Ocean Engineering 141 (2017) 25–34

Contents lists available at ScienceDirect

Ocean Engineering
journal homepage: www.elsevier.com/locate/oceaneng

An MILP method for optimal offshore oilfield gathering system MARK

a a,⁎ a b a
Haoran Zhang , Yongtu Liang , Jing Ma , Chen Qian , Xiaohan Yan
a
National Engineering Laboratory for Pipeline Safety/Beijing Key Laboratory of Urban oil and Gas Distribution Technology, China University of Petroleum-
Beijing, Fuxue Road No.18, Changping District, Beijing 102249, PR China
b
University of Calgary, 2500 University Drive, NW, Calgary, Alberta, Canada T2N 1N4

A R T I C L E I N F O A BS T RAC T

Keywords: Offshore oilfield gathering system plays a key role in oilfield exploration and production process, the rational
Offshore oilfield construction of the gathering system is directly related to the efficiency and profit of the offshore oilfield
Gathering pipeline system production. Thus, optimization of the whole gathering system is the key to lower the offshore oilfield production
Optimization cost, and it also has been one of the major challenges in offshore oilfield production. Many researches have been
Pipeline network topology
done on the optimization but pipeline network distribution, treating technical and integrated optimization still
remained unsolved. This paper focuses on the common offshore oilfield gathering network construction, taking
gathering radius, economic flowrate, terrain obstacles and production technic into consideration, building a
mixed integer linear programming (MILP) model, figuring out the globally optimal connection topology,
location of center platforms, pressure increment and dehydrating facilities and major parameters of each
pipeline, and contributing to further optimal offshore oilfield engineering mode. Last, this paper takes two
typical real cases, namely half-offshore and half-continent mode and offshore mode, to test the accuracy of the
proposed model. The optimal results indicate the proposed method can lower the general investment and testify
the liability and applicability of it.

1. Introduction 1970), shortest route (Knoope et al., 2014; Marcoulaki et al., 2012) and
many other sub questions. Most of the sub questions are typical non-
1.1. Background deterministic polynomial complete (NPC) models. Usually these sub
questions are coupled, and considering the direction of pipeline (Dan
Petroleum industry plays a vital role in the modern global economy, and Williams, 2015; Schwarz et al., 2015), topology, submarine
owing to the fact that it is the largest source of energy in the world geomorphology and obstacles (Haneberg, 2015; Rocha et al., 2015)
(Sahebi and Nickel, 2014). Across the intervening years, offshore and production technical, thus the offshore oilfield gathering pipeline
exploration and production have spread around the world, taking place optimization can be very complex. Moreover, the current models and
off the coasts of Africa, Asia, Central and South America, and Europe methods are unable to solve complex connections and general and
(Gramling and Freudenburg, 2006). Offshore gathering pipeline system globally optimal results. This paper focuses on the common offshore
collects, disposes and transfers the produced liquid and plays a key role oilfield gathering network construction, taking gathering radius, sub-
in the whole offshore production system (Kawsar et al., 2015; Wang marine geomorphology and obstacles, and production technic into
et al., 2017). The construction of offshore gathering system costs a huge consideration, building an MILP model, figuring out the global con-
part of the oilfield production investment (Kearney, 2010; Tsibulnikova nection topology, center platform location, pressure increment and
et al., 2015). The rational construction of it is directly related to the dehydrating facilities and major parameters of each pipeline, and
efficiency and profit of the offshore oilfield production. Thus, the contributing to further optimal offshore oilfield engineering mode.
optimization of the whole gathering system is the key to lower the
offshore oilfield production cost, and it is also one of the major 1.2. Related work
challenges in offshore oilfield production (Liu et al., 2015).
Offshore oilfield gathering pipeline system optimization contains Since the offshore oil field gathering system is complex, many
optimal well groups division, geometric center of points set location (Li scholars have carried out researches on different aspects of this issue.
et al., 2011), minimum spanning tree of graph theory (Rothfarb et al., In the work of And et al. (1998), he employed a general objective

⁎
Corresponding author.
E-mail address: [email protected] (Y. Liang).

http://dx.doi.org/10.1016/j.oceaneng.2017.06.011
Received 3 November 2016; Received in revised form 27 May 2017; Accepted 5 June 2017
0029-8018/ © 2017 Elsevier Ltd. All rights reserved.
H. Zhang et al. Ocean Engineering 141 (2017) 25–34

Nomenclature wellhead.
Rmax Maximum gathering radius.
Acronyms RP max Maximum gathering radius increment powered by pumps.

NPC problem Non-deterministic Polynomial Complete problem. Discrete parameters

NPV Net Present Value.
MIP Mixed Integer Programming. FIPx, y Platform construction binary variable. FIPx, y = 1 if (x, y) is
MILP Mixed-Integer Linear Programming. suitable for constructing a platform, or else, FIPx, y = 0 .
MINLP Mixed-Integer Non-Linear Programming. FDDx, y Onshore terminal binary variable. FDDx, y = 1 if (x, y) is
GDP Gross Domestic Product. onshore terminal, or else, FDDx, y = 0 .
FWPx, y Wellhead location binary variables. FWPx, y = 1 if (x, y) is
Sets and indices wellhead, or else, FWPx, y = 0 .
FPFx, y, r Pipeline installation condition binary variable. FPFx, y, r = 1
(x, y), (rxx, y, r , ryx, y, r ) ∈ XY Set of gird coordinate. if pipeline could start from (x, y) to the adjacent node
r , rrr ∈ R Set of connection direction. along direction r, or else, FPFx, y, r = 0 .
d∈D Set of pipe diameter.
v∈V Set of flowrate classes. Continuous variables

Continuous parameters Q Dx, y, r Flowrate from (x, y) to the adjacent node along direction r .
Q Px, y Platform receiving flowrate at (x, y).
CFx, y Platform construction cost at (x, y). Q PFx, y Free water volume separated at (x, y).
CPFv Separating facilities cost when separating liquid in volume Q DDx, y Final receiving flowrate at (x, y).
flowrate class v . Q PPx, y Flowrate at pressure increment facility (x, y).
CPFQ Dehydrating cost for a volume crude oil. Hx, y Equivalent connection length at (x, y).
CPv Pressure increment facilities cost when transferring liquid
in volume flowrate class v . Binary variables
CPPQ Pressure increment cost for a volume crude oil.
CBCx, y Center platform construction cost at (x, y). BFx, y Platform construction binary variable. BFx, y = 1 if platform
CDd Construction unit price of pipeline with diameter d . is constructed at (x, y), or else, BFx, y = 0 .
LDx, y, r Connecting distance from (x, y) to the adjacent node along BCx, y Center platform construction binary variable. BCx, y = 1 if
direction r . platform is constructed at (x, y), or else, BCx, y = 0 .
M A maximum value BPFx, y, v Dehydrating facility construction binary variable.
γ Dehydrating efficiency BPFx, y, v = 1 if dehydrating facility is constructed at (x, y),
Q PV max v Upper bound of flowrate class v . or else, BPFx, y, v = 0 .
Q PV min v Lower bound of flowrate class v . BPx, y, v Pressure increment facility construction binary variable.
Q DF max d Upper bound of economic flowrate of pipeline with BPx, y, v = 1 if pressure increment facility is constructed at
diameter d . (x, y), or else, BPx, y, v = 0 .
Q DF min d Lower bound of economic flowrate of pipeline with BDx, y, r , d BDx, y, r , d = 1 if pipeline with diameter d could start from
diameter d . (x, y) to adjacent node along direction r , or else,
Q WPx, y Output of wellhead (x, y). Q WPx, y = 0 if (x, y) is not a BDx, y, r , d = 0 .

function of MILP model that optimized a selected economic indicator considered the optimal layout problem for a pipeline of given length in
(e.g., net present value) and incorporated the nonlinear reservoir the presence of uncertainties in the failure rates of its constituent parts
performance, surface pressure constraints, and drilling rig resource (i.e. pipes and joints). Wu et al. (2007) established a mathematical
constraints. Carvalho and Pinto (2006) proposed a mixed integer optimization model for the problem of minimizing the cost of pipelines
programming (MIP) with the discrete and continuous decisions to incurred by driving the gas in a distributed non-linear network, based
maximize the net present value and took the pressure of each reservoir. on distribution gas pipeline networks which had a steady demand. The
Gupta and Grossmann (2012) solved the offshore oilfield development decision variables included the selection of the pipes’ diameter,
problem based on the multi-period non-convex mixed-integer non- pressure drops at each node of the network, and mass flow rate at
linear programming (MINLP) model aiming at obtaining the maximiz- each pipeline leg. Ruan et al. (2009) presented a model that minimized
ing total net present value (NPV) for long-term planning horizon. the total pipeline and compressors investment cost for onshore natural
As for the topological optimization of natural gas transmission gas transmission pipelines, and proposed a rank-optimization method
networks, Kabirian and Hemmati (2007) proposed an MINLP model to solve the model. Marcoulaki et al. (2012) established a nonlinear
combining installation and operating costs, to develop optimal network programming model for pipeline route optimization, which considered
structures by selecting the type and location of pipeline and compressor process constraints and was solved by stochastic optimization method.
stations. Sahebi and Nickel (2014) came up with a mixed integer model Zhang et al. (2015) developed several programming models respec-
to design the oilfield development and plan crude oil transportation tively for multiple connection modes of onshore gathering networks
problems. For the nonconvex mixed-integer nonlinear optimization and an improved genetic algorithm was used to figure out the model.
problems, Rose et al. (2016) established MINLP model for operating a Then Duan et al. (2016) proposed a two-stages method based on Zhang
single station and discussed several continuous reformulations of the et al. (2015), considering geographic factors. Despite the comprehen-
problem. However, researchers mentioned above had not considered sive coverage of network connection modes, those methods are hard to
geomorphology or process constraints. ensure the result calculated by intelligence algorithms has global
Intelligent algorithms have been applied to solve complex topology optimality. Yet it is difficult to guarantee the optimality of the results
structures of oil and gas gathering networks. Marseguerra et al. (2004) obtained by intelligent algorithms as the model scale increased (Jr

26
H. Zhang et al. Ocean Engineering 141 (2017) 25–34

et al., 2011; Vieira et al., 2010). Furthermore, intelligent algorithms center platform location, pressure increment and dehydrating facilities
could easily converge to the locally optimal solution as a result of the location, parameters and route of pipes and compares the optimal
strong randomness, thus the calculation results need to be further result with that of current algorithms.
discussed.
Some scholars have tried to deal with network structure topology 2. Methodology
optimization by heuristic algorithms. Rothfarb et al. (1970) put
forward three fundamental issues for topology optimization of offshore 2.1. Problem description
natural gas pipeline networks: (1) selection of pipe diameters in a
specified pipeline network to minimize the sum of investment and Offshore oilfield production usually sets wellhead platforms or
operation costs; (2) selection of minimum-cost network structures, center platforms at the wellheads. In this paper, the wellhead platforms
given gas-field locations and flow requirements; (3) optimal expansion can be defined as the satellite platforms which connect to the center
of existing pipeline networks to include newly discovered gas fields. platform by stellated connection mode and there is no connection
And techniques were attempted to put into practice, which were among wellhead platforms. Since the deck of the wellhead platform is
incorporated procedures for globally optimizing pipeline diameters small, the pressure increment and processing facilities are not allowed
for fixed tree structures and heuristic procedures for generating low- to be installed. The center platforms can be connected to each other to
cost structures. But geomorphology constraints were not taken into receive the produced liquid from wellhead platforms. Contrary to the
consideration. Dolan et al. (1989) combined the approach of Rothfarb wellhead platform, the deck of the center platform is larger, which can
et al. (1970) with simulated annealing and came up with a different place the pressure increment and processing facilities. Platforms are
approach to the pipeline network design problem. connected by submarine pipelines. When installing the submarine
Recently，Rosenthal (2010) studied the solution of the pipe pipes, designers should consider the submarine terrain and obstacles,
diameter problem by non-serial dynamic programming. Baumrucker install on ideal geological conditions with smooth slope and avoid
and Biegler (2010) developed a mathematical program with equili- trench and faults. Meanwhile, if two destinations are away from each
brium constraints (MPECs) approach for efficient operation of gas other, pressure increment facilities should be taken into consideration.
pipelines. Wang et al. (2012) established a mathematical model for If the produced liquid at wellheads contains a large amount of water,
topology optimization of the submarine cluster structures and pro- dehydrating facilities should be taken into consideration as well.
posed a heuristic algorithm based on the hyperlink matrix. Whereas Offshore oilfield engineering modes include half-offshore and half-
factors such as terrain and obstacles were not taken into account, and continent mode and offshore mode, as shown in Fig. 1. In the first
the proposed algorithm was confined to the specific structure optimiza- mode, offshore produced liquid will first be separated into oil and water
tion of pipe networks. Brimberg et al. (2003) built up a mixed integer and then transferred to onshore terminal for further process. Installing
programming model (MILP) for the optimization of nearshore pipeline submarine pipes cost a lot, thus this mode can only be applied to a
network arrangement and pipeline type. And the tabu search and short distance offshore production block. Accordingly, offshore produc-
variable neighborhood search were coupled together to solve the tion mode can be applied to a long distance offshore oilfield. In this
model. Dey et al. (2004) developed a risk-based maintenance model mode, FPSO is used for oil processing and storage instead of onshore
using a combined multiple-criteria decision-making and weight meth- terminal. The essence to optimize offshore oilfield gathering system is
od for offshore oil and gas pipelines. Lucena et al. (2014) dealt with the the selection and optimization of wellhead position, platform location,
route optimization of subsea gathering pipelines by establishing a network distribution, pipe diameter and length, pressure increment
mathematical programming model, considering the constraints of and dehydrating facilities location.
subsea obstacles. The evolutionary algorithm was determined to solve The model adopts the discrete mesh division by which the study
the model. Since only the line selection between two points could be area is divided into a number of sub-squares. The 16 node-type
carried out by the model, it failed to be directly applied for global connection structure is applied to build up the corresponding con-
optimization of pipeline networks. However, the heuristic algorithms straint equations, function objective and the programming model for
mentioned above have poor universal application for it only aims to the each sub-square. As shown in Fig. 2, R is the number set of connection
specific problems subjected to the model constraints. Thus, current direction between node and the subscript is represented by r while rrr
models and algorithms cannot solve the complex connection and as the opposite direction of r . XY is the coordination set of all the mesh
integral solution problem. nodes in study area. Supposing node (x′, y′) as the terminal of node
This paper builds an MILP model with submarine obstacles and (x, y) going through a distance along the direction r , define x′ = rxx, y, r
production technic as constraints for common offshore oilfield pipeline and y′ = ryx, y, r . Similarly, node (rxx, y, rrr , ryx, y, rrr ) is obtained by node (x, y)
network connection, and figures out the globally optimal topology, along the opposite direction of r .

Fig. 1. Offshore oilfield gathering system. (a) half-offshore and half-continent mode (b) offshore mode.

27
H. Zhang et al. Ocean Engineering 141 (2017) 25–34

2.3. MILP model

Based on MILP model structures, an optimization model for

offshore oilfield gathering system is proposed in this paper, which
aims to work out a offshore oilfield gathering system with the least
construction cost under the premise of production and hydraulic
constraints. Combined with specific actual data, the optimization issue
of offshore oilfield gathering system can be transformed into an MILP
mathematical model, and then solved by mature MILP solver. The
optimal construction scheme of offshore oilfield gathering system can
be obtained. To intuitively show the proposed model, the following
compact form is presented. The detailed content will be elaborated in
Section 3.

min F ( pi , pl , pr , Qn )
pi, pl, pr , Qn
s.t.φ ( pl , pr , Qn ) ≤ 0
ψ ( pi , Qp ) ≤ 0
ϕ ( pi , Qp , Qn ) = 0
ξ ( pi ) ≤ 0 (1)

Fig. 2. Mesh division and direction. where: pi represents design variables of submarine pipelines. pl
represents design variables of platforms. pr represents design variables
of processing facilities. Qn represents variables of node flowrate. Qp
represents variables of pipeline flowrate. F represents objective func-
2.2. Model requirements tion. φ represents universal constraints of platform construction. ψ
represents universal constraints of pipeline flowrate. ϕ represents
The model is formulated as MILP and the optimization is executed universal constraints of flowrate balance. ξ represents universal con-
using GUROBI. A detailed offshore oilfield gathering pipeline system straints of pipeline construction.
construction scheme can be obtained by solving the model. As seen from Eq. (1), the objective function of the optimization
Given: model of offshore oilfield gathering system is associated with pipelines,
platforms, processing facilities, and processing capacity of the system.
1. well group information: well group location, well group output, There are four types of constraints in the optimization model. The first
water cut of produced liquid type is constraints of platform construction φ , which mainly describe
2. obstacle information the construction location of platforms and the specification of proces-
3. gathering radius sing facilities. Those constraints involve pl , pr and Qn . The second type
4. construction unit price: pipe unit price (of different diameters), is constraints of pipeline flowrate ψ , which mainly describe the
platform construction unit price, dehydrating facilities unit price relationship between size of pipeline and pipeline flowrate. Those
Determine: constraints are related with pi and Qp . The third type is constraints of
5. pipeline connection mode flowrate balance ϕ , which mainly describe the flowrate of each node
6. pipeline diameter based on economic flowrate with specific pipeline connection structures and the relationship
7. pipeline flow direction between each pipeline flowrate. The last type is constraints of pipeline
8. center platform location construction ξ , which mainly describe the constraints on pipeline
Objective: connection structure due to gathering radius limit and are related with
The objective is to minimize the total costs, including pipeline pi .
costs, platform and affiliated facilities construction costs, pressure
increment and dehydrating facilities costs and the cost of pressure
3. Mathematical formulations
increment and dehydrating process, and to work out a detailed
offshore oilfield gathering pipeline system construction scheme
3.1. Objective function
under various operational and technical constraints.
In order to solve the model more efficiently, assuming:
The objective function is to work out the offshore oilfield construc-
9. To make sure the designed gathering system can work safely, this
tion scheme with lowest cost under each given constraint. Total costs
paper takes gathering radius as the model constraint. If the
include platform and its affiliated facilities construction costs and
gathering and transferring distance is less than the gathering
pipeline construction costs.
radius, this paper assumes the whole system meets the hydraulic
Platforms and their affiliated facilities construction costs include
and thermal requirements.
platform construction costs, dehydrating and pressure increment
10. During the submarine pipeline designing process, the first step is
facilities costs. The latter costs depend on the oil volume they
primarily selecting several pipe diameters based on economic
processed.
flowrates, then taking flow, mechanic and transfer conditions into
consideration and determining the exact diameter. This paper ⎛
selects the final diameter based on economic flowrates. The f1 = ∑ ∑ ⎜⎜CFx, y BFx, y + ∑ CPFv BPFx, y, v + CPFQ Q Px, y
x y ⎝ v
selected diameter should be further testified under the in sit
condition. ⎞
+ ∑ CPv BPx, y, v + CPPQ Q PPx, y + CBCx, y BCx, y⎟⎟
11. This paper assumes the pressure increment facilities have enough v ⎠
power to transfer the oil till the terminal, and the produced oil just (x, y) ∈ XY , v ∈ V (2)
can go through only one dehydrating and pressure incrementing
process. Pipeline construction costs depend on pipeline diameter and length.

28
H. Zhang et al. Ocean Engineering 141 (2017) 25–34

f2 = ∑ ∑ ∑ ∑ CDd LDx,y,r BDx,y,r,d (x, y) ∈ XY , r ∈ R, d ∈ D Boosted flowrate equals to platform receiving flowrate if there are
x y r d (3) no process facilities.
⎛ ⎞
3.2. Constraints of platform construction ⎜⎜1 − ∑ BPx,y,v⎟⎟ M + ∑ BPFx,y,v M + QPPx,y ≥ QPx,y (x, y) ∈ XY , v ∈ V
⎝ v ⎠ v

The platform construction binary variable must equal to 0 if this (17)

spot is not suitable for setting a platform. FIPx, y denotes the suitability of ⎛ ⎞
setting a platform and FIPx, y = 1 if this spot is suitable for setting a ⎜⎜∑ BPx, y, v − 1⎟⎟ M − ∑ BPFx,y,v M + QPPx,y ≤ QPx,y (x, y) ∈ XY , v ∈ V
platform. BFx, y denotes platform construction binary variable. If this ⎝ v ⎠ v
spot is suitable for setting a platform, then BFx, y = 1, or else, BFx, y = 0 . (18)
BFx, y ≤ FIPx, y (x, y) ∈ XY (4) Boosted flowrate equals to the platform receiving flowrate minus
dehydrating flowrate if there is a process facility.
There must be a platform at the well head.
⎛ ⎞ ⎛ ⎞
FWPx, y ≤ BFx, y (x, y) ∈ XY (5) ⎜⎜1 − ∑ BPx,y,v⎟⎟ M + ⎜⎜1 − ∑ BPFx,y,v⎟⎟ M + QPPx,y ≥ QPx,y − QPFx,y (x, y)
⎝ ⎠ ⎝ ⎠
The receiving flowrate must be zero if there is no platform. v v

∈ XY , v ∈ V (19)
BFx, y M ≥ Q Px, y (x, y) ∈ XY (6)
The receiving flowrate equals to the sum of produced flowrates at ⎛ ⎞ ⎛ ⎞
⎜⎜∑ BPx, y, v − 1⎟⎟ M + ⎜⎜∑ BPFx, y, v − 1⎟⎟ M + Q PPx, y ≤ Q Px, y − Q PFx, y (x, y)
each node and the influx flowrate from other nodes. ⎝ v ⎠ ⎝ v ⎠
∑ QDrxx,y,r,ryx,y,r ,rrr + (1 − BFx,y) M + Q WPx,y ≥ QPx,y (x, y), (rxx,y,r , ryx,y,r ) ∈ XY , v ∈ V (20)
r
The constraints of pressure increment facilities are the same as
∈ XY , r , rrr ∈ R (7)
those of dehydrating facilities.
∑ QDrxx,y,r,ryx,y,r ,rrr + (BFx,y − 1) M + Q WPx,y ≤ QPx,y (x, y), (rxx,y,r , ryx,y,r ) (1 − BPx, y, v ) M + Q PV max v ≥ Q PPx, y (x, y) ∈ XY , v ∈ V (21)
r

∈ XY , r , rrr ∈ R (8) (BPx, y, v − 1) M + Q PV min v ≤ Q PPx, y (x, y) ∈ XY , v ∈ V (22)

Dehydrating and pressure increment facilities must be installed on

a platform. 3.3. Constraints of pipeline and node flowrate
∑ BPFx,y,v ≤ BFx,y (x, y) ∈ XY , v ∈ V
v (9) Transferring flowrate must equal to zero if there are no pipes
between two spots.
Dehydrating flowrate must equal to zero if there are no dehydrating
facilities at this spot ∑ BDx,y,r,d M ≥ QDx,y,r (x, y) ∈ XY , r ∈ R, d ∈ D
d (23)
∑ BPFx,y,v M ≥ QPFx,y (x, y) ∈ XY , v ∈ V
v (10) There must be no pipes in between if the transferring flowrate
equals to zero.
Dehydrating flowrate equals to processing flowrate times dehydrat-
ing efficiency. Q Dx, y, r M ≥ ∑ BDx,y,r,d (x, y) ∈ XY , r ∈ R, d ∈ D
(24)
⎛ ⎞ d
⎜⎜1 − ∑ BPFx,y,v⎟⎟ M + QPx,y γ ≥ QPFx,y (x, y) ∈ XY , v ∈ V Determine transferring flowrate classes.
⎝ v ⎠ (11)
(1 − BDx, y, r , d ) M + Q DF max d ≥ Q Dx, y, r (x, y) ∈ XY , r ∈ R, d ∈ D (25)
⎛ ⎞
⎜⎜∑ BPFx, y, v − 1⎟⎟ M + Q Px, y γ ≤ Q PFx, y (x , y) ∈ XY , v ∈ V (BDx, y, r , d − 1) M + Q DF min d ≤ Q Dx, y, r (x, y) ∈ XY , r ∈ R, d ∈ D (26)
⎝ v ⎠ (12)
Final receiving flowrate of any node except onshore terminal nodes
There are different kinds of classes for dehydrating facilities, and in
must equal to zero.
virtue of their flowrate limits, the corresponding facilities should be
determined based on the real flowrates. If the platform has to construct Q DDx, y ≤ FDDx, y M (x, y) ∈ XY (27)
dehydrating facility (When ∑v BPFx, y, v = 1) and the receiving flowrate
The sum of production flowrates at wellheads and influx flowrates
Q Px, y is in class v′ (When ∑v BPFx, y, v = 1 and Q PV min v ′ ≤ Q Px, y ≤ Q PV max v ′),
from adjacent nodes equal to the sum of final receiving flowrate, the out
the processing capacity class of dehydrating facility must be equal to
v′(BPFx, y, v ′ = 1). flux flowrate of this node and the free water flowrate after dehydrating
process.
(1 − BPFx, y, v ) M + Q PV max v ≥ Q Px, y (x, y) ∈ XY , v ∈ V (13)
Q WPx, y + ∑ QDrxx,y,r,ryx,y,r ,rrr = QDDx,y + ∑ QDx,y,r + QPFx,y (x, y),
(BPFx, y, v − 1) M + Q PV min v ≤ Q Px, y (x, y) ∈ XY , v ∈ V (14) r r

Pressure increment facilities must be installed on a platform. (rxx, y, r , ryx, y, r ) ∈ XY , r , rrr ∈ R (28)

∑ BPx,y,v ≤ BFx,y (x, y) ∈ XY , v ∈ V The sum of transferring flowrates at onshore terminal nodes equals
v (15) to the sum of all production flowrates at wellheads minus the free water
flowrates at each dehydrating node.
Boosted flowrate must equal to zero if there are no pressure
increment facilities. (1 − FDDx, y ) M + ∑ ∑ Q WPx,y − ∑ ∑ QPFx,y ≥ ∑ QDrxx,y,r,ryx,y,r ,rrr (x, y),
x y x y r
∑ BPx,y,v M ≥ QPPx,y (x, y) ∈ XY , v ∈ V
v (16) (rxx, y, r , ryx, y, r ) ∈ XY , r , rrr ∈ R (29)

29
H. Zhang et al. Ocean Engineering 141 (2017) 25–34

(FDDx, y − 1) M + ∑ ∑ Q WPx,y − ∑ ∑ QPFx,y ≤ ∑ QDrxx,y,r,ryx,y,r ,rrr (x, y), increase the transferring distance of the gathering system. Equivalent
x y x y r largest connecting distance has a maximum value based on the
(rxx, y, r , ryx, y, r ) ∈ XY , r , rrr ∈ R efficacious workaround of pumps.
(30)
−RP max ≤ Hx, y (x, y) ∈ XY (41)
One node can have up to one flow direction and one route
connected with the adjacent node. Onshore terminal nodes must have Each route can only have one pipe.
no further connection.
∑ BDx,y,r,d + ∑ BDrxx,y,r,ryx,y,r ,rrr,d ≤ 1(x, y), (rxx,y,r , ryx,y,r ) ∈ XY , r , r
1 − FDDx, y ≥ ∑ ∑ BDx,y,r,d (x, y) ∈ XY , r ∈ R, d ∈ D d d
r d (31)
rr ∈ R, d ∈ D (42)
Wellhead nodes must have a further connection.

FWPx, y ≤ ∑ ∑ BDx,y,r,d (x, y) ∈ XY , r ∈ R, d ∈ D 4. Results and discussion

r d (32)
A platform has a maximum amount of further connecting pipes. The proposed method can be applied to the optimization of offshore
oilfield gathering system for half-offshore and half-continent mode and
(1 − BFx, y ) M + NP ≥ ∑ ∑ BDrxx,y,r,ryx,y,r ,rrr,d (x, y), (rxx,y,r , ryx,y,r ) ∈ XY , r offshore mode. This paper testifies the applicability and accuracy of the
r d
method by using the following two cases. MATLAB R2014a is taken as
, rrr ∈ R, d ∈ D (33)
programming environment and GUROBI Optimizer is chosen to solve
One node can only have up to one other node transferring liquid if the proposed MILP model. GUROBI Optimizer is a state-of-the-art
this node is neither wellheads nor onshore terminal. solver for mathematical programming. The solvers in GUROBI
Optimizer are designed from the ground up to exploit modern
BFx, y M + FDDx, y M + 1 ≥ ∑ ∑ BDrxx,y,r,ryx,y,r ,rrr,d (x, y), (rxx,y,r , ryx,y,r ) architectures and multi-core processors, using the most advanced
r d
implementations of the latest algorithms. There is the built-in im-
∈ XY , r , rrr ∈ R, d ∈ D (34) proved branch and bound algorithm solving MILP models, which is a
Connection binary variable must equal to zero if this spot is search algorithm for globally optimal solution. Thus, the globally
unsuitable for constructing pipes. optimal design solution of offshore oilfield gathering system can be
obtained by the proposed model and GUROBI Optimizer.
FPFx, y, r ≥ ∑ BDx,y,r,d (x, y) ∈ XY , r ∈ R, d ∈ D
d (35)
4.1. Example 1

3.4. Constraints of pipeline construction Taking a offshore 15 km × 15 km block for instance. Given 8 well
groups, each has certain amount of wellheads. Offshore drilling and
Equivalent largest connecting distance of one node equals to zero if production are usually based on platform, all wellheads are collected to
this node has no connection with other nodes. the platform as the center. As shown in Fig. 3, the grey area indicates
Hx, y ≤ ∑ ∑ BDx,y,r,d M + ∑ ∑ BDrxx,y,r,ryx,y,r ,rrr,d M (x, y) the obstacles and is unsuitable for installing pipeline and stations. The
r d r d black dots in the figure indicate the well group locations, and output of
∈ XY , r ∈ R, d ∈ D (36) well group is shown in Table 1. The red square indicates the onshore
terminal, and all produced liquid must be transferred to the onshore
Equivalent largest connecting distance of one node equals to the terminal. Costs of wellhead and center platforms are shown in Table 2.
sum of equivalent largest connecting distance of the previous node and Only the center platform can receive oil from other platforms, and
the connecting distance if there are no pressure increment facilities at there must be a platform at wellhead. Different pipeline diameters and
this node.

Hx, y ≤ Hrxx, y, r , ryx, y, r + LDrxx, y, r , ryx, y, r , rrr + ∑ BPx,y,v M

v
⎛ ⎞
+ ⎜⎜1 − ∑ BDrxx,y,r,ryx,y,r ,rrr,d ⎟⎟ M
⎝ d ⎠ (37)

Hx, y ≥ Hrxx, y, r , ryx, y, r + LDrxx, y, r , ryx, y, r , rrr − ∑ BPx,y,v M

v
⎛ ⎞
+ ⎜⎜∑ BDrxx, y, r , ryx, y, r , rrr , d − 1⎟⎟ M
⎝ d ⎠ (38)

(x, y), (rxx, y, r , ryx, y, r ) ∈ XY , r , rrr ∈ R, d ∈ D

Equivalent largest connecting distance of a wellhead node must be a

positive value if there are no pressure increment facilities attached.

Hx, y + (1 − FWPx, y ) M ≥ −∑ BPx, y, v M (x, y) ∈ XY

v (39)
Equivalent largest connecting distance of any node must be less
than the largest gathering radius.
Hx, y ≤ Rmax (x, y) ∈ XY (40)
Equivalent largest connecting distance of one node could be less if
there are pressure increment facilities at this node, which could Fig. 3. Oilfield wellhead and onshore terminal position of example 1.

30
H. Zhang et al. Ocean Engineering 141 (2017) 25–34

Table 1 Table 4
Well group output of example 1. Pressure increment and dehydrating facilities costs.

No. Output (m3/d) No. Output (m3/d) processing capacity class Ⅰ Ⅱ Ⅲ Ⅳ Ⅴ

1 2480 5 2340 Upper flowrate limit of pressure 14,000 18,000 22,000 24,000 28,000
2 1390 6 3180 increment facilities (m3/d)
3 2800 7 2460 Lower flowrate limit of pressure 10,000 14,000 18,000 22,000 24,000
4 3680 8 2310 increment facilities (m3/d)
Pressure increment facilities cost 14.57 16.03 17.48 18.94 21.86
(million dollar)
Table 2 Upper flowrate limit of 14,000 18,000 22,000 24,000 28,000
Platform costs. dehydrating facilities (m3/d)
Lower flowrate limit of 10,000 14,000 18,000 22,000 24,000
Category Cost (thousand dollar) dehydrating facilities (m3/d)
Dehydrating facilities costs 29.14 30.60 32.05 33.51 36.43
Infrastructure cost 48,663.8 (million dollar)
Unit price with depth 1457
Center platform 80,135
Table 5
The property of the produced liquid.
the corresponding economic flowrate classes are shown in Table 3.
Example Free Density / Viscosity / Initial Wax
Pressure increment and dehydrating facilities can only be installed on
water (kg m−3) (mPa s) boiling content / %
the center platform, and there must be supporting facilities such as ratio/% point /℃
increased floor space and electricity. The corresponding costs are
shown in Table 4. The processing flowrate class represents the facility Example 1 46 938.2 184.1 148 7.42
Example 2 53 937.4 174.5 142 9.38
processing capacity which keeps proportional to the flowrate. There are
five different kinds of classes for pressure increment and dehydrating
facilities, and in virtue of their flowrate limits, the corresponding
Schlumberger. The built-in black oil model (Trangenstein and Bell,
facilities should be determined based on the real flowrates, as shown in
1989) of the software was applied to simulation of oil liquid and the
constraints 13,14,21,22. The property of the produced liquid is shown
built-in Beggs-Brill equation to calculation of pressure drop. The dot
in Table 5.
pressure is shown in Fig. 6, which indicated the result can satisfy the
The results are shown in Fig. 4. The yellow dots and green dots
requirement.
respectively indicate the locations of satellite platforms and center
platforms. And the blue dots represent center platforms equipped with
4.2. Example 2
pressure increment and dehydrating facilities after optimization on the
original center platforms. The pipeline network after optimization
Taking a offshore 15 km × 15 km block for instance. Given the
meets all constraints requirements. The pipe length of each pipeline
location and output of 14 well groups as shown in Fig. 7, this block has
diameter is shown in Table 6.
rugged and complex terrain. The grey area indicates the obstacles and
The implementation result shows that the center platform should
is unsuitable for installing pipeline and stations. The black dots in the
be constructed at dot 8 and the pressure increment facilities (classⅠ)
figure indicate the well group locations, and the output of well group is
and dehydrating facilities (classⅢ) are needed. The main reason is that
shown in Table 7. The red square indicates the position of FPSO. All
the distance between the dot 8 and the terminal is so long that the
produced liquid must be transferred to the FPSO. The platform costs,
wellhead pressure is not enough, thereby needing to build the pressure
pipeline data and the pressure increment and dehydrating facilities
increment facilities. Meanwhile, it is not economic to transport the
costs are the same as example 1, shown in Tables 2–4. The property of
crude oil with free water to the terminal on account of the high water
the produced liquid is shown in Table 5.
content in the produced liquid. Therefore, the dehydrating facilities are
The results are shown in Fig. 8. The pipeline network after
necessary for the free water separation.
optimization meets all constraints requirements. The pipe length of
In the light of the existing topology of the pipeline network,
each pipeline diameter is shown in Table 8.
PIPESIM was used for hydrothermal check. The modelling interface
The implementation result shows there is no need for pressure
of PIPESIM is shown in Fig. 5. PIPESIM is a set of simulation
increment and dehydrating facilities. It is analyzed that the pipeline of
calculator of steady-state multiphase flow developed by
this block is short and the pressure increment is unnecessary as the

Table 3
Pipeline data.

31
H. Zhang et al. Ocean Engineering 141 (2017) 25–34

Fig. 6. Dot pressure of the network in example 1. Note: Pipe 8-0 denotes the pipe
between node 8 and 0 (terminal).

Fig. 4. Optimal result of offshore oilfield gathering system of case one. (For interpreta-
tion of the references to color in this figure legend, the reader is referred to the web
version of this article.)

Table 6
Pipe length of example 1.

Diameter 6 in. 8 in. 10 in. 12 in. 14 in.

Length (km) 12.1 3.0 0 7.8 0

wellhead pressure can satisfy the liquid lift to the terminal FPSO.
Meanwhile, there are dehydrating facilities on the terminal FPSO for
unified processing thus they are unnecessary to be set on platforms.
In the light of the existing topology of the pipeline network, the
Fig. 7. Oilfield wellhead and FPSO position of example 2. (For interpretation of the
PIPESIM modelling interface is shown in Fig. 9. The dot pressure is
references to color in this figure legend, the reader is referred to the web version of this
shown in Fig. 10, which indicated the result can satisfy the require- article.)
ment.
heuristic algorithm for deep-water gathering system. The problem can
be approximately perceived as that to be solved in this paper. Although
4.3. Method comparison the heuristic algorithm can quickly solve the model, it isn’t a global
search algorithm and thus the optimality of the calculational result still
According to literatures, there are few researches on optimization of need further discussion. Zhang et al. (2015) adopted an improved
offshore oilfield gathering system. The corresponding modification to
genetic algorithm to optimize the topological structure of coal-bed
other algorithms for gathering pipeline optimization can also solve this methane gathering pipeline networks. And the algorithm, which is used
kind of problem. For instance, Wang et al. (2012) put forward a

Fig. 5. Modelling interface of PIPESIM in example 1.

32
H. Zhang et al. Ocean Engineering 141 (2017) 25–34

Table 7
Well group output of example 2.

No. Output (m3/d) No. Output (m3/d)

1 1520 8 3180
2 2480 9 2460
3 6110 10 1310
4 2800 11 4890
5 3110 12 5630
6 3680 13 2460
7 2340 14 2310

Fig. 10. Dot pressure of the network in example 2.

for solving the connection mode of combination of well tandem and

valve, can be modified and applied to optimize the offshore oilfield
gathering system. The genetic algorithm is a global search algorithm
but it will easily converge to the locally optimal solution if the model
has strong nonlinearity. Duan et al. (2016) proposed a two-stages
method based on Zhang et al. (2015), considering geographic factors.
Firstly, he applied the improved genetic algorithm to solve the two-
dimensional optimal topology of pipeline networks. And then based on
this, he took the three-dimensional terrain into account and used the
ant colony algorithm for further resolution. The method can work out
the optimization considering complex terrain, but it is still easy to
converge to the locally optimal solution.
In order to testify the optimization effect of the proposed method,
this paper takes methods from Wang et al. (2012), Zhang et al. (2015)
Fig. 8. Optimal result of offshore oilfield gathering system of example 2. and Duan et al. (2016) to carry out the same example. On account of
the disadvantage of three methods referred above, there is a penalty
function added solving the problem to make the comparison more
Table 8 objectively. The calculation results are shown in Fig. 11.
Pipeline length of example 2. In example one, the proposed method can save 7.1 million dollar,
21.5 million dollar and 12.0 million dollar for investment compared
Diameter 6 in. 8 in. 10 in. 12 in. 14 in. with the others respectively. While in example two, 8.8 million dollar,
Length(km) 12.1 15.9 4.8 0 0 15.1 million dollar and 10.0 million dollar can be saved. Since the
algorithm is the global optimization, it avoids the poor applicability, the
unstable calculation results and the deviation in calculation effects in
the previous intelligence algorithm and heuristic algorithm. The
algorithm can be easily applied to solving optimization models of
offshore oilfield gathering system.

Fig. 11. Comparison of three algorithms.

Fig. 9. Modelling interface of PIPESIM in example 2.

33
H. Zhang et al. Ocean Engineering 141 (2017) 25–34

5. Conclusions Assessment of dropped object risk on corroded subsea pipeline. Ocean Eng. 106,
329–340.
Kearney, O.P., 2010. Production Taxation and Offshore Oil Investment: Evidence from
This paper proposes an MILP method to figure out the globally the Gulf of Mexico.
optimal network topology, location of center platform, pressure incre- Knoope, M.M.J., Raben, I.M.E., Ramírez, A., Spruijt, M.P.N., Faaij, A.P.C., 2014. The
influence of risk mitigation measures on the risks, costs and routing of CO2 pipelines.
ment and dehydrating facilities and diameter and route of pipeline with Int. J. Greenh. Gas Control 29, 104–124.
given wellheads locations and submarine terrain. This paper considers Li, Z., Sun, Y., Zhang, Z., Liu, J., 2011. A global optimization method based on genetic
the construction costs of platforms and pipes and establishes an MILP algorithms for gas gathering pipeline network in a gas field. Nat. Gas. Ind. 31, 86–89.
Liu, Y., Li, J., Wang, Z., Wang, S., Dong, Y., 2015. The role of surface and subsurface
model with the objective function to minimize total investment and the integration in the development of a high-pressure and low-production gas field.
constraints of submarine terrain and production process. And the Environ. Earth Sci. 73, 5891–5904.
model is solved by GUROBI. Finally, this paper takes offshore produc- Lucena, R.R.D., Baioco, J.S., Lima, B.S.L.P.D., Albrecht, C.H., Jacob, B.P., 2014. Optimal
design of submarine pipeline routes by genetic algorithm with different constraint
tion system with half-offshore and half-continent mode and offshore
handling techniques. Adv. Eng. Softw. 76, 110–124.
mode for instances, and works out the optimally global distribution of Marcoulaki, E.C., Papazoglou, I.A., Pixopoulou, N., 2012. Integrated framework for the
gathering pipeline network. The proposed methods can lower the total design of pipeline systems using stochastic optimisation and GIS tools. Chem. Eng.
construction cost when compared with that of the previous methods, Res. Des. 90, 2209–2222.
Marseguerra, M., Zio, E., Podofillini, L., 2004. Optimal reliability/availability of
and it testifies the accuracy of the proposed model. This model can uncertain systems via multi-objective genetic algorithms. IEEE Trans. Reliab. 53,
provide theoretical and technical support for further offshore oilfield 424–434.
gathering system optimization. Rocha, D.M., Cardoso, C.D.O., Borges, R.G., Baioco, J.S., Coutinho, D.D.C.E.S., Albrecht,
C.H., Jacob, B.P., 2015. Optimization of submarine pipeline routes considering slope
stability. pp. 761-776.
References Rose, D., Schmidt, M., Steinbach, M.C., Willert, B.M., 2016. Computational optimization
of gas compressor stations: minlp models versus continuous reformulations. Math.
Methods Oper. Res. 83, 409–444.
And, R.R.I., Grossmann, I.E., And, S.V., Cullick, A.S., 1998. Optimal planning and
Rosenthal, A., 2010. Dynamic programming is optimal for nonserial optimization
scheduling of offshore oil field infrastructure investment and operations. Ind. Eng.
problems. SIAM J. Comput. 11, 47–59.
Chem. Res. 37, 1380–1397.
Rothfarb, B., Frank, H., Rosenbaum, D.M., Steiglitz, K., Kleitman, D.J., 1970. Optimal
Baumrucker, B.T., Biegler, L.T., 2010. MPEC strategies for cost optimization of pipeline
design of offshore natural-gas pipeline systems. Oper. Res. 18, 992–1020.
operations. Comput. Chem. Eng. 34, 900–913.
Ruan, Y., Liu, Q., Zhou, W., Batty, B., Gao, W., Ren, J., Watanabe, T., 2009. A procedure
Brimberg, J., Hansen, P., Lin, K.W., Mladenovic, N., Breton, M., 2003. An oil pipeline
to design the mainline system in natural gas networks. Appl. Math. Model 33,
design problem. Oper. Res. 51, 228–239.
3040–3051.
Carvalho, M.C.A., Pinto, J.M., 2006. An MILP model and solution technique for the
Sahebi, H., Nickel, S., 2014. Offshore oil network design with transportation alternatives.
planning of infrastructure in offshore oilfields. J. Pet. Sci. Eng. 51, 97–110.
Eur. J. Ind. Eng. 8, 739.
Dan, A., Williams, M., 2015. GIS-based optimal route selection for oil and gas pipelines
Schwarz, L., Robl, K., Wakolbinger, W., Mühling, H., Zaradkiewicz, P., 2015. GIS Based,
in Uganda. Adv. Comput. Sci. 4, 93–104.
Heuristic Approach for Pipeline Route Corridor Selection. Springer International
Dey, P.K., Ogunlana, S.O., Naksuksakul, S., 2004. Risk‐based maintenance model for
Publishing.
offshore oil and gas pipelines: a case study. J. Qual. Maint. Eng. 10, 169–183.
Trangenstein, J.A., Bell, J.B., 1989. Mathematical structure of the black-oil model for
Dolan, W.B., Cummings, P.T., Levan, M.D., 1989. Process optimization via simulated
petroleum reservoir simulation. SIAM J. Appl. Math. 49, 749–783.
annealing: application to network design. Aiche J. 35, 725–736.
Tsibulnikova, M.R., Kupriyanova, O.S., Strelnikova, A.B., 2015. Economic assessment of
Duan, Z., Liao, Q., Wu, M., Zhang, H., Liang, Y., 2016. Optimization of pipeline network
environmental impact in the course of oil field development and production. In: IOP
structure of cbm fields considering three-dimensional geographical factors. In:
Conference. p. 012076.
Proceedings of the 11th International Pipeline Conference. Calgary, Alberta, Canada.
Vieira, I.N., Albrecht, C.H., Lima, B.S.L.P.D., Jacob, B.P., Rocha, D.M., Oliveiracardoso,
Gramling, R., Freudenburg, W.R., 2006. Attitudes toward offshore oil development: a
C.D., 2010. Towards a computational tool for the synthesis and optimization of
summary of current evidence. Ocean Coast. Manag. 49, 442–461.
submarine pipeline routes.
Gupta, V., Grossmann, I.E., 2012. An efficient multiperiod MINLP model for optimal
Wang, Y., Duan, M., Xu, M., Wang, D., Feng, W., 2012. A mathematical model for subsea
planning of offshore oil and gas field infrastructure. Ind. Eng. Chem. Res. 51,
wells partition in the layout of cluster manifolds. Appl. Ocean Res. 36, 26–35.
6823–6840.
Wang, Y., Zhang, X., Zhao, Y., Chen, H., Duan, M., Estefen, S.F., 2017. Perturbation
Haneberg, W.C., 2015. Evaluating the Effects of Input Cost Surface Uncertainty on Deep-
analysis for upheaval buckling of imperfect buried pipelines based on nonlinear pipe-
Water Petroleum Pipeline Route Optimization. Springer International Publishing.
soil interaction. Ocean Eng., 92–100.
Jr, M.H.A.D.L., Baioco, J.S., Albrecht, C.H., Lima, B.S.L.P.D., Jacob, B.P., Rocha, D.M.,
Wu, Y., Lai, K.K., Liu, Y., 2007. Deterministic global optimization approach to steady-
Cardoso, C.D.O., 2011. Synthesis and optimization of submarine pipeline routes
state distribution gas pipeline networks. Optim. Eng. 8, 259–275.
considering on-bottom stability criteria. In: ASME 2011 30th International
Zhang, H., Liang, Y., Wu, M., Qian, C., Li, K., Yan, Y., 2015. Study on the optimal
Conference on Ocean, Offshore and Arctic Engineering. pp. 307-318.
topological structure of the producing pipeline network system of CBM fields. In:
Kabirian, A., Hemmati, M.R., 2007. A strategic planning model for natural gas
International Petroleum Technology Conference. Doha, Qatar.
transmission networks. Energy Policy 35, 5656–5670.
Kawsar, M.R.U., Youssef, S.A., Faisal, M., Kumar, A., Seo, J.K., Paik, J.K., 2015.

34