You are on page 1of 6

Off-line Energy Optimization Model for

Crude Distillation Unit


Predrag Domijan*, Damir Kalpi**
*

INA Rijeka Lube Oil Refinery, Milutina Baraa 26, HR-51000 Rijeka, Croatia
predrag.domijan@ina.hr
** University of Zagreb, Faculty of Electrical Engineering and Computing, Department of Applied Mathematics
Unska 3, HR-10000 Zagreb, Croatia
damir.kalpic@fer.hr

AbstractThis paper presents a model that uses the crude


true boiling point curve and other routinely made
laboratory measurements on crude distillation unit (CDU)
for product properties prediction. The input to model is
different type of crude. The objective of optimization is cost
of energy consumption per tons of crude. The detailed
model is based on a real refinery, a first-principle,
nonlinear. The CDU model is a non-stage-by-stage, steadystate model and corrected by real process values. The model
calculates the yields and properties of the products based on
the feed information and product specifications. The
proposed model has the next benefits in compares with
commercials simulation models: adapted to real plant
conditions, open source, flexible and fast execution. The
model is very useful in the next applications: compare actual
plant performance to the design case and to the optimal
case, identify operating changes needed to maximize profits,
ensure the highest return on investment for additional
capital expenditures on the unit, identify the fouling level
and for planning optimal shutdowns and maintenances the
unit. The optimal solution obtained shows that the energy
consumption could be decreased 3.2% over the normal
operating condition.

I.

INTRODUCTION

The oil industry (or petroleum industry) is normally


divided into two parts: upstream (concerned with finding
oil deposits (explorations) and getting the crude oil out of
the ground (production)) and downstream (concerned with
turning crude oil into usable products (refining and
petrochemicals) and delivering them to customers
(distributions)). Petroleum industry is one of the most
prolific and dynamic industries of modern civilization
where a highly competitive market and stringent
environmental laws, strict quality control of refinery
products is a must. Crude distillation unit (CDU) is one
through which entire crude entering a refinery must be
processed.
Crude oil as well as all the CDU products is complex
mixtures of hydrocarbons, so it is not convenient to
characterize them in terms of individual components, see
in [1], [2] and [3].
The property prediction methodology makes use of
crude true boiling point (TBP) curve (available from
refinery laboratory) and steady state data such as operating
flowrates, temperatures and pressures. Based on draw

plate temperatures of different products the algorithm


calculates five equilibrium flash vaporization (EFV)
temperatures. These EFVs are then converted to TBPs
which, when superimposed on the crude TBP curves,
allow estimation of five more product TBP temperatures.
Then TBP curves are used as input to model. Thus without
making the simulation over the entire column, product
properties are determined in an easy and instant way. This
method is mainly based on as is proposed in [2] and [3];
and is not the same done in [1].
II.

CRUDE DISTILLATION UNIT (CDU)

Figures (1)-(6) show process flow diagrams of the


CDU. The refinery processes blends of REB, Brent, Iraq
Kirkuk, Libyan EsSider and Syrian light crude oils.
During the unit performance test (test-run), the
atmospheric bottoms (Residue) yield was 42% of crude.
The column's lowest side cut is heavy gas oil. Light gas
oil and Kerosene are side products. Gasoline C5-180, LPG
and Still gas combine the C01 top product.
The CDU is compound of six sections and is shown in
the same way as is on Honeywell TPS distributed control
system (DCS) screens. On figure (1) is given Desalter
(V04) and Preflash column (C06) with heat exchangers
for heating crude oil before furnace (F01). Figure (2)
show Furnace (F01) and Atmospheric column (C01)
section and Reflux section (V01 and C03) is given on (3).
Debutanizer (C04) and Stabilizer (C05) sections are given
on (4) and (5). Heavy distillate cooling section is shown
on figure (6).
An example of optimizing CDU is given on figures (7)
where the results are flow rates (t/h), heat exchangers
energy duty (MJ) and temperatures (C).

III.

MODEL

Proposed model is an off-line, first principle: based on


mass and energy balance and validated by actual data,
nonlinear and steady-state [4].
Mathematical Programming (MP) in the oil industry,
and in oil refining in particular, is well established. The oil
industry is the largest single user of Mathematical

Programming [5]. This is because the technique is


effective in tackling the problems which the industry
faces.
The characteristics of a typical MP problem are:
many potential solutions;
some measure of the quality of solutions;
interconnectedness between the variable
elements of the system.
Oil refineries face an enormous number of options in
their operations:
which crude to refine;
what processing condition to use;
how to economize with the energy.

Figure 4. Debutanizer column section

Figure 5. Stabilizer column section


Figure 1. Desalter and preflash section

Figure 6. Heavy distillate cooling section


Figure 2. Furnace and Atmospheric column section

Figure 7. An example of optimizing CDU


Figure 3. Reflux section

Nonlinear Programming is a part of the mathematical


programming. In this project nonlinear programming is
used because of inherently existing nonlinearity in
description the CDU equipments [6]. In some paper
Genetic algorithms is used to solve nonlinear
dependencies [7].
Simulation programs are very popular in oil refineries
[8] but because of their weaknesses (see [9] and [10]) the
model developed in the paper gives the better (i.e. closer
to real operation) solution for the energy optimization
application presented in the papers.
The model of CDU is given in figure (8). Preprocessing
input data was calculated in MS Excel file named
Technological Preprocessing file (TPP file). The data in
the file is changed in laboratory when was changing the
crude oil or basic properties of crude oil. In the file are
data: TBP, EFV, density, characterization factor (K) and
volume (and mass) flow of all crude oil fractions.

TPP file
(MS Exc el)

Inp ut/Output file


(MS Exc el)

MODEL
(Lingo v3.1)
Figure 8. Crude Distillation Unit Model

The next MS Excel file: Input/Output file read the


prepared data from TPP file. The file combines data that
characterize equipments in the CDU, crude oil, distillate
and energy prices.
Nonlinear CDU model was writing in LINGO. Total
variables are 210 of which 116 nonlinear. Total constrains
are 266 of which 78 nonlinear. Memory used is 191 kBy
and elapsed time is less than 5 sec (PC: Pentium 2.8 GHz).
Local optimum was achieved in 99 iterations.
The results of the nonlinear optimization is transferred
to Input/Output file (see figure (8)). In the file all products
streams are calculated. Specially, F01 furnace duty is
depictured separately because of the higher energy (fuel
gas and oil) consumptions.

A. Performance Function
Performance (or goal) function of the model is
minimizing cost of energy and chemicals consumptions
per ton in hour of input crude oil:
min(Cost/Crude) = Cost($) / Crude(ton)
Cost($) = Chemicals($/ton) Crude(ton)
+ H2O_Desalter($/ton) H2O(ton)
+ Steam_5bar($/ton) Steam_5bar(ton)
+ Steam_12bar($/ton) Steam_12bar(ton)
+ H2O_Cool($/ton) H2O_Cool(ton)
+ Fule _Oil($/ton) Fuel_Oil(ton)

+ Fuel_Gas($/ton) Fuel_Gas(ton)
+ Elect_Power($/MW) Elect_Power(MW)
The typical value of performance function was 2.5~3.5
($/ton) depending on the CDU capacity and crude oil type.
B. Process Equipments Model
All process equipments in the CDU are modeled based
on first principle (energy and mass balance) and validated
by actual data. Equations are nonlinear and model
represents the steady-state.
Energy balance is based on heat contents (enthalpy) of
petroleum fractions; see [2] and [3]. An USER@ function
is defined in LINGO for calculating enthalpy of petroleum
fractions and H2O (for both: liquid and vapor state).
Example 1:
Enthalpy (h) of liquid petroleum fractions:
hHC(t, , K )[MJ/ton]= (0.00295682247 t2(C)
- 17.1523 2(ton/m3)
- 0.0010328 t2(C) (ton/m3)
- 0.437829 t (C) 2 (ton/m3)
- 0.607645 t (C) (ton/m3)
+ 2.698276 t (C)
+ 3.032970 (ton/m3)
+ 41.86634) (0.054 K + 0.352)
Example 2:
Enthalpy of steam (1-12 bars):
hSteam(t, )[MJ/ton]= 2.034506 t (C)
+ 2460.255
Where is:

t(C)
- temperature
(ton/m3) - density

1) T-pipe type 1
T-pipe type 1 is the first element to be modeled. On
figure (9) is given the example where two similar streams
make one.

F3,t 3,h 3

F1,t 1,h 1

F2,t 2,h2
Figure 9. T-pipe type 1

F (ton/h) h (MJ/ton) -

flow
enthalpy

4) Destillate cooling fan


F3 = F2 + F1
t3 = (F1 t1 + F2 t2) / F3

On figure (12) is given the example of typical distillate


cooling fan (a type of cooler).

2) T-pipe type 2
On figure (10) is given the example where one stream
makes two.

Figure 12. Distillate cooling fan

F3 ,t3 ,h 3

F1,t1,h1

F2,t2,h2

Q1 = F1 (h1 - h2)
Q1 < max Duty
t2 < t1 - 1
5) Distillate-Water cooller

Figure 10. T-pipe type 2

On figure (13) is given the example of typical distillatewater cooler (exchanger).

F3 = F1 - F2
t3 = t1 = t2
h3 = h2 = h1

(Ak2,f%)

3) Exchanger

F1 1,t 11,h11

On figure (11) is given the example of typical


exchanger.

F2,t2 1,h 21

(Ak,f%)

t 22,h22

Figure 11. Exchanger

- construction parameter
- fouling factor
- energy transfer from first stream
- energy transfer to second stream

Q1 = F1 (h11 - h12)
Q2 = F2 (h22 - h21)
Q1 = Q2 = Ak f% LMTD
LMTD = [(t11 - t22) - (t12 - t21)]
/ ln [(t11 - t22) - (t12 - t21)]
t11 > t22 +1
t12 > t21 + 2
Model of the exchanger is based on [3] and [6]. Fouling
is discussed in [11] and [12].

t 12,h1 2

FH2O 22 ,t22 ,h22

t1 2,h12

F1,t 11,h1 1

Ak(MJ)
f%
Q1(MJ)
Q2(MJ)

t2,h2

F1,t1,h1

T21,h21

Figure 13. Distillate-Water cooler

Q1 = F1 (h11 - h12)
QH2O = FH2O 22 (h22 - h21)
Q1 = Ak2 f% QH2O
6) Vessel
On figure (14) is given the example of a vessel.

F1,t 1,h 1

F2,t 2,h2

F3,t3,h3
Figure 14. Vessel

F2 ~ 0
F3 = F1, t3 = t1, h3 = h1

F4 > 0.2 F1
F5 = F6
F1 + F4 = F3 + F2
F4 = F2 - F7
F3 = F1 - F7
Qr = F4 (h2 - h4)
Qb = QE22
QF7 = F7 (h2 - h1)
QF3 = F3 (h3 - h1)
Qb = QF7 + Qr + QF3
t3 < t6
t3 > t5

7) Furnace
On figure (15) is given the furnace.

FSte a m_5 ba r,t21 ,h 21

t22 ,h22

F01
t 12,h1 2

FCrud e ,t 11,h1 1
FHO

FHG

Atmospheric distillation (C01), Debutanizer (C04) and


Stabilizer (C05) columns are very similar in modeling, see
more in [2] and [3].

FSte am _12 bar


C. Initial values

Figure 15. Furnace

Q1 = FCrude (h12 - h11) + FSteam_5bar (h22 - h21)


Q2 = QHG FHG + QHO FHO
Q1 = F01 Q2
Q2 < maxDutyF01
FSteam_12bar = FHO Steam_12bar_%
HG Fuel Gas
HO Fuel Oil

IV.

8) Stabilizer column
On figure (16) is given the example of Stabilizer
column.

F2,t2,h2

F1,t1,h 1

(V03,E20,E31)

F4,t4,h4

C05
F5,t5,h5

(QE2 2)
F6,t6,h6

For nonlinear models in particular, the values with


which the solver begins its search may determine the
quality and speed of the solution.
In the CDU model initial value are put very close to
"guess or expected" value. It is necessary to try a few
values before the solver find the optimal solutions.
Otherwise, usually the solution is far from optimum or
even infeasible.

F3,t3,h3
Figure 16. Stabilizer tower

F7 ,t4,h4

TEST REZULTS

The described CDU model has been tested on real CDU


in INA Refinery Rijeka. Iraq Kirkuk ( = 0.8547 t/m3, K =
11.9) crude oil was used for test-run.
Results of comparing the process data of real operation
(from Honeywell TPS DCS screens) and optimize Model
CDU is given in Table 1.
The results are divided in three parts: The first part is
the division of the crude oil in eight fractions. The
divisions in both columns are nearly the same.
The energy consumption is in the second part. The
results of optimization the CDU Model gives lower
energy demand. The main reason for higher energy
demand in real CDU is fouling in exchangers and furnace.
The last part is process data. It could be notice that
reflux flow in C01 column is lower in the CDU model
than in operation example. The flashed crude temperature
at the entry of the furnace is higher in the CDU model. It
leads to lower energy demand in the furnace.
The performances of optimization are given in the last
two rows. Finally, the 3.17% of energy cost savings is
possible in ideal situation (with no fouling and good
condition of all equipment in the CDU).

TABLE 1:
Crude oil: Iraq Kirkuk
Crude oil fractions
Still Gas
LPG
Gasoline C5- 78C
Gasoline 78 - 180 C
Kerosene
Light Gas oil
Heavy Gas oil
Residue
ENERGY
Steam 5 bar (ton/h)
Steam 12 bar (ton/h)
Cooling water (ton/h)
Desalter water (ton/h)
Fuel gas (ton/h)
Fuel oil (ton/h)
Electrical Power (MJ)
PROCESS DATA
- Furnace F01 (%)
Crude oil in Desalter (C)
Crude oil in F01 (C)
Crude oil in C01 (C)
Gasoline in V01 (C)
Flow reflux in C01 (ton/h)
Flow reflux in C04 (ton/h)
Flow reflux in C05 (ton/h)
Flash zone C01 (at)
Energy cost ($) /
Crude oil flow (ton)
Energy cost savings:

DCS
%
0.07
1.15
5.94
15.70
9.69
18.17
6.99
42.29

CDU Model
%
0.05
1.25
5.32
15.68
9.75
18.29
7.48
42.19

6.34
1.98
603.90
16.91
2.30
6.29
3.47

7.11
2.26
441.53
24.00
2.45
5.66
3.50

shutdown planning for maintenance and crude


management.
The possible saving in energy of 3.17% is valuable
because CDU is one of the biggest energy consumers [13]
in typical refinery and there is the best opportunity for
energy management.
The model can be in addition used for testing different
revamping scenarios and to solve production bottlenecks
in the CDU.
Finally the model presented in the paper it could be
useful at the stage of planning production in case to use
new crude oil. Accounting for more than 85% of the
typical refinery's operating cost structure, the crude is one
key "operating parameter" that must be set early in the
planning process [14]. The benefit of using the CDU
model for that application is much higher than is presented
in the paper.
REFERENCES
[1]

[2]

75.3
125
256
377
156
191.3
34.85
31.04
3.21

78.0
134
253
371
140
137.2
26.7
39.6
2.72

2.84

2.75

---

3.17%

[3]
[4]

[5]

[6]
[7]

[8]

V.

CONCLUSION

[9]
[10]

In this paper we have proposed nonlinear, steady-state,


first principle model of Crude Distillation Unit (CDU).
Although the CDU Model is not rigorous (no column
stage by stage simulation) it can be used by plant
engineers on daily work (setting operation points of all
equipments, calculating energy consumptions and
fouling).
Presenting CDU model gives the answers on the
questions related to equipment condition in the CDU,

[11]
[12]
[13]
[14]

S. Bhat, T. Chatterjee, and D. N. Saraf, On-line Data Processing


and Product Properties Prediction for Crude Distillation Units,
AIChE 2003 Spring National meeting New Orleans, Louisiana
March 30 - April 3, 2003, Paper No.115c.
J. B. Maxwell, Data Book on Hydrocarbons, Fourth printing, D.
Van Nostrand Company, Inc, Princeton, New Jersey, 1957.
W. L. Nelson, Petroleum Refinery Engineering, Fourth Edition,
McGraw-HILL Book Company, Inc, New York, 1958.
S. M. Ranade, Achieve sustained business benefit from rigorous
offline process models, Hydrocarbon Processing, pp. 57-61,
December 1997.
R. Simons Mathematical Programming in the Oil Industry, The
Newsletter of Mathematical Programming in Industry and
Commerce, June 1996.
J. E. Albers Nonlinear model-based control of a heat exchanger,
Hydrocarbon Processing, pp. 101-106, November 1999.
S. V. Inamdar, S. K. Gupta, and D. N. Saraf, Multi-objective
Optimization of an Industrial Crude Distillation Unit Using the
Elitist Non-dominated Sorting Genetic Algorithm, Chemical
Engineering Research and Design, Vol. 82, Issue A5, pp. 611623, May 2004.
J. Forrest, and E. Reyes, Companies find new value with
refinery-wide rigorous simulation solutions, ERTC 8th Annual
Meeting, pp. 9-10, November 2003.
R. Sowell Why a simulation system doesn't match the plant,
Hydrocarbon Processing, pp. 102-107, March 1998.
J. C. M. Hartmann Decision-making and modeling in petroleum
refining, Hydrocarbon Processing, pp. 77-81, November 1997.
H. M. Joshi, Mitigate fouling to improve heat exchanger
reliability, Hydrocarbon Processing, pp. 93-95, January 1999.
K. Bailey, Optimize heat exchanger operations by minimizing
fouling, Hydrocarbon Processing, pp. 113-117, June 1999.
S. M. Brown, The drive for refinery energy efficiency, Refining,
pp. 45-55, PTQ Autumn 1999.
K. G. Waguespack, and J. F. Healey Manage crude oil quality for
refining profitability, Hydrocarbon Processing, pp. 133-140,
September 1998.

You might also like