The Optimal Gasoline Blending into Romanian Refinery

Case Study

Cristian Patrascioiu and Nicolae Nicoleta

Petroleum-Gas University of Ploiesti, Bucuresti 39 Av., Ploiesti, Romania

Keywords: Gasoline Blending, Mix Formula, Linear Optimization, Objective Function, Restrictions, MATLAB

Optimization, Industrial Optimization Software.

Abstract: The paper presents the research into Romanian industrial in line gasoline blending. The paper is structured

in three parts. First part presents an industrial bending system of a Romanian refinery. The second part

contents a research of the blending mathematical model for petroleum products. The last part is a case study,

regarding an industrial problem of the gasoline blending in a Romanian refinery. The study has reported

similitude and differences between the optimal solution calculated by Blend Optimization and Supervisory

System (BOSS) and the optimal solution obtained by using the author’s model.

1 INTRODUCTION

Refineries are not producing final products

according with quality norms, it’s producing

different components with various chemical

specifications from whom is produced the final

products. In this category is included the commercial

gasoline. The actual technology used to produce a

commercial gasoline presents two main processes:

tank blending and in line blending.

In order to control these blending processes, the

authors detect two types of issues. The first problem

is the automation system with which is controlled

the blending process. In literature there are presented

various automation systems used for petroleum

products blending (Hydrocarbon Processing 1987;

1989; 1991; 1995).

The second problem is the blend recipe. This can

be determinate based on experiments and is

unchangeable, or it can be determinate by

calculations, based on a mathematical model. In the

second variant, a particular interest it is represented

by a test of an optimum blend recipe. The Romanian

refineries have automation systems which can

provide a control of this blending process, and

software system which can determine an optimal

recipe to produce commercial gasoline

(http://iom.invensys.com).

In this context, authors formed a blending

process in order to obtain a commercial gasoline

and, by solving this, they could compare the numeric

results derived from the blending model purposed

with dates generated by the software of the

optimization system used in refineries. The

validation of this system software is the objective of

this article.

2 INDUSTRIAL SYSTEM FOR IN

LINE BLENDIG OF

PETROLEUM PRODUCTS

The in line blending is a process where the entire

components are blended simultaneously, in a

common pipeline, and then send in a final product

tank. The advantages of this method are (Manescu

1970); (Lambert 2006):

- significant reduction of the blending process

time;

- realized the blend in a closed and pressurized

system;

- eliminate the blends in other tanks;

- control the blending process during cycle

mixing, with on line analyzers characteristics;

- the possibility to obtain batches according with

optimum recipes.

Figure 1 presents an industrial system used for in

line blending of petroleum products in a refinery

from Romania. This system has two subsystems: one

subsystem for the automation of the in line blending

527

Patrascioiu C. and Nicoleta N..

The Optimal Gasoline Blending into Romanian Reﬁnery - Case Study.

DOI: 10.5220/0004146405270532

In Proceedings of the 9th International Conference on Informatics in Control, Automation and Robotics (ICINCO-2012), pages 527-532

ISBN: 978-989-8565-22-8

 2012 SCITEPRESS (Science and Technology Publications, Lda.)

process and the other subsystem used to calculate

the optimal blending recipes.

Figure 1: Industrial system for petroleum products in line

blending.

2.1 The Industrial Blending Control

System

The blending control system includes the following

control elements: regulatory control system for

flows composed by flow transmitter with outlet

pulse and regulatory control valve, temperature and

pressure measurement system, components

specification measurement system and final product

characteristics measurement system. The

measurement system used for blending components

and final product is related to the NIR analyzer

(http://www.unityscientific.com). The NIR analyzers

can measure many characteristics, for large

petroleum products diversity. Some of

characteristics of the NIR analyzer are:

− API density;

− distillation temperature at 10%, 50% and 90% ;

− research octane number (RON) and motor octane

number (COM);

− aromatic content;

− Reid pressure vapors;

− benzene and olefins content.

2.2 Optimization Software System

The name of the Optimization subsystem of the in

line blending for petroleum products is Blend

Optimization and Supervisory System (BOSS) and

this is used in the refinery to mix an optimum recipe

for crude oil and to obtain optimum recipes for

gasoline and diesel products. The optimization

system is using a linear programming to make the

optimum recipe for gasoline blending.

The main objective is to minimized blend costs

and quality giveaway of the final product. The

structure of the optimal control system – BOSS is

illustrated in figure 2.

3 MATHEMATICAL

MODELLING OF THE

GASOLINE BLENDING

PROCESS

In scope to validate the optimization system BOSS,

the authors have studied the mathematical model of

the gasoline in line blending process and they have

solved the linear optimization problem.

Figure 2: The optimal control system BOSS structure.

ICINCO2012-9thInternationalConferenceonInformaticsinControl,AutomationandRobotics

528

3.1 Mathematical Model

The gasoline in line blending is formulated so: a

refinery has n types of gasoline,

n,,i,b

K1= . Each

type of gasoline is characterized by the next

properties: research octane number o

; density d

;

distillation index r

and components prices c

. The

scope is to obtain a commercial gasoline with a

minimum price, in quantity of

b tones, with

research octane number

o , density d and distillation

index

The model purpose by the authors contains an

objective function and a restriction array. The

objective function has a financial order and with a

next formula:

∑

xcF

(1)

where x

represents the quantity from component i

witch is included in blend.

The restriction array contains the next calculation

formulas for estimation and checks the next gasoline

blend properties:

• Research octane number

∑∑

≥

xoxo

(2)

• distillation index

∑∑

≥

xrxr

(3)

• density

∑

≥

(4)

• components quantity

n,,,i,bx

K21=≤

(5)

• available quantity

≥

∑

(6)

• non-negativity condition of the variables

n,...,,i,x

210 =≥

(7)

3.2 The Solving Model

To solve the linear optimization problem, the

authors have used the MATLAB function lp,

function which determines the minimum point for a

multivariable objective function, for who, the

restriction are presented in a matrix form (Ghinea

1997); (Patrascioiu, 2005). The optimization

problem has the next formula:

()

[]

⎥

⎦

⎤

⎢

⎣

⎡

×=

min

cccF

X ;

(8)

⎥

⎦

⎤

⎢

⎣

⎡

≤

⎥

⎦

⎤

⎢

⎣

⎡

⎥

⎦

⎤

⎢

⎣

⎡

mnmm

aaa

KKKK

22221

11211

;

(9)

and

n,,ixxx

imax,iimin,

K1=

≤

(10)

From these four variants of the lp function

utilization, the authors have selected the variant

where the restriction array contains equality

restrictions and inequality restrictions, subject to

have this formula:

x=lp(c,a,b,VMI,VMS,x0,k);

The description of these variables from this lp

function is:

x - variables row vector;

c - row vector coefficients;

a - restrictions coefficients matrix;

b - row vector of restrictions free element;

x0 - initial values of variables vector;

VMI-

imin,

x values vector, associated to simple

skirts restrictions;

VMS-

imax,

values vector, associated to simple

skirts restrictions;

k – equality restrictions number from the

restriction array.

4 CASE STUDY

The authors have analyzed the optimal software

system BOSS from Invensis Corp. used into

Romanian refinery. The case study developers the

nest problems:

– the industrial mixing problem;

– the mathematical modeling;

TheOptimalGasolineBlendingintoRomanianRefinery-CaseStudy

529

– the comparison between the BOSS results and

the author’s model results.

The validation of the software system BOSS has

been the objective of the case study.

4.1 The Industrial Mixing Problem

The industrial mixing problem is the production of

4000 t commercial gasoline at a minimum price with

the next properties:

− Research octane number to be minimum 94.8;

− Density at 15°C to be maximum 0.7673g/cm

;

− Benzene content to be maximum 0.795 ppm;

− Sulfur content to be maximum 7.4 ppm;

− Final boiling temperature to be maximum 202°C.

To produce this type of commercial gasoline there

are available eight components with the properties

presented in table 1. The quantity of i component

which is mixed is x

4.2 The Industrial and the Solving

Model Problem

The application of the general model presented in

the section 3 and in according to the conditions

formulated in table 1 conduces to mathematic model

of the commercial gasoline production, respectively

the relations (11) – (12).

The restrictions used by the authors system have

the next components:

– Restriction (2) relates to research octane number.

– Restriction (4) associated to density.

– Restriction developed by the mixing benzene

content. This restriction is an additive expression.

– Restriction what relates the sulfur content of the

commercial gasoline. This restriction is an additive

expression, too.

– Restriction associated to total quantity of the

commercial gasoline.

– The restriction group (5) refers to the available

quantity of each gasoline component.

– The restriction group (7) associated to non-

negativity condition of each variable.

Using the MATLAB function lp, they is obtained

the optimal solution

⎥

⎦

⎤

⎢

⎣

⎡

3184

3761

5160

42105

2357

3431

opt

X .

This solution satisfies the restrictions system (12). In

same mode, all the restriction of the commercial

gasoline are satisfied, table 2. This fact reflects the

robust character of the optimization algorithm and

the validity of the mathematical model.

4.3 Comparison between the BOSS

Results and the Author’s Model

Results

The authors have tested the BOSS optimization

system for the same example presented in section

4.1. The quantities of the eight gasoline components

are presented in table 3.

87654321

8446414655160421059561457534319528 x.x.x.x.x.x.x.x.Fmin +

++=

;

(11)

⎪

⎩

⎪

⎨

⎧

≥

≤

=+++++++

≤+++++++

≥+++++++

,81, = i 0, x

446.8 x

1465.4 x

160.5 x

2105.4 x

561.9

575.4 x

431.3 x

528.9 x

4000

80800364794792112101043135787

29601511510130145

31800200024017200376010620

52163912591340126611611307140815311

379200961101157915105610274194

87654321

xxxxxxxx

xx.x.xxxx.x.

xxxxxxxx.

x.xxx.x.x.xx.

x,x,x,x,x,x,x.x,

xxxx.x.x.xx.

(12)

ICINCO2012-9thInternationalConferenceonInformaticsinControl,AutomationandRobotics

530

Table 1: The properties of the mixing components of the gasoline commercial product.

Component of

gasoline

Research octane

number

Density

[g/cm

]

Benzene

concentration [%

vol]

Sulfur

concentration

[ppm]

Final

boiling point

[°C]

Quantity

[ t ]

Cost

[RON/t]

1 94.1 0.653 0.062 5.4 87.7 528.9 677.8

2 74.0 0.710 1.0 1 135.3 431.3 617.1

3 102.6 0.764 0.376 30 104 575.4 735

4 105.5 0.861 0.720 1 210 561.9 647.8

5 91.7 0.790 1.240 10 211 2105.4 677.8

6 115.0 0.746 0 15 79.4 160.5 1032.8

7 110.0 0.793 0 1 79.4 1465.4 923

8 96.0 0.610 0.02 15 36 446.8 677.8

Table 2: The values of the restriction functions.

Restriction Value

Quantity

0 431.3 357.2 0 2105.4 160.5 761.3 184.3 4000

+++ +++=

Research octane number

95.38 94.8≥

Density

0.7673 0.7673≥

Benzene concentration

0.795 0.795≥

Sulfur concentration

9.53 7.4≥

Final boiling point

154.89 202≤

The optimal BOSS solution is different in accord to

MATLAB optimal solution, especially to the

component 1 and 4. For the component 7 and 8, the

calculated quantity with MATLAB program is 50%

in rapport to the quantity calculated by BOSS

software. However, the quantity of the component 2,

3, 5 and 6 are identical values.

The source of the differences between the

optimal solution calculated by BOSS software and

MATLAB software is the mathematical model.

Table 3: Comparison between the BOSS solution and the

MATLAB solution.

Gasoline

component

Quantity

[tons]

BOSS MATLAB

528.9 0

431.3 431.3

575.4 357.2

561.9 0

2105.4 2105.4

160.5 160.5

1465.4 761.3

446.8 184.3

The obtained results indicate that the BOSS

mathematical model is more complicate than the

author’s mathematical model. The analysis of the

mathematical model developed by the authors has

evidenced other deficiencies according to BOSS

software. In table 4 there is presented a comparison

between the gasoline properties estimated by BOSS

software and the author’s model. The author’s model

is in total according with the BOSS software for

density and benzene concentration.

Unfortunately, the author’s model is in des-

according with the final boiling point of the

gasoline.

Table 4: Comparison between the gasoline properties

estimated by BOSS software and the author’s model.

Gasoline properties Unit BOSS Author’s model

Research octane number - 94.8 95.38

Density g/cm

0.7673 0.7673

Benzene concentration % vol 0.795 0.795

Sulfur concentration ppm 7.4 9.53

Final boiling point

C 202 154.89

5 CONCLUSIONS

The paper presents the author’s research into

industrial in line gasoline blending. The study has

reported differences between the optimal solution

calculated by Blend Optimization and Supervisory

System and the optimal solution obtained by using

the author’s model. The authors have identified the

gut components of the proposed mathematical model

TheOptimalGasolineBlendingintoRomanianRefinery-CaseStudy

531

(Research octane number, density and benzene

concentration) but in same time, they have

incriminated the major deficiencies of model (sulfur

concentration and final boiling point).

The research of the authors will continue with

the next steps:

− experimental determination of the physical

properties of the compounds used to the gasoline

reformulating;

− the blend of the gasoline with know composition

and experimental determination of the physical

properties of the blend;

− the verification of the calculus relations used in

the gasoline blend mathematical model.

REFERENCES

Ghinea, M., Fireteanu, V., Matlab. Numeric calculus,

Editura Teora, Bucuresti, 1997 (Romanian)

Lambert, D., Sanchez, M., Ribero, B., All in the blend,

Hydrocarbon Processing, September, 2006

Manescu, M., Dumitru, V., Ionescu, V., Barbatu, G. I.,

Mathematical programming in the oil industry,

Editura Academiei Republicii Socialiste Romania,

Bucuresti, 1970 (Romanian);

Patrascioiu, C.,

Optimization techniques, Editura

MatrixRom, Bucuresti, 2008 (Romanian);

* * * Advanced Process Control Handbook II,

Hydrocarbon Processing, Gulf Publishing Co., March

1987.

* * * Advanced Process Control Handbook IV,

Hydrocarbon Processing, Gulf Publishing Co., March

1989.

* * * Advanced Process Control Handbook VI,

Hydrocarbon Processing, Gulf Publishing Co.,

September 1991.

* * * Advanced Control and Information Systems’95,

Hydrocarbon Processing, Gulf Publishing Co.,

September 1995.

http://iom.invensys.com/EN/Pages/Solutions_InventoryM

anagement_and_ProductBlending_PB_BOSS.aspx

http://www.unityscientific.com/industries/chemical

ICINCO2012-9thInternationalConferenceonInformaticsinControl,AutomationandRobotics

532