∈ gg, dg.
D: depletion rate of crude oil and natural gas reserves.

3. Parameters:

3.1. Yield parameters:

GORoi jt : Gas-oil ratio of crude oil type o produced during time period t from reservoir i linked to GOSP j; where (i, j) ∈ (ro, n).
Piojt : Yield of crude oil of type o liberated during time period t at node i transported to node j, Percentage; where (i, j) ∈ (ro, n), (go, po).
Pigjt : yield of gas product g obtained during time period t at node i transported to node j, Percentage; where (i, j) ∈ (gg, pg).

3.2. Capacity parameters:

C oj : capacity of node j for crude oil o, Mbarrel/period; where j ∈ n, go, po, do.
C gj : capacity of node j for gas product g, Mcft/period; where j ∈ gg, pg, dg.
cioj : capacity of the route linking node i to node j of crude oil o, Mbarrel/period; where (i, j) ∈ (ro, n), (n, go), (go, po), (po, do).
cigj : capacity of the route linking node i to node j for gas product g, Mcft/period; where (i, j) ∈ (rg, gg), (n, gg), (gg, pg), (pg, dg).

3.3. Volume parameters:

Roi : amount of reserves in reservoir at node i for oil type o, Mbarrel; where i ∈ ro.
Rgi : amount of reserves in reservoir at node i for gas byproduct g, Mcft; where i ∈ rg.
Cmax : maximum amount of CO2 to be emitted to the environment in time period t, Mcft/period.
OPECQ: OPEC quota or market share per planning time period t, Mbarrel/period.

3.4. Cost parameters:

eciojt : production cost per unit of stream xoi jt , at node i during time period t, $/barrel; where (i, j) ∈ (ro, n).
ecigjt : production cost per unit of stream ygi jt , at node i during time period t, $/cft; where (i, j) ∈ (rg, gg).
pciojt : processing cost per unit of stream xoi jt , at node j during time period t, $/barrel; where (i, j) ∈ (ro, n), (go, po).
pcigjt : processing cost per unit of stream ygi jt , at node j during time period t, $/cft; where (i, j) ∈ (gg, pg).
tciojt : transportation cost per unit of stream xoi jt , from node i to node j during time period t, $/barrel; where (i, j) ∈ (ro, n), (n, go), (go, po), (po, do).
tcigjt : transportation cost per unit of stream ygi jt , from node i to node j during time period t, $/cft; where (i, j) ∈ (rg, gg), (n, gg), (gg, pg), (pg, dg).
cgjt : cost per unit of emitting CO2 to environment at plant i during time period t, $/barrel; where j ∈ pg.
wo+ jt
: penalty cost per unit for producing oil of type o above the specified demand at node j during time period t (i.e., holding cost), $/barrel;
where j ∈ go, do.
wo−
jt
: penalty cost per unit for producing oil of type o below the demand at node j during time period t (i.e., penalty of filling part of the demand
from the outside market), $/barrel; where j ∈ go, do.
wg+
jt
: penalty cost per unit for producing gas product g above the specified demand at node j during time period t (i.e., holding cost), $/cft;
where j ∈ gg, dg.
wg−
jt
: penalty cost per unit for producing gas product g below the demand at node j during time period t (i.e., penalty of filling part of the demand
from the outside market), $/cft; where j ∈ gg, dg.

3.5. Demand and price parameters:

dojt : demand at terminal j for oil type o in time period t, Mbarrel/period; where j ∈ do.
dgjt : demand at terminal j for gas byproduct g in time period t, Mcft/period; where j ∈ dg.
pr ojt : selling price per unit of oil o during time period t at demand node j, $/barrel; where j ∈ do.
pr gjt : selling price per unit of gas byproducts g during time period t at demand node j, $/cft; where j ∈ dg.
dr: discount rate per period t.
 A.M. Attia, A.M. Ghaithan and S.O. Duffuaa / Computers and Chemical Engineering 128 (2019) 216–227 221


quota or the market share. Emissions of CO2 should be within   
the range established by environmental regulations. Eq. (20) limits + pr gjt ygi jt − yg+
jt
CO2 emissions. Oil and gas are natural resources and deplete after g;(i, j )∈( pg,dg)
a certain time of consumption, Eqs. (21) and (22) are used as minimize D (26)
sustainability constraints. Where D represents the depletion rate; it
should be minimized to ensure a longer lifetime for the reserves.
4. Case study
Eventually, Eq. (23) represents the non-negativity constraint.

xoi jt − xo+
jt
+ xo−
jt
= dojt − xo+
jt−1
∀ o, ∀ j ∈ do (17)
In this section, a real HCSC from Saudi Arabia was chosen
i∈ po to elucidate the utility of the proposed MOO model, and the
 numerical results were analyzed. Also, a sensitivity analysis was
ygi jt − yg+
jt
+ yg−
jt
= dgjt − yg+
jt−1
∀ g, ∀ j ∈ dg (18) conducted to study the effect of key parameters of the model
i∈ pg
on planning decisions and to recommend some managerial in-
  sights. The network in the case study is depicted in Figs. 2 and
xoi jt + xo+
jt−1
≤ OP ECQ (19) 3; showing a representation of the figure in McMurra (2011).
o;(i, j )∈( po,into) o; j∈into The network considers only the main production areas (high
 production reservoirs), where, the real data are collected from
ygjt ≤ Cmax (20) many publications (Facts and Figures, 2015; General Authority
g∈CO2; j∈ pg for Statistics, 2019; OPEC : OPEC Basket Price, 2019; U.S. Energy
 o
Information Administration (EIA), 2019).
o;(i, j )∈(ro,n ) xi jt The network contains 20 oil reservoirs, connected to 64 GOSPs
 o
≤D (21)
o;i∈ro Rit for separation of gases, water and salt from crude oil. The pro-
 duced crude oil is transported to 8 stabilization and sweetening
g;(i, j )∈(rg,gg) ygi jt plants via pipelines. Whereas, the associated gas from GOSPs and
 g ≤D (22) the non-associated gas from 13 gas reservoirs are transported to
g;i∈rg Rit
9 gas plants for impurities removal and recovery of H2 S which
xoi jt , xo+ , xo− , ygi jt , yg+ , yg− ,D ≥ 0 (23) converted to elemental sulfur. The obtained sweet-dry gas (e.g.,
jt jt jt jt
methane) is used to satisfy industrial demand and feed stock,
and the NGL and ethane are piped to 5 fractionation plants. The
3.4. Model objective functions
outputs from the fractionation plants are ethane, butane, propane,
and natural gasoline.
The first objective considers the minimization of the total costs
The produced crude oil (i.e., AXL, AL, AM, and AH) are used to
over the planning horizon, expressed in Eq. (24). Costs include
satisfy the local demand of different refineries located in 4 regions
cost of production from reservoirs in terms 1 and 2, cost of
in the Kingdom (i.e., East, West, Middle, and South regions) and
processing in each plant in terms 3 and 4, in terms 5 and 6 cost of
satisfy the international demand as constrained by the OPEC quota.
transportation through all the existing routes, penalty cost of over-
The total proven crude oil reserves in Saudi Arabia is 268 Bbbl
or under- the specified demand at terminals and inventory cost
with 17.33% AXL, 53.31% AL, 10.99% AM, and 18.36% AH. Whereas,
at gathering centers in terms 7 and 8, and the final term accounts
the local demand of each type is 26.28% AXL, 44.11% AL, 2.99% AM,
for carbon dioxide emission cost. The total cost is discounted back
and 26.61% AH and the international demand is 10.10% AXL, 56.56%
to its present value based on the discount rate dr per planning
AL, 22.22% AM, and 11.11% AH.
period t.
The gas byproducts are used as follows: methane and ethane
Minimize T otal Cost are used to satisfy the local industrial demand and thus ensures
 the survival of the local industry. NGL, propane, butane and nat-
 −(t−1 )
 
= (1 + dr ) eciojt xoi jt + ecigjt ygi jt ural gasoline are used to satisfy the international demand, while,
t o;(i, j )∈(ro, )n g;(i, j )∈(rg,gg) propane and butane are used for domestic supply.
  The data required to run the model include the following:
+ pciojt xoi jt + pcigjt ygi jt
o;(i, j )∈(ro,n ), (n,po) g;(i, j )∈(rg,n ), ( pg) (1) GOR corresponding to crude oil type for different reservoir
     streams.
+ t ciojt xoi jt + t cigjt ygi jt + wo+
j
xo+
jt
+ wo−
j
xo−
jt (2) Crude oil composition; yield of main components (e.g., nat-
o;i; j g;i; j o; j∈(go, do)
 ural gas, H2 S) at each entity.
    (3) Natural gas composition, for instance, the yield of CO2 , H2 S,
+ wg+
j
yg+
jt
+ wg−
j
yg−
jt
+ cgjt ygjt (24) methane, and ethane.
g; j∈(gg, dg) g∈CO2; j∈ pg (4) The demand for crude oil and gas byproducts by local cus-
The second objective is to maximize the total revenue obtained tomer, local industry and international customer and the
from selling crude oil and gas byproducts; subtracting the over- corresponding selling prices.
production quantities, formulated in Eq. (25). Eq. (26) represents (5) International market share specified by the OPEC quota.
the third objective of minimizing the rate of depletion of the (6) The capacity of each entity, the capacity of routes connecting
reserves, and consequently maximizing the sustainability of the the entities and the transportation modes utilized through
crude oil and natural gas reserves. these routes.
(7) Cost elements: production and processing costs at each en-
maximize Revenue = (25) tity, transportation costs, and penalty costs of producing
above and below the demand. The penalty of producing
 above the required demand is the cost of holding the prod-
 −(t−1 )
   ucts and is estimated to be 25% of the international price.
(1 + dr ) pr ojt xoi jt − xo+
jt While, the below penalty is the international market price
t o;(i, j )∈( po,do)
plus costs of delivering the product to the demand terminal
222 A.M. Attia, A.M. Ghaithan and S.O. Duffuaa / Computers and Chemical Engineering 128 (2019) 216–227
Fig. 2. The upstream crude oil supply chain network in Saudi Arabia.
Table 3
Model statistics.
Blocks of Equations 95 Single Equations 1833
Blocks of Variables 47 Single Variables 1760
Non-Zero Elements 8433
(i.e., assuming that shortage is not allowed) and estimated
to be 125% of the international price.
4.1. Numerical results
The proposed model based on the above data was coded in
GAMS 24.1.2 r40979 and solved with commercial solver CPLEX
12.5.1.0. The tactical planning horizon is three months with one
month planning period; Table 3 summarizes the model statistics.
To generate the efficient Pareto-optima AUGMECON 2
(Improved Augmented ε -Constraint) algorithm proposed by
Mavrotas and Florios (2013) based on the ε -constraint method
was used. The first step of the algorithm is to apply a lexico-
graphic optimization to specify the range of Pareto-optima, as
follows. First, the model is optimized based on minimizing the
total cost f1 (11,487.61). Then, the total revenue is maximized f2
(36,574.97) subject to f1 value as an equality constraint and the
other eight sets of constraints. Subsequently, the depletion rate
f3 (0.001141) is minimized considering both f1 and f2 as equality
constraints and the other sets of constraints. The procedure is
repeated considering different orders of the objective functions;
the results are listed in Table 4.
The second step is to pick out the efficient points from the
pay-off matrix, by dividing the range of f2 and f3 equidistantly.
A sensitivity analysis was conducted to specify the efficient res-
olution that provides precise solutions. The analysis started by
dividing the range of (f2 , f3 ) by 25 equidistant segments (26 points)
and keep increasing resolution by 25. As expected, the execution
time increases and new efficient points added. Values of (f1 , f2 , f3 )
were normalized on the range [0, 1], then, the Euclidean distance
between the new points and the old points were calculated. The
procedure was continued until the maximum Euclidean distance
became less than 0.05; the results are shown in Table 5. As a re-
sult, a systematic search based on dividing each interval into 100
equidistant segments (i.e., 101 × 101 = 10,201 possible points) was
applied. Where, the coordination of the searched points (e2 , e3 )
represents the right-hand side of (f2 , f3 ). Also, to force the solver
to minimize the surplus and slack, eps were chosen to be 10−3 ,
which is the highest value from the range, eps ∈ [10−6 , 10−3 ],
proposed by Mavrotas and Florios (2013). Eventually, the model
was solved, where, efficient points provides a feasible solution and
is only considered as a feasible plan.
The surface of the obtained Pareto-optima is depicted in Fig. 4
with 148 efficient points. As expected, the worst plan for the total
cost and revenue M$ (34,774.49, 19,299.20) /3months occurred
at a high reserves sustainability 144.20 years (i.e., low depletion
rate). The total cost is at its highest levels because the production
is very low; consequently, the penalty of producing less than the
 A.M. Attia, A.M. Ghaithan and S.O. Duffuaa / Computers and Chemical Engineering 128 (2019) 216–227 223

Fig. 3. The natural gas supply chain network in Saudi Arabia.

Table 4
Pay-off matrix applying lexicographic optimization.

Total Cost (M$/month) Total Revenue (M$/month) Depletion Rate Sustainability (Years)

Minimizing Total Cost 11,487.61 36,574.97 0.001141 73.01

Maximizing Total Revenue 11,673.71 37,145.98 0.001141 73.01
Minimizing Depletion Rate 34,774.49 19,299.20 0.000578 144.20

Table 5
Sensitivity analysis for resolution.

Resolution (segments) Pareto Points New Points Maximum Euclidean distance

5 8 – –
25 30 22 0.087
50 57 27 0.075
75 101 44 0.063
100 148 47 0.049
125 155 7 0.042
150 163 8 0.023
175 220 57 0.021
200 426 206 0.004

Table 6
Preferred plan from the MOD model.

Total cost = M$ 11,709.04/3months Oil production = 913.94 MMbbl/3months

Revenue = M$ 36,236.58/3months 10.15 MMbbl/day
Profit = M$ 24,527.54/3months Gas production = 55,3251.39 MMcft/3months
Depletion rate = 0.00113568 6147.24 MMcft/day
Sustainability = 73.38 year

required demand is very high. Referring to Fig. 5 this plan is
non-profitable. On the other vertex of the Pareto-surface, low total
costs and high revenue cannot be achieved without affecting the
reserves sustainability. As a conclusion, to achieve the extreme of
the sustainability of the natural resources, this affects the cash
flow required for sustaining the development projects in the
Kingdom. The break-even production of oil is 6.96 MMbbld and
of gas 6570.46 MMcftd, so, to achieve profit the kingdom should
produce more of crude oil and less of natural gas.
Examining the relationship between crude oil and natural
gas productions and their effect on profit. Fig. 5 shows that oil
production has an impact on gas production because part of the
gas demand can be met from associated gas. Under high levels
of oil production, the Kingdom can reach the highest level of
224 A.M. Attia, A.M. Ghaithan and S.O. Duffuaa / Computers and Chemical Engineering 128 (2019) 216–227
Fig. 4. Efficient Pareto-optima surface.
profit and keep a sufficient amount of natural gas reserves to the
coming generations. While under this production level crude oil
reserves will deplete within 73.01 years. To sustain oil reserves the
Kingdom has a range of production until reaching the break-even
point. At this case, gas production increases to compensate for the
reduction in associated gas.
The effect of oil production levels on the total cost elements
(production, processing, and transportation costs, penalty of pro-
ducing above the demand, and penalty of producing below the
demand) is shown in Fig. 6. As oil production increases the costs
of production, processing, and transportation increase. However,
the penalty of producing below the required demand decreases,
while the penalty of producing above demand is constant at
zero (i.e., the solver forces the solution towards the minimum
depletion).
Fig. 7 demonstrates the effect of production levels of both oil
and gas on the total cost. The total cost decreases as oil production
levels increase, whether gas production is at high or low levels. In
addition, the highest and lowest levels of total costs are related to
the highest and lowest oil production levels, respectively. As the
oil production increases and the gas production is at low levels,
the total cost is low, because we can satisfy both oil and gas (from
the associated gas) demand. At the same time, the revenue from
selling crude oil allows Saudi Arabia to cover the below production
penalties of gas by-products. Whereas, the total cost is higher if
oil production decreases and gas production is increased and the
penalty cost is higher because oil demand is not met in this case.
Fig. 5. Relation between oil production, gas production, and profit.
From the set of Pareto-optima, the preferred tactical plan was
chosen using TOPSIS (Technique for Order Preference by Similarity
to Ideal Solution) based on equally weighted objectives. TOPSIS
technique selects the nearest plan to the ideal one, (Clemen and
Reilly, 2004). The values of the objective functions, the quantity
of oil production, and the quantity of gas production are listed in
Table 6.
4.2. Sensitivity analysis
In this section, we investigate the sensitivity of the HCSC
tactical decisions against four key parameters: OPEC quota, crude
oil price, crude oil demand, and CO2 emission limit, to get more
insights in managerial decisions. The preferred tactical plan was
chosen using TOPSIS based on equally weighted objectives.
4.2.1. OPEC quota
To investigate the effect of changing OPEC quota on the utiliza-
tion of the key processing plants, we evaluated the results based
on ten levels of the quota; from 1 to 10 Mbbld, incrementing by 1.
For the crude oil processing plants (sweetening and stabiliza-
tion plants) the utilization is increasing as the quota increases, as
expected, until satisfying the demand or reaching the CO2 emis-
sion limit, and then become constant, as shown in Fig. 8(a). Except
in Khursaniyah processing plant, which has a fixed utilization set
at zero. Khursaniyah feeds oil to the west region and Ju’aymah
international terminal. Where, the West region requirements are
satisfied from Safaniya and Abqaiq plants, and Ju’aymah demand
is satisfied from Khurais, Safaniya, and Qatif plants.
Whereas, the utilization of gas plants does not necessarily
increase (e.g., Berri gas plant), because as crude oil production
increases with quota, the Kingdom has enough gas from associated
gas and therefore reduces the production of non-associated gas, as
shown in Fig. 8(b). However, some gas plants are not connected
to GOSPs, therefore, the utilization of gas plants that process the
non-associated gas decreases as OPEC quota for oil increases.
4.2.2. Crude oil price
As the price increases, the solver increases the production to
satisfy the demand (Fig. 9). Intuitively, if the prices decreased,
the solver decreases the production and satisfies the demand at
under production penalties. Under production penalties, mean
it is cheaper to get the products from the outside market than
producing it domestically; below production penalty less than pro-
duction cost. But in real situations, the Kingdom have to satisfy the
demand to avoid losing the market share even under low prices.
 A.M. Attia, A.M. Ghaithan and S.O. Duffuaa / Computers and Chemical Engineering 128 (2019) 216–227 225

Fig. 6. Relation between crude oil production and total cost elements.

Fig. 7. Relation between crude oil and natural gas productions with the total cost.

Fig. 9. Effect of crude oil prices on oil and gas production levels.

Fig. 8. Effect of OPEC quota on utilization of key processing plants.

226 A.M. Attia, A.M. Ghaithan and S.O. Duffuaa / Computers and Chemical Engineering 128 (2019) 216–227
Fig. 10. Effect of crude oil demand on oil and gas production levels.
Crude oil demand: The results obtained by altering crude oil
demand on oil and gas production levels is shown in Fig. 10. The
dependency of natural gas production on crude oil production
is clear, we have to increase gas production if oil production
is decreased to compensate for the reduction in associated gas
supply.
4.2.3. CO2 emission limit
The results obtained from solving the model based on different
CO2 emission limits are shown in Fig. 11. Fig. 11(a) depicts that at
low allowable emission levels we have to produce low quantities
of both oil and gas. As the emission levels increase we can produce
more from oil and satisfy gas from the associated gas. At a high
levels, we can produce more from the non-associated gas to satisfy
the demand within the maximum allowable emission levels.
Fig. 11(b) show the trends of both total cost and total rev-
enue under different CO2 emission limits. At low emission and
production (oil and gas) levels the total cost is reaching the
Fig. 11. Effect of CO2 emission limit.
highest point as a result of below production penalties. As the
production increases penalties decrease and the revenue increases
until satisfying the demand and both curves become stable. At 150
MMcft/month of allowable CO2 emissions, Saudi Arabia can reach
the break-even point, and at 250 MMcft/month reach the highest
level of profit. At greater levels the increase in profit is almost
insignificant.
5. Conclusions and suggested future research
A deterministic multi-objective optimization model is pre-
sented for the tactical planning decisions of crude oil and natural
gas by-products supply chains. The proposed model is an attempt
to address a gap identified in the literature review. To the best
of our knowledge, this paper is the first to optimize the HCSC
in a multi-objective perspective and in an integrated framework
(oil and gas simultaneously). Another aim of the paper is to
study the trade-off among different objectives of the Saudi Arabia
HCSC, since, all the existing models are deterministic and single
objective. The results show that gas production depends on oil
production, and it is better for the two to be studied simultane-
ously. Considering the depletion rate results in a situation where it
is profitable to produce oil domestically and buy gas by-products
from the international market. The model has been shown to be
practical and offers opportunities for in-depth analysis. The pro-
posed model provides the decision maker with many production
plans compromising between environmental impact, depletion of
the natural resources, minimize the expenses and maximize the
returns. In addition, it does not take time to produce alternative
plans and provide the decision maker the chance to change any
parameter and get new plans (Pareto efficient points) within a few
seconds. For future research, instead of examining the sensitivity
of the model against some parameters, a stochastic programming
model should be formulated to account for the uncertainty of
these parameters (i.e., price, demand, GOR, …, etc). Furthermore,
the model should formulate the nonlinear behavior of some pro-
duction and processing plants; for instance, reservoir performance.
More processing units may be considered such as refineries, which
adds, oil by-products (e.g., gasoline, diesel, jet fuel, …, etc.) and
end customers (e.g., gas stations) to the model.
 A.M. Attia, A.M. Ghaithan and S.O. Duffuaa / Computers and Chemical Engineering 128 (2019) 216–227
Supplementary material
Supplementary material associated with this article can be
found, in the online version, at doi:10.1016/j.compchemeng.2019.
06.016.
References
Aizemberg, L., Kramer, H.H., Pessoa, A.A., Uchoa, E., 2014. Formulations for a prob-
lem of petroleum transportation. Eur. J. Oper. Res. 237, 82–90.
Al-Othman, W.B.E., Lababidi, H.M.S., Alatiqi, I.M., Al-Shayji, K., 2008. Supply chain
optimization of petroleum organization under uncertainty in market demands
and prices. Eur. J. Oper. Res. 189, 822–840.
Al-Qahtani, K., Elkamel, A., 2008. Multisite facility network integration design and
coordination: an application to the refining industry. Comput. Chem. Eng. 32,
2189–2202.
Al-Qahtani, K., Elkamel, A., 2009. Multisite refinery and petrochemical network de-
sign: optimal integration and coordination. Ind. Eng. Chem. Res. 48, 814–826.
Carvalho, M.C.A., Pinto, J.M., 2006a. A bilevel decomposition technique for the opti-
mal planning of offshore platforms. Braz. J. Chem. Eng. 23, 67–82.
Carvalho, M.C.A., Pinto, J.M., 2006b. An MILP model and solution technique for the
planning of infrastructure in offshore oilfields. J. Pet. Sci. Eng. 51, 97–110 Spe-
cial Issue on Petroleum Production Research in BrazilAn update on the develop-
ments in petroleum production research in Brazil.
Chen, J., Lu, J., Qi, S., 2010. Transportation network optimization of import crude
oil in China based on minimum logistics cost. In: 2010 IEEE International Con-
ference on Emergency Management and Management Sciences (ICEMMS). Pre-
sented at the 2010 IEEE International Conference on Emergency Management
and Management Sciences (ICEMMS), pp. 335–338.
Clemen, R.T., Reilly, T., 2004. Making Hard Decisions With Decision Tools Suite Up-
date Edition, first ed. Cengage Learning, Pacific Grove, CA.
Duffuaa, S.O., Al-Zayer, J.A., Al-Marhoun, M.A., Al-Saleh, M.A., 1992. A linear pro-
gramming model to evaluate gas availability for vital industries in saudi arabia.
J. Oper. Res. Soc. 43, 1035–1045.
Elkamel, A., Ba-Shammakh, M., Douglas, P., Croiset, E., 2008. An optimization ap-
proach for integrating planning and CO2 emission reduction in the petroleum
refining industry. Ind. Eng. Chem. Res. 47, 760–776.
Facts & Figures 2015. [WWW Document], n.d. Saudi Aramco. http://www.
saudiaramco.com/en/home/news-media/publications/corporate-reports/
facts- figures- 2015.html (accessed 3.19.19).
Fernandes, L.J., Relvas, S., Barbosa-Póvoa, A.P., 2013. Strategic network design of
downstream petroleum supply chains: single versus multi-entity participation.
Chem. Eng. Res. Des. 91, 1557–1587 Special Issue: Computer Aided Process En-
gineering (CAPE) Tools for a Sustainable World.
Fernandes, L.J., Relvas, S., Paula Barbosa-Póvoa, A., 2011. Strategic planning of
petroleum supply chains. In: Pistikopoulos, E.N., M.C.G., A.C.K. (Eds.), Computer
Aided Chemical Engineering, 21st European Symposium on Computer Aided
Process Engineering. Elsevier, pp. 1738–1742.
General Authority for Statistics, 2019. [WWW Document], n.d. Gen. Auth. Stat. https:
//www.stats.gov.sa/en (accessed 3.19.19).
Ghaithan, A.M., Attia, A., Duffuaa, S.O., 2017. Multi-objective optimization model for
a downstream oil and gas supply chain. Appl. Math. Model. 52, 689–708.
Gupta, V., Grossmann, I.E., 2012. An efficient multiperiod MINLP model for optimal
planning of offshore oil and gas field infrastructure. Ind. Eng. Chem. Res. 51,
6823–6840.
Guyonnet, P., Grant, F.H., Bagajewicz, M.J., 2009. Integrated model for refinery
Planning, oil Procuring, and product distribution. Ind. Eng. Chem. Res. 48,
463–482.
Hamedi, M., Zanjirani Farahani, R., Husseini, M.M., Esmaeilian, G.R., 2009. A distri-
bution planning model for natural gas supply chain: a case study. Energy Policy
37, 799–812.
227
Iakovou, E.T., 2001. An interactive multiobjective model for the strategic maritime
transportation of petroleum products: risk analysis and routing. Saf. Sci. 39,
19–29.
Iyer, R.R., Grossmann, I.E., Vasantharajan, S., Cullick, A.S., 1998. Optimal planning
and scheduling of offshore oil field infrastructure investment and operations.
Ind. Eng. Chem. Res. 37, 1380–1397.
Jiao, J.L, Zhang, J.L, Tang, Y.S, 2010. A model for the optimization of the petroleum
supply chain in china and its empirical analysis. In: 2010 International Confer-
ence on E-Business and E-Government (ICEE). Presented at the 2010 Interna-
tional Conference on E-Business and E-Government (ICEE), pp. 3327–3330.
Kim, Y., Yun, C., Park, S.B., Park, S., Fan, L.T., 2008. An integrated model of supply
network and production planning for multiple fuel products of multi-site re-
fineries. Comput. Chem. Eng 32, 2529–2535 Enterprise-Wide Optimization.
Kuo, T.-H., Chang, C.-T., 2008a. Application of a mathematic programming model
for integrated planning and scheduling of petroleum supply networks. Ind. Eng.
Chem. Res. 47, 1935–1954.
Kuo, T.-H., Chang, C.-T., 2008b. Optimal planning strategy for the supply chains of
light aromatic compounds in petrochemical industries. Comput. Chem. Eng. 32,
1147–1166.
Mavrotas, G., Florios, K., 2013. An improved version of the augmented ε -constraint
method (AUGMECON2) for finding the exact pareto set in multi-objective inte-
ger programming problems. Appl. Math. Comput. 219, 9652–9669.
McMurra, S., 2011. Energy to the world: The story of Saudi ARAMCO Volume 2, 1st
edition. ed. Aramco Services Company, Houston, Texas, USA.
Moradi Nasab, N., Amin-Naseri, M.R., 2016. Designing an integrated model for a
multi-period, multi-echelon and multi-product petroleum supply chain. Energy
114, 708–733.
Neiro, S.M.S., Pinto, J.M., 2004. A general modeling framework for the operational
planning of petroleum supply chains. Comput. Chem. Eng. 28, 871–896 FOCAPO
2003 Special issue.
Nygreen, B., Christiansen, M., Haugen, K., Bjørkvoll, T., Kristiansen, Ø, 1998. Mod-
eling Norwegian petroleum productionand transportation. Ann. Oper. Res. 82,
251–268.
OPEC, 2019. OPEC Basket Price [WWW Document], n.d. http://www.opec.org/opec_
web/en/data_graphs/40.htm (accessed 3.19.19).
Persson, J.A., Göthe-Lundgren, M., 2005. Shipment planning at oil refineries using
column generation and valid inequalities. Eur. J. Oper. Res. 163, 631–652 Supply
Chain Management and Advanced Planning.
Rocha, R., Grossmann, I.E., de Aragão, M.V.S.P., 2017. Petroleum supply planning: re-
formulations and a novel decomposition algorithm. Optim. Eng. 18, 215–240.
Rocha, R., Grossmann, I.E., Poggi de Aragão, M.V.S., 2009. Petroleum allocation at
PETROBRAS: mathematical model and a solution algorithm. Comput. Chem. Eng.
33, 2123–2133 FOCAPO 2008 – Selected Papers from the Fifth International Con-
ference on Foundations of Computer-Aided Process Operations.
Sahebi, H., Nickel, S., Ashayeri, J., 2014. Strategic and tactical mathematical pro-
gramming models within the crude oil supply chain context—a review. Comput.
Chem. Eng. 68, 56–77.
Sear, T.N., 1993. Logistics planning in the downstream oil industry. J. Oper. Res. Soc.
44, 9–17.
Ulstein, N.L., Nygreen, B., Sagli, J.R., 2007. Tactical planning of offshore petroleum
production. Eur. J. Oper. Res. 176, 550–564.
U.S. Energy Information Administration (EIA), 2019. [WWW Document], n.d. https:
//www.eia.gov/ (accessed 3.19.19).
van den Heever, S.A., Grossmann, I.E., 20 0 0. An iterative aggregation/disaggregation
approach for the solution of a mixed-integer nonlinear oilfield infrastructure
planning model. Ind. Eng. Chem. Res. 39, 1955–1971.
van den Heever, S.A., Grossmann, I.E., Vasantharajan, S., Edwards, K., 20 0 0. Inte-
grating complex economic objectives with the design and planning of offshore
oilfield infrastructures. Comput. Chem. Eng. 24, 1049–1055.
van den Heever, S.A., Grossmann, I.E., Vasantharajan, S., Edwards, K., 2001. A La-
grangean decomposition heuristic for the design and planning of offshore hy-
drocarbon field infrastructures with complex economic objectives. Ind. Eng.
Chem. Res. 40, 2857–2875.