Accepted Manuscript

Energy Efficiency Optimization in Scheduling Crude Oil Operations of Refinery

Based on Linear Programming

Naiqi Wu, Zhiwu Li, Ting Qu

PII: S0959-6526(17)31678-5

DOI: 10.1016/j.jclepro.2017.07.222

Reference: JCLP 10225

To appear in: Journal of Cleaner Production

Received Date: 13 September 2016

Revised Date: 10 March 2017

Accepted Date: 29 July 2017

Please cite this article as: Naiqi Wu, Zhiwu Li, Ting Qu, Energy Efficiency Optimization in
Scheduling Crude Oil Operations of Refinery Based on Linear Programming, Journal of Cleaner
Production (2017), doi: 10.1016/j.jclepro.2017.07.222

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form.
Please note that during the production process errors may be discovered which could affect the
content, and all legal disclaimers that apply to the journal pertain.
 ACCEPTED MANUSCRIPT

Energy Efficiency Optimization in Scheduling Crude Oil Operations of Refinery Based on Linear

Programming

Naiqi Wu1, Zhiwu Li1,* and Ting Qu2

1. Institute of Systems Engineering, Macau University of Science and Technology, Taipa, Macau
2. School of Electrical and Information Engineering, Jinan University (Zhuhai Campus), Zhuhai 519070,
China.
E-mail: [email protected] (Naiqi Wu); [email protected] (Zhiwu Li); and [email protected] (Ting
Qu)
* Corresponding author: Zhiwu Li

Abstract: For sustainable development, a refinery is required to save energy as much as possible so as to reduce

the emission of greenhouse gas. In crude oil operations, oil transportation from storage tanks to charging tanks

via a pipeline consumes a large portion of energy. It is vitally important to minimize energy consumption for this

process. Since the oil flow resistance is proportional to the square of oil flow rate, the relation between energy

efficiency and flow rate is nonlinear, which makes the problem complicated. This work addresses this important

issue by formulating a linear programming model for the considered problem such that it can be efficiently

solved. A real-world industrial case study is used to demonstrate the applications and significance of the

proposed method.

Keywords: Oil refinery, crude oil operations, scheduling, energy efficiency

I. Introduction

Facing with global and increasingly intensive market competition, great attention is paid to the scheduling

operations and control of discrete manufacturing systems [Bai et al., 2016; Chen et al., 2017a and 2017b; Qiao et

al., 2015; and Wu et al., 2012b, 2012c, and 2013] to name a few of studies made in recent years. Also, a plant in

the process industry has to be well operated such that it is competitive. It is known that, with advanced

information technology applied to modify the operations, a process plant can be made more profitable [Moro,

2003]. During the last two decades, extensive attention from both academia and industry community has been

paid to the optimization of the operations in refineries, a type of most important process industries. A refinery

can be operated in a hierarchical way with three layers: production planning, short-term scheduling, and unit

control at the upper, middle, and lower layers, respectively. With linear programming-based commercial

1
 ACCEPTED MANUSCRIPT

software tools, such as PIMS (Process Industry Modeling System) [Aspen Technology Inc., 1999] and RPMS

(Refinery and Petrochemical Modeling System) [Bonner & Moore, 1979], an optimal plan can be efficiently

found at the upper layer. Meanwhile, at the lower layer, advanced process control techniques are widely realized

for unit control in refineries, resulting in significant productivity improvement. However, thanks to the lack of

efficient techniques and software tools for short-term scheduling at the middle layer, the three layers cannot be

integrated such that a global optimum cannot be achieved and a short-term schedule has to be obtained manually

by a planner. Hence, extensive efforts have been made for developing effective and efficient techniques for

scheduling of refineries.

Since the operations in a refinery contains a discrete-event process, its short-term scheduling problem is

essentially combinatorial and NP-hard [Floudas and Lin, 2004]. Moreover, to obtain such a schedule, one needs

to define activities to be performed and sequence them simultaneously, leading to the fact that the widely used

heuristics and meta-heuristics cannot be applied since these methods require that the jobs to be scheduled should

be known in advance. Hence, mathematical programming models are adopted to formulate the scheduling

problem of a refinery with two categories, namely discrete and continuous-time representations.

By discrete-time representation, the scheduling horizon is discretized into a number of uniform slots such

that the scheduling problem can be formulated as a mixed integer linear programming (MILP) model [Shah,

1996; Lee et al., 1996; Pinto, et al., 2000; Glismann and Gruhn, 2001; Jia et al., 2003; Rejowski and Pinto,

2003; Saharidisa et al., 2009; Mendez et al., 2006; and Yuzgee et al., 2010]. To make a schedule obtained by

such a model practically applicable, the time duration for the time slots should be short enough, leading to a huge

number of discrete variables for a practical application problem. Hence, it is almost impossible to be solved by

the existing software tools such as the famous CPLEX and LINGO [Floudas and Lin, 2004; and Wu et al., 2011].

To make the problem computationally tractable and solvable, models with continuous time-representation

are developed such that the time points for the start and end of an event can be exactly found [Jia et al., 2003; Jia

and Ierapetritou, 2004; Karuppiah et al., 2008; Li et al., 2002; Li et al., 2012; Mouret et al., 2009 and 2011; and

Shah et al., 2009]. By such a model, it is indeed that the number of discrete variables can be significantly

reduced. Especially, by the priority-slot-based modeling method presented by Mouret et al. [2009 and 2011], the

problem size can be significant reduced. Nevertheless, it pays an expense of introducing non-linear constraints,

resulting in a mixed integer non-linear programming (MINLP) model. It is well-known that an MINLP problem

is very difficult to solve. Furthermore, in order to build such a model, one needs to decide the number of

operations to be performed in advance, which is unrealistic for practical applications in general [Floudas and
2
 ACCEPTED MANUSCRIPT

Lin, 2004].

In addition to different modeling methods, efforts have been made to improve computational efficiency by

developing solution methods. For example, event-based methods [Yuzgee et al., 2010; and Furman et al., 2007],

an outer-approximation algorithm [Jia et al., 2003], and a decomposition algorithm [Shah et al., 2009] to

improve the efficiency of solving such a problem. Indeed, the aforementioned solution methods can reduce the

computation requirements to some extent. However, all these methods belong to enumeration in nature. With the

NP-hard nature of the problem, they cannot resolve the computational complexity problem for real-life

applications. To avoid solving an MINLP problem, based on the model obtained by the priority-slot-based

modeling method, a two-stage algorithm is proposed to solve the MINLP problem [Mouret et al., 2009 and

2011]. At Stage 1, an MILP model is formed by removing all the non-linear constraints from the MINLP model.

This MILP problem is then solved to obtain a solution for the integer variables. Generally the solution obtained

from solving the MILP is not feasible for the original MINLP, since the non-linear constraints are removed.

Then, at Stage 2, the solution for the integer variables obtained at Stage 1 is substituted into the original MINLP

to obtain a non-linear programming (NLP) model without an integer variable. Then, the NLP is solved to obtain

a schedule. However, a solution may not be found for the NLP problem even if a feasible schedule exists for the

original MINLP problem.

Thus, to make the problem formulated by mathematical programming models solvable, in modeling, some

constraints are ignored, which, unfortunately, results in an inefficient or unrealistic schedule for real-life

scenarios [Mendez et al., 2006]. This means that there is a gap between academic research and applications. As

the best knowledge of the authors, up to now, there is no software tool of short-term scheduling of oil refineries

for practical use.

To bridge such a gap, the key is to develop a computationally efficient approach for a good solution other

than an exactly optimal solution. Notice that, with a large number of discrete variables, the solution space is very

large. On the other hand, with a large number of constraints for the problem, the feasible solution space must be

small. Thus, it is extremely difficult to solve such a problem by any enumeration method. With this observation,

a Petri net-based control-theoretic approach for crude oil operations is proposed [Wu et al., 2008a, 2009, and

2015]. By this approach, the dynamic behavior of the process is modeled with a hybrid Petri net model by

treating the plant as a hybrid system with the interaction of discrete-event and continuous processes [Wu et al.,

2007 and 2008b].

From the viewpoint of control theory, schedulability is analyzed to establish schedulability conditions for
3
 ACCEPTED MANUSCRIPT

different scenarios [Wu et al., 2011, 2008a, 2009, 2010a, 2010b, and 2016a]. These conditions determine the

feasible space, i.e., the feasibility conditions, such that the problem can be decomposed into two sub-problems:

refining scheduling and detailed scheduling. With the schedulability conditions as constraints, the refining

scheduling problem can be solved to obtain a realizable and optimal refining schedule by using linear

programming-based methods [Wu et al., 2012a and 2016b]. Then, given a realizable and optimal refining

schedule, a detailed schedule to realize it can be found in a recursive way [Wu et al., 2011, 2008a, 2009, 2010a,

2010b, and 2016a]. Accordingly, a short-term schedule for crude oil operations can be efficiently found although

it may not be globally optimal.

It should be pointed out that all the aforementioned studies mainly focus on searching for techniques to

cope with computational complexity while some objectives are optimized without taking sustainability into

account. With continuous climate change resulting from anthropogenic greenhouse gas emissions, there is a

great concern on energy saving for sustainable development. It is well-known that process industries are

characterized by high energy consumption. It is vitally important to minimize energy consumption in operating a

refinery such that greenhouse gas emissions can be reduced as much as possible. A refinery process is very

complex and energy-intensive, and significant energy can be saved if it is well-operated in the sense of energy

efficiency. Thus, it is very important to reduce energy consumption in the sense of cost reduction and sustainable

development. This work focuses on the energy saving issue in scheduling the process of crude oil operations.

In the existing studies on scheduling crude oil operations, a variety of objectives are optimized. They

include: 1) minimizing cost resulting from crude oil inventory, oil tanker waiting, and oil unloading [Lee et al.,

1996; Jia et al., 2003; and Jia and Ierapetritou, 2004]; 2) maximizing productivity [Pinto et al., 2000; and Wu et

al., 2012 and 2016b] and minimizing the number of tanks used [Pinto et al., 2000]; 3) minimizing the number of

oil type switches in oil transportation via a pipeline [Lee et al., 1996]; 4) minimizing the remaining oil in a tank

when it is unloaded [Shah, 1996]; and 5) maximizing the processing effectiveness of different oil types by

different distillers [Wu et al., 2012 and 2016b]. However, as far as the authors know, there is no research report

on how to save energy in scheduling crude oil operations, which is crucial to our society. This motivates us to

conduct this study.

As pointed out in [Wu et al., 2005], in crude oil operations, crude oil transportation from storage tanks to

charging tanks via a pipeline consumes a large portion of energy. In order to save energy in the process of crude

oil operations, it is significant to minimize the energy consumption for oil transportation. The aim of this work is

two folds: 1) to develop a method to minimize the energy consumption in oil transportation from storage tanks to
4
 ACCEPTED MANUSCRIPT

charging tanks, and 2) the proposed method should be computationally efficient such that it is practically

applicable.

Generally, in a refinery, crude oil is transported from storage tanks to charging tanks via a pipeline. When

delivering liquid material through a pipeline, the flow resistance in a pipeline is proportional to the square of

fluid velocity, i.e., is highly non-linear as pointed out by Cafaro et al. [2015]. Hence, the transportation rate of

the pipeline is not proportional to the power applied, in other words, the energy consumption is non-linear with

respect to oil flow rate. This makes the problem of optimizing the energy efficiency in oil transportation very

complicated. Thus, a computationally efficient method for optimizing energy consumption in oil transportation

should avoid non-linearity and at the same time obtain satisfactory performance.

For a long-distance pipeline, at each pumping station, there are a number of sets of machines for pumping

the oil. In practice, the oil flow rate is regulated by changing the number of sets of machines that are in

operation. Based on such an operation mode, this work tackles the energy efficiency issue and a linear

programming-based approach is proposed to minimize the energy consumption in crude oil operations. It is done

as follows: a short-term schedule is found with the maximal flow rate of the pipeline to maximize the

productivity without taking energy efficiency into account. This problem can be efficiently solved by the Petri

net-based theoretic-control approach proposed in [Wu et al., 2008a, 2009, 2012, and 2016b]. Based on the

obtained schedule, this work minimizes the energy consumption for oil transportation via a pipeline. A linear

programming-based technique is developed to solve it. By doing so, non-linearity is avoided and it is

computationally very efficient.

The remainder of this work is organized as follows. Next section briefly introduces the process of crude oil

operations and its short-term scheduling problem. Section III states the energy optimization problem in crude oil

operations and presents the linear programming formulation for it. A real-world industrial case study is given to

demonstrate the application and significance of the proposed method in Section IV. Section V concludes the

work.

II. The Process and its Short-term Schedule

Before presenting the problem discussed in this work and the method for it, we briefly introduce the

processes of a refinery. An illustrative view of general/typical oil refinery processes can be depicted in Fig. 1. A

refinery contains a number of storage tanks located at a port near the plant, a pipeline, a number of charging

tanks in the plant, a number of distillers for oil distillation, and a variety of other production units. In the

5
 ACCEPTED MANUSCRIPT

viewpoint of operations, there are two phases: 1) crude oil operations and 2) production. Crude oil arrives at the

port through oil tankers by sea. It is then unloaded into storage tanks. From the storage tanks, oil is delivered to

charging tanks through the pipeline. Charging tanks are used to feed oil to the distillers for distillation. This

process forms the first phase, called crude oil operations. The products obtained from distillation are further

processed by other production units to obtain various components. These components are blended to form the

final products, which is the production phase. A short-term scheduling problem for the process of crude oil

operations is well recognized as being the most difficult one in a refinery. That answers why we find no

implementable theory and software tools to enable industrial-size applications.

Crude oil operations Production

Other
Charging Distillers production Blend Finished
Tanker Storage Pipeline
tanks tanks units header products

Figure 1. The illustrative view of a general oil refinery.

To meet the market demands, a refinery should process a number of crude oil types with different

components. Each distiller can process some crude oil types, but not all, which in turn requires a tank (storage or

charging tank) to hold one oil type at any time. Before oil can be processed by a distiller, brine must be separated

from oil. To do so, it requires that, after filling a storage or charging tank, crude oil must stay in it for some time

before it can be discharged. This is called an oil residency time constraint. Besides, any tank cannot be charged

and discharged simultaneously. There is another requirement that a distiller must work continuously and cannot

be stopped unless there is a planned maintenance. The above requirements pose a large number of constraints on

the process of crude oil operations.

To schedule the process of crude oil operations is to decide the tasks to be performed and sequence them. A

task is a discrete event for the process. In the execution of a task, oil is delivered in a continuous way, resulting

in a hybrid system with both discrete-event and continuous processes. When a task is executed, the system is

transformed from a state to another such that a task can be seen as a control command. Thus, the scheduling

problem of crude oil operations is to determine the commands (tasks) and can be studied from a perspective of

hybrid system control as done by Wu et al. [2011, 2008a, 2009, 2010a, 2010b, 2016a, 2012, and 2016b]. By the

6
 ACCEPTED MANUSCRIPT

control-theoretic-based approach, a task is defined as follows.

Definition 2.1: A task (TS) is defined as TS = {OT, SP, DP, V, , }, where OT denotes an oil type; SP the

source from which the oil comes, DP the device to which the oil is delivered; V the amount of oil to be

processed; and  and  the start and end time points for a task.

For easy implementation and simplicity for finding a schedule, the oil flow rate in a task is set to be a

constant, i.e., f = V/( -). In crude oil operations, there are three types of TSs: UTSs for oil unloading from a

tanker to storage tanks, DTSs for oil delivering from storage tanks to charging tanks, and FTSs for oil feeding

from charging tanks to distillers. With the definition of TSs, a short-term schedule SCHD for crude oil operations

can be described as

SCHD = {UTS1, UTS2, …, UTSw, DTS1, DTS2, …, DTSx, FTS1, FTS2, …, FTSk} (2.1)

Thus, the scheduling problem of crude oil operations is to find an SCHD such that all the aforementioned

requirements and constraints are met, while some objectives are optimized. With the maximal oil flow rate of a

pipeline, such a schedule can be efficiently found by the control-theoretic-based approach to optimize

productivity and oil type processing effectiveness [Wu et al., 2008a, 2009, 2012a, and 2016b]. A schedule for a

scenario from a refinery obtained by using this approach is shown in Figs. 2 and 3, where DSi represents Distiller

i. In Fig. 2, distiller feeding schedule is given by presenting the case when a charging tank is used to feed a type

of oil into a distiller. For example, during time interval [0, T1], Charging Tank #1 is used to feed Oil Type #1

into Distiller 1. While, in Fig. 3, an oil transportation schedule from storage tanks to charging tanks is given by

presenting the case when a type of oil is charged into a charging tank. For instance, during time [0, A1], Oil Type

#1 is charged into Charging Tank #2.

TT11 TT22 TT33

DS TK
TK #1
#1 TK
TK #2
#2 TK
TK #3
#3 TK
TK #4
#4
DS11
#1
#1 #2
#2
TT44 TT55 TT66
TK
TK #5
#5 TK
TK #6
#6 TK
TK #7
#7 TK
TK #1
#1
DS
DS22 #3
#3 #4
#4
TT77 TT88 TT99 TT10
10
TK
TK #8
#8 TK
TK #9
#9 TK
TK #5
#5 TK
TK #8
#8 TK
TK #9
#9
DS
DS33 #5
#5 #6
#6 #7
#7

Time
Time (H)
(H)
00
20
20 40
40 60
60 80
80 100
100 120
120 140
140 160
160 180
180 200
200 220
220 240
240

Figure 2. The Gantt chart for distiller feeding schedule.

To make a schedule feasible, a charging tank can feed a distiller only when the oil in the charging tank is

ready for feeding. For example, the charging of Oil Type #1 into TK #3 ends at A2 as shown in Fig. 3 such that it

7
 ACCEPTED MANUSCRIPT

can be ready at T2 when TK #3 starts to feed DS1 as shown in Fig. 2. Note that, for the schedule shown in Figs. 2

and 3, TK #1 is used to feed Oil Type #1 into DS1 at time zero, which means that Oil Type #1 should be charged

into TK #1 before time zero, i.e., it is done during the last scheduling horizon.

A
A11 AA22 A
A33 A
A44 A
A55 A
A66 A
A77 A
A88
#1 #1 #7 #5 #4 #7 #2 #4
TK #2 TK #3 TK #5 TK #8 TK #1 TK #9 TK #2 TK #6

00 10 20
20 30
30 40
40 50
50 60
60 70
70 80
80 90
90 100
100 110
110 120
120 130
130 140
140 150
150 160 170 180
160 170 180
Time
Time(H)
(H)

Figure 3. The Gantt chart for oil transportation schedule via a pipeline.

By observing the schedule shown in Fig. 3, during time [A2, A3], the pipeline is stopped and no oil is

transported. Note that, in transporting oil via the pipeline, the oil flow resistance is proportional to the square of

the oil velocity and is highly non-linear to the velocity [Cafaro et al., 2015]. This implies that the higher the oil

flow rate in the pipeline is, the more energy is consumed. Thus, when there is idle time for the pipeline, one may

reduce the oil flow rate to save energy if the obtained schedule is still feasible. Also, the transportation of some

oil parcels can be delayed by reducing the oil flow rate without affecting the distiller feeding schedule. In this

way, energy consumption can be further reduced. This problem is coped with in the next section.

III. Problem Formulation and Solution Method

As aforementioned, to find a schedule for crude oil operations is to decide a series of TSs and, by the

control-theoretic-based approach, the oil delivering rate for each TS is set to be a constant. Since the scheduling

problem of crude oil operations is extremely complicated, it is difficult to efficiently find such a schedule by

optimizing productivity and energy consumption simultaneously. However, with the maximal oil transportation

rate via a pipeline, a schedule to maximize the productivity can be efficiently found. Based on such a schedule,

this section discusses how to optimize energy consumption by regulating the oil transportation rate in the

obtained DTSs.

A. Problem Statement

Given a schedule with maximal oil transportation rate for DTSs, to minimize energy consumption, we

examine whether some parcels of oil in the DTSs can be delayed without impact on the feasibility of the

schedule. If so, we can delay the transportation of some parcels by reducing the transportation rate.

A pipeline system in a refinery used to transport crude oil from storage tanks to charging tanks can be

illustrated by Fig. 4. It is composed of a pipeline and a number of pumping stations that provide the power for

8
 ACCEPTED MANUSCRIPT

the oil transportation. There are a number of sets of machines at each pumping station. The power provided by

the system is dependent on the number of sets of machines that are in operation at each station. In such a system,

given the number of sets of machines in operation at each station, the oil transportation rate can be regulated in a

range. Since the relation between the pumping power and oil rate is highly non-linear [Cafaro et al., 2015] and

the range that can be regulated is generally small. In practice, when the oil transportation rate needs to be

increased, one puts more sets of machines in operation at each pumping station. In fact, the oil transportation rate

is not proportional to the number of sets of working machines either and is highly nonlinear instead. Take a

refinery as an example. When one set of machines is used at each pumping station, the transportation rate of the

pipeline is 20,000 tons per day. When two sets of machines are used, the rate is 30,000 tons per day. If three sets

are used, it is 33,000 tons per day only.

Storage tanks
Pipeline Charging tanks

Pumping Pumping Pumping

station 1 station 2 station 3

Figure 4. Illustration of a pipeline system.

Note that the scheduling of crude oil operations is a routine job. To make a method for the routine

scheduling problem applicable to complex and practical application problems, it must be simple and

computationally efficient. With high non-linearity, it is extremely difficult to optimize the energy consumption if

the domain of oil transportation rate is in a real number space. Thus, to search for an efficient method for the

scheduling problem of crude oil operations, one must avoid the non-linearity. To do so, it is reasonable to adopt

the operational mode used in the practice, which regulates the oil transportation rate by changing the number of

sets of machines at each station. Also, given the number of sets of machines to be operated at each station, there

is a most energy-efficient oil transportation rate. Thus, it is justified that, given the number of sets of machines to

be operated, the most energy-efficient oil transportation rate is applied. Based on this analysis, we present a

simple and computationally efficient method to minimize the energy consumption in scheduling crude oil

operations as follows.

Based on the above analysis, to minimize energy consumption is to decide the oil transportation rate for

each DTS. Since the number of sets of machines at each station is limited and known, there are only a limited

number of selections on oil transportation rate for the amount oil for each DTS. To do so, we assume that there

are n selections. Then, given a DTS = {OT, SP, DP, V, , }, we divide V into n parcels V1, V2, …, and Vn such

9
 ACCEPTED MANUSCRIPT

that each parcel is delivered with different rate as shown in Fig. 5. Nevertheless, the feasibility needs to be

guaranteed, i.e., when a charging tank is scheduled to be charged, it must be emptied, and when the oil in a

charging tank is scheduled to be fed into a distiller, the oil residency time constraint must be satisfied.

Specifically, for the schedule given in Figs. 2 and 3, by regulating the oil transportation rate of DTSs, Charging

Tank #5 can be charged only after time point T4 and its charging should be ended  time units earlier than time

point T8 when Tank #5 that holds Oil Type #7 is used to feed DS3, where  is the oil residency time.

Flow
Flow rate
rate
C11 DTS
DTS 11
level
level 11
C12
C
C21
21
Flow
Flow rate
rate
C22 DTS
DTS 22
level
level 22
C1k
C
C2k
2k
Cn2
C
Cn1
n1

Flow
Flow rate
rate Cnk DTS
DTS kk
level
level nn

Figure 5. Optimization of energy consumption by regulating the flow rate of DTSs.

In summary, to minimize energy consumption for oil transportation via a pipeline, for each DTS = {OT, SP,

DP, V, , }, one needs to optimally divide V into n parcels V1, V2, …, and Vn such that they are transported with

flow rate levels 1, 2, …, and n, respectively. A linear programming-based method can be developed to achieve

this purpose.

B. A Linear Programming-based Method

Given a short-term schedule for crude oil operations obtained by the control-theoretic-based approach,

assume that there are k DTSs, each of which is performed to charge a charging tank. With a single pipeline, these

DTSs are sequenced such that DTS i+1 should be performed just after DTS i. Notice that if DTS i+1 is scheduled

to be performed immediately after DTS i is performed, then when DTS i is delayed, DTS i+1 must be delayed

too. However, if there is an interruption for the pipeline between performing DTSs i and i+1, a delay of

performing DTS i does not necessarily result in a delay of performing DTS i+1. Thus, to present the method, the

DTSs are grouped.

The DTSs obtained by the control-theoretic-based approach are divided into d groups such that, in group Gi,

there are ki DTSs with k1 + k2 + … + kd = k. We use DTSij to denote the j-th DTS in group Gi, and Aij and Bij to

denote the time points when DTSij starts to charge a charging tank and ends the charging, respectively. Note that

Aij and Bij are given by the schedule obtained by the control-theoretic-based approach, i.e., they are known. Then,

10
 ACCEPTED MANUSCRIPT

with this grouping, we have Bij = Ai(j+1), i.e., in the same group, the DTSs are performed one after another

without interruption. However, Bi(ki) < A(i+1)1 must hold. In other words, between groups Gi and G(i+1), the

pipeline is schedule to be idle for some time. Since the schedule obtained by the control-theoretic-based

approach is known, given i, whether Bi(ki) < A(i+1)1 holds or not is known too. This implies that the number of

groups is known, i.e., d is decided parameter. Based on this grouping of DTSs, we present the following

notations to formulate the considered problem.

Parameters and sets

n: the number of sets of machines usable at each pumping station;

S = {1, 2, …, n}: the set of the number of sets of machines;

d: the number of groups of DTSs;

G = {1, 2, …, d};

Nki ={1, 2, …, ki};

DTSij: the j-th DTS in Group iG and jNki that is decided by a given schedule;

Gi = {DTSi1, DTSi2, …, DTSi(ki)};

Vij: the amount of oil to be transported by performing DTSij;

TKij: the charging tank to be charged by performing DTSij that is decided in the given schedule;

Aij: the time point when TKij starts to be charged by performing DTSij as given by the schedule;

Bij: the time point when charging TKij ends as given by the schedule;

Tij: the time point when TKij charged by performing DTSij begins to feed a distiller as given by the schedule;

: the oil residency time;

fi: the most energy-effective oil transportation flow rate when i sets of machines are used at each pumping

station;

Ci: the cost coefficient when i sets of machines are used at each pumping station;

Decision variables

xijh: the amount of oil in DTSij to be transported by using the most energy-efficient flow rate with h sets of

machines being used at each pumping station, iG, j Nki, and hS;

i1: the time point when TKi1 starts to be charged by performing DTSi1 after oil transportation rate is

regulated.

Given a schedule obtained by the control-theoretic-based approach, the above listed sets and parameters are

known except Ci. To formulate the addressed problem, one needs to determine Ci. With high non-linearity, the
11
 ACCEPTED MANUSCRIPT

higher the oil transportation rate is, the more power is consumed by transporting one unit of crude oil, i.e., Ci >

Ci-1 holds. For the linear programming problem presented below, if we set Cd > Ch with d  h, by solving the

following linear programming problem, higher priority must be given to the operation mode that uses h sets of

machines than the one that uses d sets of machines. Thus, to optimize the problem addressed in this paper, we

need only set Ci > Ci-1 regardless of the actual values of Ci’s. Let  denote the nominal power in KWs when one

set of machines is in operation at each station. Then, with i sets of machines are in operation, the power is i in

KWs, and the energy consumed for delivering one ton of oil is i/fi in KWhours. However, in solving an

optimization problem, we can obtain the same result no matter we set Ci to be i/fi or i/fi. Thus, it is reasonable to

let Ci = i/fi such that we have Ci > Ci-1. Then, we formulate the problem as follows.

Minimize J =    C h xijh
i G j  N ki h S
(3.1)

Subject to

11  A11 (3.2)

 x jh = V , j N
h S
1 1j k1 (3.3)

g hS x gh
j
11 + 1 1 fh +   T1j, j Nk1 (3.4)


i1  Ai1, iG\{1} (3.5)

i1  (i-1)1 +  xi

j  N k ( i  1 ) h S
(  1)jh fh , iG\{1} (3.6)

 xijh = V , iG\{1}, j N ,
h S
ij ki (3.7)

g hS xigh
j
i1 + 1
fh +   Tij, j Nki, iG\{1} (3.8)


xijh  0 and i1  0 (3.9)

Since Ch represents the energy consumed for transporting one unit of crude oil, by Objective (3.1), the total

energy consumption is minimized by regulating oil transportation rate. Constraint (3.2) guarantees that oil

transportation can be done when a charging tank is available as specified by the given schedule. Constraint (3.3)

states the conservativeness property of crude oil in a DTS. Constraint (3.4) guarantees that the time when the oil

in the charged charging tank is ready is earlier than Tij when the charging tank is used to feed a distiller, i.e., it

guarantees that the time delay caused by regulating the oil transportation is in a permissive range such that the oil

residency time constraint is satisfied. Constraint (3.9) presents the non-negative requirement.

As above discussed, between two Groups G(i-1) and Gi, there is an idle time, or we have Ai1 > B(i-1)(k(i-1)).

However, after delaying the transportation of oil of DTSs in G(i-1), this may no longer hold. Since a pipeline
12
 ACCEPTED MANUSCRIPT

cannot be used to perform two DTSs simultaneously, a DTS in Gi can be performed only after all the DTSs in G(i-

1) have been executed. Constraints (3.5) and (3.6) state that when a DTS in Gi is performed, the pipeline is

available, and at the same time, charging tank TKi1 that is necessary for performing DTSi1 is released. Constraints

(3.7) and (3.8) have the same meaning as that of (3.3) and (3.4).

Notice that the domain of xijh’s and i1’s is real number, and the objective and constraints are linear. Hence,

this is a linear programming formulation and can be efficiently solved by commercial software tools. Notice that

a frequent switch from transporting one type of oil to another via a pipeline is very costly and the number of such

switches should be minimized in scheduling crude oil operations. The number of DTSs for an obtained schedule

is not large in general. Consequently, the proposed linear programming formulation cannot be large. Thus, the

proposed method is simple and computationally efficient. In this way, the proposed method minimize energy

consumption and at the same time it is practically applicable due to its computational efficiency.

IV.Industrial Case Study

This section uses a real-life scenario from a refinery in southern China to show the application of the

proposed method. The refinery is located at the southern China and is one of the largest refineries in China. It

has three distillers and a long-distance pipeline for delivering oil from storage tanks to charging tanks. These

distillers are designed for processing different types of oil such that multiple types of oil should be processed by

the refinery. The distance from the storage tanks to charging tanks is more than 20 kilometers, so is the pipeline.

The maximal oil processing capacity of the three distillers is 375 tons, 230 tons, and 500 tons per hour,

respectively. For the pipeline, there are three sets of machines at each pumping station. If one set, two sets, and

three sets of machines are put into operation, the corresponding most energy-efficient oil transportation rate via

the pipeline is 20,000 tons, 30,000 tons, and 33,000 tons per day (or 833.333 tons, 1250 tons, and 1375 tons per

hour), respectively.

Table 4.1: The initial state of the charging tanks

Tank Capacity (Ton) Type of oil filled Volume (Ton) Distiller feeding
Tank #129 34,000 Crude oil #3 27,000 Distiller 1
Tank #128 34,000 Crude oil #2 30,000 Distiller 2
Tank #116 34,000 Crude oil #4 27,000 Distiller 3
Tank #117 34,000 Crude oil #5 30,000
Tank #115 34,000 Crude oil #5 25,000
Tank #127 34,000
Tank #182 20,000

13
 ACCEPTED MANUSCRIPT

Tank #180 20,000

Tank #181 20,000

As a routine, the refinery needs to present a short-term schedule every 10 days. The case presented here is

one of the scenarios and a schedule is found by the control-theoretic-based method [Wu et al., 2008a, 2009, and

2012a]. For this case problem, since the total oil processing capacity is 375 + 230 + 500 = 1105 tons per hour

that is less than 1250 tons per hour by using two sets of machines at each pumping station, we can treat 1250

tons per hour as the maximal oil transportation rate via the pipeline for scheduling the process. In other words,

there are two oil transportation modes: 1) one set of machines at each station is in operation and 2) two sets of

machines at each station are in operation with f1 = 20,000/24 = 833.333 tons per hour and f2 = 30,000/24 = 1250

tons per hour, respectively. In this way, with the initial state of the charging tanks shown in Table 4.1, the

obtained schedule is shown in Figs. 6 and 7.

TT13
13 TT14
14 TT23
23
Distiller
Distiller 11 #129
#129 #127
#127 #182
#182 #129
#129
Oil
Oil #3
#3 (27000)
(27000) Oil
Oil #1
#1 (63000)
(63000)

TT11
11
TT12
12
Distiller
Distiller 22 #128
#128 #180
#180 #181
#181
Oil
Oil #2
#2 (55200)
(55200)

TT21
21
TT22
22
Distiller
Distiller 33 #116
#116 #117
#117 #115
#115 #116
#116 #117
#117
Oil
Oil #4
#4 (27000)
(27000) Oil
Oil #5
#5 (55000)
(55000) Oil
Oil #6
#6 (38000)
(38000)

Time
Time (H)
(H)
00
20
20 40
40 60
60 80
80 100
100 120
120 140
140 160
160 180
180 200
200 220
220 240
240

Figure 6. The distiller feeding schedule for the case problem.

For the obtained schedule by the control-theoretic-based method, there are nine DTSs and they form two

groups with G1 = {DTS11, DTS12, DTS13, DTS14} and G2 = {DTS21, DTS22, DTS23, DTS24, DTS25}. For this

schedule, Charging Tanks #129 with Oil #3, #128 with Oil #2, and #116 with Oil #4 have been already charged

during the last scheduling horizon such that they can be used to feed DS1, DS2, and DS3 at time zero,

respectively. For example, 34000 tons of oil #2 in tank #128 are fed into Distiller 2 first. At the same time,

40,000 tons of oil #2 are transported into tanks #180 and #181. Then, 21,200 tons of oil #2 are fed into Distiller 2

and 18,800 tons of oil #2 in tank #181 are left for the use of the next scheduling horizon. It should also be

noticed that, among the DTSs, the oil transported to Charging Tanks #128 and #127 by performing DTS24, DTS25

is not processed during the current scheduling horizon but for the next horizon and the time when it is processed

is unknown. Hence, we do not need to consider these two DTSs for energy reduction.

14
 ACCEPTED MANUSCRIPT

A11 A12 A13 A14 B14 A21 A22 A23 A24 A25
#2 #2 #1 #1 #6 #6 #1 #2 #3
#180 #181 #127 #182 #116 #117 #129 #128 #127

Time (H)
0
20 40 60 80 100 120 140 160 180 200

Figure 7. The oil transportation schedule for the case problem.

From the given schedule, by using ST to stand for storage tanks, we have DTS11 = {#2, ST, #180, 20000, 0,

16}, DTS12 = {#2, ST, #181, 20000, 16, 32}, DTS13 = {#1, ST, #127, 34000, 32, 59.2}, DTS14 = {#1, ST, #182,

20000, 59.2, 75.2}, DTS21 = {#6, ST, #116, 34000, 92.8, 120}, DTS22 = {#6, ST, #117, 34000, 120, 147.2},

DTS23 = {#1, ST, #129, 9000, 147.2, 154.4}, T11 = 130.4 hour, T12 = 217.4, T13 = 72, T14 = 162.7, T21 = 164, T22 =

232, and T23 = 216.03. Also by definition, we have C1 = 1/f1 = 0.0012 and C2 = 2/f2 = 0.0016. For this case

problem, we have  = 6 hours. From Fig. 7, we can observe that only B14 < A21 holds, which implies that there

are two groups of DTSs, i.e., d = 2. Then, we can formulate the linear programming model for the problem as

follows.

Minimize J = C1(x111+x121+x131+x141+x211+x221+x231) + C2(x112+x122+x132+x142+x212+ x222+ x232)

subject to
11  0
x111 + x112 = 20000
11 + x111/833.333 + x112/1250 + 6  130.4
x121 + x122 = 20000
11 + x111/833.333 + x112/1250 + x121/833.333 + x122/1250 + 6  217.4
x131 + x132 = 34000
11 + x111/833.333 + x112/1250 + x121/833.333 + x122/1250 + x131/833.333 + x132/1250 + 6  72
x141 + x142 = 20000
11 + x111/833.333 + x112/1250 + x121/833.333 + x122/1250 + x131/833.333 + x132/1250 + x141/833.333 +
x142/1250 + 6  162.7
21  11 + x111/833.333 + x112/1250 + x121/833.333 + x122/1250 + x131/833.333 + x132/1250 + x141/833.333
+ x142/1250
21  92.8
x211 + x212 = 34000
21 + x211/833.333 + x212/1250 + 6  164
x221 + x222 = 34000
21 + x211/833.333 + x212/1250 + x221/833.333 + x222/1250 + 6  232
x231 + x232 = 9000
21 + x211/833.333 + x212/1250 + x221/833.333 + x222/1250 + x231/833.333 + x232/1250 + 6  216.03
xijh  0 and i1  0
As pointed out above, for the obtained formulation, there are only 16 variables and 17 constraints, which is

small and easy to solve. It is solved by using CPLEX with x111 = x121 = x141 = x212 = x222 = x232 = 0, x112 = 20000,

15
 ACCEPTED MANUSCRIPT

x122 = 20000, x131 = 17000, x132 = 17000, x142 = 20000, x211 = 34000, x221 = 34000, and x231 = 9000. The obtained

schedule is illustrated by the Gantt chart in Fig. 8, where an orange bar presents that the oil is transported by

using two sets of machines, while a green bar represents that one set of machine is used. By this schedule, Aij’s

and Bij’s are modified as A12 = 16, A13 = 32, A14 = 66, B14 = 90, A21 = 92.8, A22 = 133.6, A23 = 174.4, and A24 =

185.2 such that it is feasible. Notice that A14 + 6 = 72 = T13 to guarantee that the Oil Type #1 charged to

Charging Tank #117 is usable at time T13. To do so, Oil Type #2 charged into Charging Tanks #180 and #181,

and part of Oil Type #1 charged to #117 should be transported by using two sets of machines, otherwise an

obtained schedule is infeasible.

A11 A12 A13 A14 B14 A21 A22 A23 A24 A25
#2 #2 #1 #1 #6 #6 #1 #2 #3
#180 #181 #127 #182 #116 #117 #129 #128 #127

Time (H)
0
20 40 60 80 100 120 140 160 180 200 220 240

Figure 8. Gantt chart for the optimized oil transportation schedule.

By observing Fig. 8, it can be seen that higher priority is given to the use of one set of machines since this

mode is more energy-efficient. By the schedule obtained by using the proposed method, we have J = 228.

However, by the schedule given in Fig. 7, we have J = 273.6. This implies that the objective is reduced by

(273.6-228)/273.6  16.7%, i.e., significant energy is saved.

V. Conclusions

It is commonly recognized that, to be competitive in a global market, an oil refinery should be well

operated. Since the scheduling problem of a refinery is extremely complicated and challenging, much attention

has been paid to this issue. In this research field, the main focus is on finding an efficient approach such that a

scheduling problem is computationally solvable and some objectives are optimized. In the existing methods, the

objectives include maximizing productivity, minimizing oil inventory, minimizing changeover, and so on.

However, no much work is found to take energy efficiency as an objective in scheduling an oil refinery. Due to

the great effect of greenhouse on the global climate, an enterprise is required to be sustainable, i.e., energy

efficiency is vitally important. This work addresses this critical issue in scheduling a refinery.

In our previous work, we present a control-theoretic-based approach to the scheduling problem of crude oil

operations, by which a schedule can be efficiently found. Based on the approach, this work studies the energy

efficiency problem in crude oil operations. Since oil transportation from storage tanks to charging tanks

consumes a large part of energy in crude oil operations, the objective of this work is to reduce the energy

16
 ACCEPTED MANUSCRIPT

consumption in transporting oil via a pipeline. With the relation between oil transportation rate and energy

consumption being highly non-linear, to obtain a practically applicable method, it is crucial to avoid non-

linearity. By considering the fact that, in practice, the oil transportation rate is regulated by changing the number

of sets of machines at each pumping station and, for a given number of sets of machines, there is a most energy-

efficient oil transportation rate, only a limited number of rates to be selected. Based on such an operation mode,

a method is proposed to formulate the energy efficiency problem of scheduling crude oil operations as a linear

programming problem. By this formulation, energy saving is realized by delaying the transportation of some oil

parcels if possible with lower oil flow rate. In this way, we avoid nonlinearity and integer variables in the model.

A real-life case study is presented to show the application of the proposed method. It is found that, for such a

case problem, there are less than 20 constraints and it is easy to solve. Also, significant energy can be saved.

Thus, the proposed method not only optimizes the energy consumption, but also is applicable to real-life

problems due to its computational simplicity.

Note that the linear programming formulation is an approximate model for the non-linear process. It is

useful to do comparison study with a non-linear model. Also, there are more issues for energy saving in

scheduling crude oil operations. For example, there are high fusion oil types whose fusion point is higher than

30C. Hence, when such oil types are transported from one place to another via a pipeline, they need to be

heated. Then, they are stored in tanks and cool down. When they are to be processed, they need to be heated

again. Also, when the middle products come just from a device, they are very hot. Then, they are stored in tanks

and cool down. However, when they go to the next processing step, they need to heat up. In this way, large

amount of energy is consumed, which can be greatly saved if the operations are properly scheduled. This is the

issues for our future research work.

Acknowledgments:
This work is supported in part by FDCT of Macau under Grants 065/2013/A2 and 078/2015/A3, and National
Natural Foundation of China under Grants 61273036 and 61603100.

References
[1] Aspen Technology Inc., PIMS system reference, version 11.0, Cambridge, MA, USA, 1999.
[2] L. P. Bai, N. Q. Wu, Z. W. Li, and M. C. Zhou, Optimal one-wafer cyclic scheduling and buffer space
configuration for single-arm multicluster tools with linear topology, IEEE Transactions on Systems, Man,
& Cybernetics: Systems, vol. 46, no. 10, 1456-1467, Oct. 2016.
[3] Bonner & Moore, RPMS (Refinery and Petrochemical Modeling System): a system descriptive, Houston,
17
 ACCEPTED MANUSCRIPT

TX, USA: Bonner & Moore Management Science, 1979.

[4] V. G. Cafaro, D. C. Cafaro, C. A. Mendez, and J. Cerda, MINLP model for the detailed scheduling of
refined products pipelines with flow rate dependent pumping costs, Computers & Chemical Engineering,
vol. 72, 210-221, 2015.
[5] Y. F. Chen, Z. W. Li, A. Al-Ahmari, N. Q. Wu, and T. Qu, Deadlock recovery for flexible manufacturing
systems modeled with petri nets, Information Sciences, vol. 381, pp. 290–303, Mar. 2017a.
[6] Y. F. Chen, Z. W. Li, K. Barkaoui, N. Q. Wu, M. C. Zhou, Compact supervisory control of discrete event
systems by Petri nets with data inhibitor arcs, IEEE Transactions on Systems, Man, & Cybernetics:
Systems, vol. 47, no. 2, pp. 364–379, Feb. 2017b.
[7] C. A. Floudas and X. Lin, Continuous-time versus discrete-time approaches for scheduling of chemical
processes: a review, Computers & Chemical Engineering, vol. 28, 2109-2129, 2004.
[8] K. C. Furman, Z. Jia, and M. G. Ierapetritou, A robust event-based continuous time formulation for tank
transfer scheduling, Industrial & Engineering Chemical Research, vol. 46, 9126–9136, 2007.
[9] K. Glismann and G. Gruhn, Short-term scheduling and recipe optimization of blending processes,
Computers & Chemical Engineering, vol. 25, 627-634, 2001.
[10] Z. Jia and M. Ierapetritou, Efficient short-term scheduling of refinery operations based on a continuous
time formulation, Computers & Chemical Engineering, vol. 28, 1001-1019, 2004.
[11] Z. Jia, M. Ierapetritou, and J. D. Kelly, Refinery short-term scheduling using continuous time formation:
crude oil operations, Industrial & Engineering Chemical Research, vol. 42, 3085-3097, 2003.
[12] R. Karuppiah, K. C. Furmanb, and I. E. Grossmann, Global optimization for scheduling refinery crude oil
operations, Computers & Chemical Engineering, vol. 32, 2745–2766, 2008.
[13] H. Lee, J. M. Pinto, I. E. Grossmann, and S. Park, Mixed integer linear programming model for refinery
short-term scheduling of crude oil unloading with inventory management, Industrial & Engineering
Chemistry Research, vol. 35, 1630-1641, 1996.
[14] W. K. Li, W. H. Chi, and B. Hua, Scheduling crude oil unloading, storage, and processing, Industrial &
Engineering Chemistry Research, vol. 41, 6723-6734, 2002.
[15] J. Li, R. Misener, and C. A. Floudas, Continuous-time modelling and global optimization approach for
scheduling of crude oil operations, AIChE Journal, vol. 58, no. 1, 205–226, 2012.
[16] C. A. Mendez, I. E. Grossmann, I. Harjunkoski, and P. Kabore, A simultaneous optimization approach for
off-line blending and scheduling of oil-refinery operations, Computers & Chemical Engineering, vol. 30,
no. 4, 614-634, 2006.
[17] L. F. L. Moro, Process technology in the petroleum refining industry – current situation and future trends,
Computers & Chemical Engineering, vol. 27, 1303-1305, 2003.
[18] S. Mouret, I. E. Grossmann, and P. Pestiaux, A novel priority-slot based continuous-time formulation for
crude-oil scheduling problems, Industrial & Engineering Chemistry Research, vol. 48, no. 18, 8515–8528,
2009.
[19] S. Mouret, I. E. Grossmann, and P. Pestiaux, Time representations and mathematical models for process
scheduling problems, Computers & Chemical Engineering, vol. 35, no. 6, 1038-1063, 2011.
[20] J. M. Pinto, M. Joly, and L. F. L. Moro, Planning and scheduling models for refinery operations,
Computers & Chemical Engineering, vol. 24, no. 9, 2259-2276, 2000.
[21] Y. Qiao, N. Q. Wu, and M. C. Zhou, Schedulability and scheduling analysis of dual-arm cluster tools with
wafer revisiting and residency time constraints based on a novel schedule, IEEE Transactions on Systems,
Man, & Cybernetics: Systems, vol. 45, no. 3, 472-484, Mar. 2015.
[22] R. Rejowski and J. M. Pinto, Scheduling of a multiproduct pipeline system, Computers & Chemical

18
 ACCEPTED MANUSCRIPT

Engineering, vol. 27, no. 8-9, 1229-1246, 2003.

[23] G. K. D. Saharidisa, M. Minoux, and Y. Dallery, Scheduling of loading and unloading of crude oil in a
refinery using event-based discrete time formulation, Computers & Chemical Engineering, vol. 33, 1413-
1426, 2009.
[24] N. Shah, Mathematical programming techniques for crude oil scheduling, Computers & Chemical
Engineering, vol. 20, Suppl. S1227-1232, 1996.
[25] N. Shah, G. K. D. Saharidis, Z. Jia, and M. G. Ierapetritou, Centralized–decentralized optimization for
refinery scheduling, Computers & Chemical Engineering, vol. 33, 2091-2105, 2009.
[26] N. Q. Wu, L. P. Bai, and C. B. Chu, Modeling and conflict detection of crude-oil operations for refinery
process based on controlled-colored-timed Petri net, IEEE Transactions on Systems, Man, & Cybernetics,
Part C, vol. 37, no. 4, 461-472, 2007.
[27] N. Q. Wu, L. P. Bai, and M. C. Zhou, An efficient scheduling method for crude oil operations in refinery
with crude oil type mixing requirements, IEEE Transactions on Systems, Man, & Cybernetics: Systems,
online, vol. 46, no. 3, 413-426, 2016b.
[28] N. Q. Wu, L. P. Bai, M. C. Zhou, F. Chu, and S. Mammar, A novel approach to optimization of refining
schedules for crude oil operations in refinery, IEEE Transactions on Systems, Man, & Cybernetics, Part C,
vol. 42, no. 6, 1042-1053, 2012a.
[29] N. Q. Wu, F. Chu, C. B. Chu, and M. C. Zhou, Short-term schedulability analysis of crude oil operations in
refinery with oil residency time constraint using Petri net, IEEE Transactions on Systems, Man, and
Cybernetics, Part C, vol. 38, no. 6, 765-778, 2008a.
[30] N. Q. Wu, F. Chu, C. B. Chu, and M. C. Zhou, Short-term schedulability analysis of multiple distiller crude
oil operations in refinery with oil residency time constraint, IEEE Transactions on Systems, Man, and
Cybernetics, Part C, vol. 39, no. 1, 1-16, 2009.
[31] N. Q. Wu, F. Chu, C. B. Chu, and M. C. Zhou, Hybrid Petri net modeling and schedulability analysis of
high fusion point oil transportation under tank grouping strategy for crude oil operations in refinery, IEEE
Transactions on Systems, Man, and Cybernetics, Part C, vol. 40, no. 2, 159-175, 2010a.
[32] N. Q. Wu, F. Chu, C. B. Chu, and M. C. Zhou, Tank cycling and scheduling analysis of high fusion point
oil transportation for crude oil operations in refinery, Computers & Chemical Engineering, vol. 34, no. 4,
529-543, 2010b.
[33] N. Q. Wu, C. B. Chu, F. Chu, and M. C. Zhou, Schedulability analysis of short-term scheduling for crude
oil operations in refinery with oil residency time and charging-tank-switch-overlap constraints, IEEE
Transactions on Automation Science and Engineering, vol. 8, no. 1, 190-204, 2011.
[34] N. Q. Wu, F. Chu, C. B. Chu, and M. C. Zhou, Petri net modeling and cycle time analysis of dual-arm
cluster tools with wafer revisiting, IEEE Transactions on Systems, Man, & Cybernetics: Systems, vol. 43,
no. 1, 196-207, Jan. 2013.
[35] N. Q. Wu and M. C. Zhou, Modeling, analysis and control of dual-arm cluster tools with residency time
constraint and activity time variation based on Petri nets, IEEE Transactions on Automation Science and
Engineering, vol. 9, no. 2, 446-454, Apr. 2012b.
[36] N. Q. Wu and M. C. Zhou, Schedulability analysis and optimal scheduling of dual-arm cluster tools with
residency time constraint and activity time variation, IEEE Transactions on Automation Science and
Engineering, vol. 9, no. 1, 203-209, Jan. 2012c.
[37] N. Q. Wu, M. C. Zhou, L. P. Bai, and Z. W. Li, Short-term scheduling of crude oil operations in refinery
with high fusion point oil and two transportation pipelines, Enterprise Information Systems, online, vol. 10,
no. 6, 581-610, 2016a.

19
 ACCEPTED MANUSCRIPT

[38] N. Q. Wu, M. C. Zhou, and F. Chu, Short-term scheduling for refinery process: bridging the gap between
theory and applications, International Journal of Intelligent Control and Systems, Vol. 10, No. 2, pp. 162-
174, June 2005.
[39] N. Q. Wu, M. C. Zhou, and F. Chu, A Petri net based heuristic algorithm for realizability of target refining
schedule for oil refinery, IEEE Transactions on Automation Science and Engineering, vol. 5, no. 4, 661-
676, 2008b.
[40] N. Q. Wu, M. C. Zhou, and Z. W. Li, Short-term scheduling of crude-oil operations: Petri net-based
control-theoretic approach, IEEE Robotics and Automation Magazine, vol. 22, no. 2, 64-76, June 2015.
[41] U. Yuzgee, A. Palazoglu, J. A. and Romagnoli, Refinery scheduling of crude oil unloading, storage and
processing using model predictive control strategy, Computers and Chemical Engineering, vol. 34, 1671-
1686, 2010.

20