Detailed Modelling and Optimisation of an Ethylene Plant

Francisco José Orantos Borralho

francisco.o.borralho@ist.utl.pt

Instituto Superior Técnico, Lisbon, Portugal

Supervisors: Prof. Dr. Carla Pinheiro, Dr. Maarten Nauta

October 2013

Abstract

This work comprises the detailed modelling and optimisation of an ethylene plant processing fresh propane and
recycle streams of ethane and propane, using gPROMS ModelBuilder ®.
Propane is converted in a steam cracker, operating at low pressure, but high temperature. Then, the effluent is
quickly chilled in one or more transfer line exchangers and in a quenching tower. After that, the cracked gas goes into a
compression train, with intercoolers, to remove water, and is dried in molecular sieves. Finally, the water-free cracked gas
goes into the distillation column train, in order to separate and recover valuable products, namely ethylene and propylene.
This process is energy demanding: temperatures go as high as 900 ºC and as low as -121 oC and the key of it
is the choice and design of the right utilities. Due to the cryogenic temperatures, at least one refrigerant has to be used in
order to chill the cracked gas and condense some of the distillate streams. In this case, two refrigeration cycles were
designed: an ethylene refrigeration cycle and a propylene refrigeration cycle. Ethylene refrigerant is condensed using
propylene refrigerant, whereas propylene refrigerant is condensed using cooling water.
Finally, some key aspects of the designed flowsheet were subjected to an optimisation, and the total annualized
cost was reduced from 87.67 M$/yr to 50.55 M$/yr, corresponding to a reduction of 42.34% on the total annualized cost.

Keywords:
Ethylene plant, steam cracking, olefins, alkenes, gPROMS, refrigeration

1. Introduction using products derived from the pyrolysis of gas oil. This
type of facility only made significant headway in Western
Natural gas and the petroleum fractions obtained Europe, beginning with the United Kingdom, and in Japan,
after the primary fractionation of crude oil by distillation after the end of the Second World War. In 1942, British
consist chiefly of saturated, paraffinic and naphthenic Celanese built the first European steam cracking unit at
hydrocarbons, whose chemical reactivity is mediocre, Spondon operating on gas oil, with a production capacity
precluding the development of diversified families of of 6 kton/year of ethylene. In 1946, Shell Chemical built the
chemical compounds of varying complexity. This can only first petrochemical complex at Stanlow, using refinery
be achieved by using unsaturated aliphatic or aromatic gases as the pyrolysis feedstock. Between the 40’s and the
hydrocarbons which, due to their many reactive 50’s, the minimum capacity of ethylene production plants
potentialities, offer outstanding flexibility for organic grew progressively from 10 to 50 kton/year. Giant
synthesis. In this respect, acetylene, which was for many installations subsequently appeared, routinely producing
years the most widely used basic hydrocarbon in aliphatic 300 kton/year of ethylene from petrochemical naphtha [1].
industry, has gradually been superseded by ethylene, Steam cracking primarily produces ethylene, but
propylene and butadiene, owing to its high production cost. also propylene and, as secondary products, depending on
Despite the fourfold increase in the price of crude oil which the feedstock employed, a C4 cut rich in butadiene and a
occurred in 1973 and its subsequent steady increase, C5+ cut with a high content of aromatics, particularly
ethylene still retained its economic advantage over benzene [1].
acetylene from natural gas or from coal [1]. At the industrial
level, this technique was first developed in the United
States. As early as 1920, Union Carbide and Carbon Co.
built a pilot plant operating on ethane and propane, and
this company went on to create the first chemical complex

1
2. Background 3. Implementation

Over 97% of the annual volume of ethylene

3.1 Software
produced is based on the thermal cracking of petroleum
hydrocarbons with steam, as known as steam cracking or The software used for the simulation, modelling
pyrolysis. First, the hydrocarbon feedstock is heated, and optimisation of all the flowsheets was gPROMS, a
mixed with steam and further heated to incipient cracking process simulator software that is used to create or
temperature (500-650 oC). The mixture is then fed to a fired simulate models, optimize flowsheets, solving a system of
tubular reactor, where under controlled residence time, equations meanwhile.
temperature profile and hydrocarbon partial pressure, is A model created in gPROMS includes a set of
o
heated until 750-900 C. These reactions are highly equations and both variables and parameters. The value
endothermic, so a high energy input is required. of the parameters is defined on the SET section of the
In the downstream section, the cracked gas is model. On the other hand, variables can be either
treated and valuable products are recovered. Main calculated in an equation or its value can be defined in the
processing of the cracked gas includes a chilling section, ASSIGN section. Each variable is related to a variable type
comprising transfer line exchangers and a quenching and the bounds and default value of the variable will be the
tower to stop side reactions, the removal of heavy same as for the variable type.
compounds, compression, the removal of acid gases, Also in the model entity, there’s the TOPOLOGY
drying, cryogenic and conventional fractionation and the section, where the connections between objects are
selective hydrogenation of minor components, like defined. These connections can be either written by code
acetylene, methylacetylene or propadiene. or by dragging and dropping objects and connecting them
With a gaseous feedstock, the downstream in the topology window. Each connection has a certain
processing starts with transfer line exchangers and a water connection type.
quench, followed by a multistage compression, comprising Another fundamental part of a gPROMS project
4 to 6 stages. Before the last compression stage, the is the PROCESS entity, where all the assigned variables
cracked gas is treated in an acid gas removal unit, are stored. To simulate a flowsheet, the process is run,
involving a scrubbing process, where caustic soda (NaOH) while gPROMS calculates the degrees of freedom and
reacts with hydrogen sulphide (H2S) and carbon dioxide solves the problem.
(CO2) [2]. Another feature of gPROMS ModelBuilder is the
After the last compression stage, the cracked optimisation toolkit, where the objective function can be
gas is chilled and dried by the use of molecular sieves. minimized (e.g., total cost) or maximized (e.g., profit),
Methane and hydrogen are then removed in the first through varying a series of control variables (assigned
fractionator, and they can be used as fuel or purified (in a variables), while defining some constraints (not assigned
PSA unit, for instance) and sold. In the deethaniser, a C2 variables).
stream is produced overhead and a C2+ stream is a bottom All the models used on this project were already
product. The overhead stream is then hydrogenated to available in PML-SS library, a gPROMS model library from
remove acetylene and fractionated to recover ethylene on Process Systems Enterprise. The cracker model was
the top and ethane on the bottom, which generally is updated from a previous version.
recycled to the steam cracking reactor. On the other hand,
the bottom product of the deethaniser is fed to the
3.2 Physical Properties Package
depropaniser, where a C3 stream is produced overhead
and a C3+ stream is a bottom product. The overhead All the physical properties were calculated by
stream is hydrogenated to remove methylacetylene and Multiflash and propagated to gPROMS.
propadiene and then fractionated to recover propylene on The key calculation carried out in Multiflash is
the top and propane on the bottom, which generally is the determination of phase equilibrium. This is based on
recycled to the reactor as well. the fundamental relationship that at equilibrium the fugacity
of a component is equal in all phases. For a simple vapour-
liquid system:
fiv = fil (Eq. 1)

2
where fiv is the fugacity of component i in the vapour phase, 3.3 Binary Interaction Parameters Estimation
and fil is the fugacity of component i in the liquid phase.
The models used in Multiflash to represent the fugacities in Binary interaction parameters (kij), or “BIPs” are

terms of temperature, pressure and composition fall into adjustable factors that are used to alter the predictions

two groups: equation of state (EoS) methods and activity from a model until these reproduce as closely as possible

coefficient methods. the experimental data. For the components of the system,

When using an equation of state method, all Multiflash lacked some binary interaction parameters,

thermodynamic properties for any fluid phase are derived setting its default value as 0, considering no interaction at

from the equation of state. When using an activity all.

coefficient method, the vapour phase properties are The best way to get realistic BIPs is to get

derived from an equation of state, whereas the liquid vapour-liquid equilibrium data for a binary or ternary

properties are determined from a combination of models system and compare it to Multiflash’s predictions, so

which include a representation of the excess properties. several datasets of vapour-liquid equilibria were collected.

Equations of state can be used over wide ranges Then, the vapour-liquid equilibrium data for each pair was

of temperature and pressure, including the subcritical and validated against Multiflash’s predictions. For isothermal P-

supercritical regions. They are frequently used for ideal or x data, the pressure for the bubble point of the mixture was

slightly non-ideal systems such as those related to the oil calculated, whereas for isothermal P-y data, the pressure

and gas industry where modelling of hydrocarbon systems, for the dew point was calculated. The residue between

perhaps containing light gases such as H2S, CO2 and N2, both experimental and predicted pressure profiles was

is the norm. minimized, by varying the BIP for the pair of components.

The simple cubic equations of state, Peng- Figure 1 shows the P-x experimental equilibrium data for a

Robinson (PR) and Redlich Kwong Soave (RKS), are 218.15 K isotherm for the binary pair ethylene-ethane,

widely used in engineering calculations. They require compared against Multiflash’s predictions with the default

limited pure component data and are robust and efficient. BIP and the optimized BIP.

For this project, the RKS Equation of State was

950
chosen, since it is better for fugacities estimation. The total
850
pressure and each variable needed for its calculation, with
the standard (Van der Waals 1-fluid) mixing rules, are 750
P (kPa)

given by Equations 2 to 5. Experimental

650
BIP = 0.013
NRT a 550
P= + (Eq. 2)
V−b V(V−b) BIP = 0
450
N= ∑components
i ni (Eq. 3) 0 0,2 0,4 0,6 0,8 1
xethylene
a = ∑ij √ai aj (1 − k ij ) ni nj (Eq. 4) Figure 1: 218.15K isotherm P-x for the ethylene-ethane binary pair.

components The binary interaction parameters can have a

b = ∑i bi ni (Eq. 5)
constant, linear or quadratic dependence on the
where ni is the number of moles of the component i and k ij temperature, according to Equation 9.
is the binary interaction parameter between component i
and component j. k ij = k 0 + k T T + k T2 T 2 (Eq. 9)
For each component, Multiflash calculates ai
According to the datasets available, when a
and bi , from its critical temperature (Tci ), its critical pressure
serious temperature dependence was found, the BIP
(Pci ) and its acentric factor (ωi ):
2
variation with the temperature was fitted to a linear or
Tci
ai = aci [1 + k i (1 − √ )] (Eq. 6) second degree polynomial equation and the different k
T
constants were introduced on the Multiflash GUI.
k i = 0.48 + 1.574 ωi − 0.176 ω2i (Eq. 7) Otherwise, only the constant k (k 0 ) was introduced.
RTci
bi = 0.08644 (Eq. 8)
Pci

3
3.4 Process flowsheet comprises a water quenching tower that cools down the
cracked gas to a temperature near ambient (about 40 oC).
First, an ethylene capacity of 850 kton/year was After the chilling section, the cracked gas is
chosen from an industrial capacity (Exxon Mobil), keeping compressed in a five stage compressor with intercooling in
in mind that usually ethylene capacities vary between 500 between. The inlet pressure of the first stage is 1 bar, the
kton/yr and 1500 kton/yr. Based on the capacity value and outlet pressure of the last one is 32 bar and a constant
assuming that the plant operates 24 hours a day, 330 days pressure ratio of 2 was chosen (1 – 2 – 4 – 8 – 16 – 32
a year, a flowrate of propane was chosen to feed the bar). There is a physical constraint in this section, since
reactor. On the other hand, only the radiation zone is that temperature can’t go up past 100 oC when pressurizing
o
considered in the reactor, so the temperature (650 C) and the system, since olefins can polymerize [2]. Between the
pressure (3 bar) of the feed are typical outlet conditions of compression stages, there are intercoolers that condense
the convection zone of the reactor [2],[3]. part of the water content and it is removed in interstage
For the reactor, the geometry and its properties knock-out drums. In the intercoolers, the cracked gas is
were chosen by assigning typical values presented on cooled with cooling water that enters at 25 oC and leaves
[2],[3], and [4] The coils have a length of 70m, an internal the cooler at 40oC, in order to be possible to cool it back
diameter of 0.108 m and a wall thickness of 0.008 m. again in a cooling tower.
The kinetics were obtained from [4], [5], and [6], Then, the cracked gas is dried. Since the outlet
which present the reactions that occur for an ethane feed, of the last compression stage is at about 90
a propane feed or a mixed feed and the activation energies o
C, which is the maximum allowable temperature to dry the
and Arrhenius pre-exponential factor for each reaction and gas, the cracked gas is first cooled to an intermediate
for each feedstock. For a mixed feed, the following temperature and then is dried. After it is dried, the water-
reactions were considered: free cracked gas is cooled again to 30oC with water and
then to 15 oC with propylene refrigerant [7].
C3 H8 → C2 H4 + CH4 (Eq. 10) The cooled cracked gas is then sequentially

C3 H8 ↔ C3 H6 + H2 (Eq. 11) cooled in a cold box [7]. The cold box is a sequence of four
multi-stream heat exchangers where the cracked gas is
C3 H8 + C2 H4 → C2 H6 + C3 H6 (Eq. 12)
cooled while the distillate of the demethaniser (as known
2 C3 H6 → 3 C2 H4 (Eq. 13) as tail gas) and the vapour outlet of the last knock-out drum

2
(as known as hydrogen rich gas) are heated. There are
2 C3 H6 → C6 H6 + 2 CH4 (Eq. 14)
3 four temperature levels, as four feeds on the demethaniser:

C3 H6 ↔ C2 H2 + CH4 (Eq. 15) -121 oC, -96 oC, -71 oC and -43 oC [7]. Since the cracked
gas can’t transfer all its heat to the cold streams (due to
C3 H6 + C2 H6 → 1 − C4 H8 + CH4 (Eq. 16)
crossover reasons), there is a cooler after the first, second
C2 H6 ↔ C2 H4 + H2 (Eq. 17) and third multi-stream heat exchangers. In the first cooler,
the heat is transferred to propylene refrigerant, while in the
C2 H4 + C2 H2 → C4 H6 (Eq. 18)
second and in the third, the heat is transferred to ethylene
C2 H6 + C2 H4 → C3 H6 + CH4 (Eq. 19) refrigerant [7].

A constant heat profile of 90 kW/m2 [2] was After the cold box, the cracked gas is submitted

assumed across the reactor, since a more detailed heat to the fractionation train in order to separate the several

profile would require industrial data. components. The number of stages and feed stages were

The chilling section comprises two transfer line taken from [8] and the reflux and boil up ratios were varied

exchangers that chill the cracked gas from around 900 C o in order to achieve the desired purities in both top and
o
to 250 C [2]. In the first one, the outlet temperature of the bottom of each distillation column. The pressure of the

cracked gas is 450 o

C and high pressure steam is columns was chosen in order to use the refrigeration levels

produced, while in the second, the outlet temperature is presented by [8] and [9].
o
250 C and medium pressure steam is produced. In a more The demethaniser, which has 65 stages and 4

detailed flowsheet, the two steam streams would receive feed locations (15th, 20th, 25th and 33rd stages), operates at

heat in the convection zone, leaving the convection zone 32 bar and its top product is high purity methane with

with a higher energy content. The chilling section also traces of hydrogen, whereas the bottom is a mixture of C1+.
The demethaniser bottom is then expanded to 26 bar and
enters the deethaniser, which has 60 stages and 1 feed

4
location (27th stage). The top comprises a high content of On the other hand, the bottom stream of the
C2, whereas the bottom comprises a mixture of C2+. The deethaniser is expanded to 8 bar and fed to the
distillate of the deethaniser is then fed to the acetylene depropaniser, which has 60 stages and 1 feed location
hydrogenation reactor, which operates at 26 bar and 340 (25th stage). The top is rich in C3, whereas the bottom
K, where acetylene is converted to ethane and ethylene, comprises a high content of C3+. Finally, the top of the
reacting with added hydrogen. It was assumed a 37% yield depropaniser is compressed to 15 bar and fed to the C3
in ethylene and a 63% yield in ethane [2]. Splitter, which has 230 stages and 1 feed location (120th
C2 H2 + H2 → C2 H4 (Eq. 20) stage). The top product is high purity propylene, whereas
the bottom comprises a high content of propane. Both
C2 H2 + 2 H2 → C2 H6 (Eq. 21)
ethane and propane are recycled to the stream cracking
The effluent of the acetylene hydrogenation reactor.
reactor is then expanded to 19 bar and fed to the C2 Figure 2 shows the process flowsheet of the
Splitter, which has 120 stages and 1 feed location (90th ethylene plant.
stage). The top comprises high purity ethylene, whereas
the bottom comprises a high content of ethane.

C2 Splitter condenser

First, second and third cold box interstage coolers C2 Splitter reboiler
Demethaniser
condenser Deethaniser
Third cracked condenser
gas cooler

Demethaniser reboiler
Depropaniser
condenser

Figure 2: Main process flowsheet.

3.5 Cascade refrigerant cycle

The ethylene refrigeration cycle (golden loop) condenser (12) and the rest is fed to the third cold box
includes two temperature levels, namely -73 oC (4.6 bar) interstage cooler (13). The sub-streams are then mixed
o
and -101 C (1.18 bar). (14) and fed to a compressor stage (16). Both compressed
Superheated ethylene, at 20.2 bar (18) is chilled streams are then mixed (17) and the loop is closed.
o
first with water (1) to 33 C and then with two different levels The propylene refrigeration cycle (green loop)
of propylene refrigerant (2/30 and 3/37) to 8 oC and - 17oC, includes five temperature levels, namely 12.8 oC (8.5 bar),
respectively. After that, ethylene refrigerant is completely 5 oC (6.8 bar), - 20 oC (3.1 bar), -35 oC (1.7 bar) and -46 oC
condensed by another level of propylene refrigerant (4/45), (1.08 bar).
leaving the last heat exchanger at – 30.6 oC. Ethylene is Superheated propylene, at 16.5 bar (54) is
then expanded to 4.6 bar and fed to the only user on this partially condensed with cooling water (19), to a
level (6), the second cold box interstage cooler, leaving it temperature of 39.1 oC. This stream is then expanded to
with an increased vapour fraction. The partially vaporized 8.5 bar and fed and condensed in the only user on this level
stream is fed to a knock-out drum (8), where the vapour is (21), which is the demethaniser reboiler. Propylene
sent to a compressor stage (7) and the liquid is split in two vapours are then separated in the first knock-out drum (22)
streams (9), since there are two users on this temperature and sent to a compressor stage (23). On the other hand,
level. Both expanders (10 and 11) expand both streams to the liquid is split (24) in three sub-streams, which are
1.18 bar. Part of the stream is fed to the demethaniser expanded to 6.8 bar (25, 26 and 27). The sub-stream from

5
 expander 25 is fed to the first user on this temperature level which are expanded to 1.7 bar in expanders 42 and 43.
(28), which is the depropaniser condenser. The sub-stream The sub-stream that comes from expander 42 is fed to the
from expander 26 is fed to the second user on this first user on this temperature level (44), the C2 Splitter
temperature level (29), which is the third cracked gas condenser, while the sub-stream that comes from
cooler. The last sub-stream that comes from expander 27 expander 43 is fed to the second user on this temperature
is partially vaporized in the third user on this temperature level (4/45), the ethylene condenser. Both sub-streams are
level, the second ethylene chiller (2/30). All the three mixed (46) and fed to the last knock-out drum (47), where
partially vaporized sub-streams are then mixed (31) and the vapour is separated and sent to a compressor stage
totally condensed in the last user on this temperature level (48), while the liquid is expanded (49) to 1.08 bar and fed
(32), which is the C2 Splitter reboiler. The liquid stream is to the only user on this level (50), the first cold box
then split again (33) in two sub-streams, which are interstage cooler. Then, the vaporized stream is sent to a
expanded to 3.1 bar in expanders 34 and 35. The first sub- compressor stage (52). Both compressed streams are then
stream, which comes from expander 34, is fed to the first mixed (53) and the loop is closed.
user on this temperature level (36), the deethaniser The cascade refrigerant cycle was obtained by
condenser. The second sub-stream, which comes from connecting both refrigerant cycles through heat
expander 35, is fed to the second user on this temperature exchangers 2/30, 3/37 and 4/45, where ethylene is
level (3/37), the third ethylene chiller. Then, both sub- condensed and propylene is vaporized.
streams are mixed together (38) and fed to a second
knock-out drum (39). The vapour is sent to a compressor
stage (40), while the liquid is split (41) in two sub-streams,

Figure 3: Cascade refrigerant cycle flowsheet.

3.6 Coupled final flowsheet

The coupled final flowsheet resulted from taking out

the single stream heat exchangers from the main process
flowsheet and the cascade refrigerant cycle flowsheet and
connecting both flowsheets through two-stream heat
exchangers. Both flowsheets connect through the first,
second and third cold box interstage coolers (50, 6 and 13
respectively), the demethaniser condenser (12), the
demethaniser reboiler (21), the depropaniser condenser
(28), the third cracked gas cooler (29), the C2 Splitter reboiler
(32), the deethaniser condenser (36) and the C2 Splitter
condenser (44).

6
4. Main Results Table 2: Demethaniser top and bottom results.

Table 1: Reactor inlet and outlet results.

Top Bottom
Mass flowrate (kg/s) 10.5 60.7
Feed Outlet Pressure (bar) 32 32
Mass flowrate (kg/s) 103 103 Temperature (oC) -97 10.5
Pressure (bar) 3 1.5 Mass composition (%)
Temperature (oC) 650 892 Methane 99.1 0
Mass composition (%) Ethane 0 4.1
Methane 0 16.7 Ethylene 0.1 48.7
Ethane 2.6 2.5 Acetylene 0 0.5
Ethylene 0 29.0 Propane 0 14.1
Acetylene 0 0.3 Propylene 0 19.8
Propane 74.9 8.3 But-1-ene 0 0.4
Propylene 0 11.7 1,3-Butadiene 0 3.5
But-1-ene 0 0.3 Benzene 0 8.9
1,3-Butadiene 0 2.1 Hydrogen 0.8 0
Benzene 0 5.2
Hydrogen 0 1.5 In the deethaniser condenser, 9.3 MW of heat is
Water 22.5 22.5 exchanged, whereas in the deethaniser reboiler, 20.6 MW
is exchanged.
Regarding the conversion, an overall conversion
Table 3: Deethaniser top and bottom results.
of 84.8% was achieved. On the other hand, the ratio between
steam and hydrocarbons is 0.29 kg/kg. Figures 4 and 5 show Top Bottom
the gas temperature and the pressure profile, respectively. Mass flowrate (kg/s) 32.3 28.4
Pressure (bar) 26 26
900 Temperature (oC) -17.6 76.2
850
Gas temperature (oC)

Mass composition (%)

800
Ethane 7.7 0.1
750
700 Ethylene 91.4 0
650 Acetylene 0.9 0
600 Propane 0 30.1
0 0,2 0,4 0,6 0,8 1
Relative axial length
Propylene 0 42.4
But-1-ene 0 0.9
Figure 4: Reactor temperature profile.
1,3-Butadiene 0 7.5
Benzene 0 19.0
3

2,5
Pressure (bar)

In the C2 Splitter condenser, 25.5 MW of heat is

2
exchanged, whereas in the C2 Splitter reboiler, 19.9 MW is
1,5
exchanged.
1
0 0,2 0,4 0,6 0,8 1 Table 4: C2 Splitter top and bottom results.
Relative axial length
Top Bottom
Figure 5: Reactor pressure profile.
Mass flowrate (kg/s) 29.7 2.7
To compress the outlet of the water quenching, Pressure (bar) 19 19
40.8 MW of energy is needed. In the demethaniser Temperature (oC) -31.1 -9.7
condenser, 1.5 MW of heat is exchanged, whereas in the Mass composition (%)
demethaniser reboiler, 15.4 MW is exchanged. Ethane 0.1 99.8
Ethylene 99.9 0.2

7
 In the depropaniser condenser, 5.9 MW of heat is be optimized (objective function), the variables that will vary
exchanged, whereas in the depropaniser reboiler, 9.8 MW is in order to reach the optimal solution (control variables) and
exchanged. can still specify equality or inequality constraints for variables
that have to assume a fixed value (equality) or a value within
Table 5: Depropaniser top and bottom results.
a range of values (inequality).
Top Bottom In this project, the objective of the optimisation
Mass flowrate (kg/s) 20.6 7.8 was to minimize the total annualized cost of the coupled
Pressure (bar) 8 8 final flowsheet. That said, the objective function was
Temperature ( C) o
13.5 96 defined as the total annualized cost (CAPEX + OPEX).
Mass composition (%) For the calculation of CAPEX, the costs for the

Ethane 0.1 0 compressors, distillation columns, expanders and heat

Propane 41.5 0 exchangers were considered and respective costs were

Propylene 58.4 0 calculated by gPROMS built-in models. The total CAPEX

But-1-ene 0 3.4 was then annualized, assuming a linear amortization over

ten years. On the other hand, the hourly OPEX was
1,3-Butadiene 0 27.5
calculated by multiplying the energy spent on the
Benzene 0 69.1
compressors by the energy cost per kWh. The annualized
In the C3 Splitter condenser, 53.4 MW of heat is OPEX was then obtained by multiplying the hourly OPEX
exchanged, whereas in the C3 Splitter reboiler, 50.5 MW is by 24 hours/day of operation and 330 days/year of
exchanged. operation.
In the coupled final flowsheet, several
Table 6: C3 Splitter top and bottom results.
operational variables can be used as control variables for

Top Bottom the above mentioned optimisation objective, but due to

Mass flowrate (kg/s) 12.0 8.6 time constraints, only three cases were considered. All the
Pressure (bar) 15 15 cases were run sequentially. Case 2 was run with the new

Temperature ( C) o
35.2 43.4 assigns that resulted from case 1 and the same applies to

Mass composition (%) case 3.

Ethane 0.1 0 Case 1: Mass flowrate of ethylene and propylene

refrigerants and ethylene splitter (9) fractions;
Propane 0 99.9
Case 2: Pressure ratios and intercooling temperatures
Propylene 99.9 0.1
of the compressor train;
Case 3: Number of stages and feed location of the
In the ethylene refrigeration cycle, 30 kg/s of
columns.
ethylene are compressed at the expense of 5.3 MW of
For case 1, it was assumed that all the duties in
energy.
the main process flowsheet remain the same, hence all the
In the propylene refrigeration cycle, 400 kg/s of
duties of the heat exchangers that connect both main
propylene are compressed at the expense of 80.9 MW of
process and the cascade refrigeration cycle remain the
energy.
same as well. On the other hand, all the heat exchangers
with outlet temperatures specified instead of duty kept the
5. Optimisation
same assigned variable.
For case 2, it was considered that regardless the
5.1 Problem Formulation and Assumptions
pressure ratios of the stages on the compressor train and
intercooling temperatures, the outlet pressure of the last
One of the main downsides of the trial and error
compressor stage must be 32 bar. It was considered as
optimisation approach is that it is difficult to manually satisfy
well that the dryer will absorb all the remaining water,
all the process constraints and know if the value found is the
regardless if it is less or more, depending on the water
real optimal one. Additionally, that task can get really time-
removed in the intercooling process (the split fraction of
consuming when the optimisation problem is complex.
water in the dryer remained assigned as 1).
The capacity of solving an optimisation problem is
For case 3, the normalized reflux and boil up
another asset of gPROMS ModelBuilder. Instead of applying
ratios used as assigned variables were fixed and the
the trial and error approach, one can specify the variable to
number of stages and feed location were varied, using as

8
 constraints the purities on the top and bottom of the 6. Conclusions
columns.
From the literature review that was done in
5.2 Optimisation Summary
this project, one can see that this subject, the steam
cracking of hydrocarbons in order to get olefins, is a
Table 7: Optimisation summary.
well-known process studied in the chemical industry,
Before After
Variable since it’s been the preferred project to get olefins, for
optimisation optimisation
Ethylene mass flowrate (kg/s) 30 27.611 more than 30 years. Although there is a lot of
Propylene mass flowrate (kg/s) 400 217.977 information about this process, its flexibility turns it into
Ethylene splitter split fraction 0.6 0.614
something more complex. There are several possible
Compressor stage 1 pressure ratio 2 1.85935
Compressor stage 2 pressure ratio 2 2.04609
feedstocks for this process and the operational
Compressor stage 3 pressure ratio 2 2.06445 variables on a plant aren’t the same for all of them.
Compressor stage 4 pressure ratio 2 2.0101 Apart from that, the investment in different
Compressor stage 5 pressure ratio 2 2.02695
technologies (for instance, for the cracking furnaces)
First intercooler outlet T (oC) 40 30.35
Second intercooler outlet T (oC) 40 30.35 turns the steam cracking into a more and more
Third intercooler outlet T (oC) 40 30.35 innovative process.
Fourth intercooler outlet T (oC) 43.85 41.1
To the extent of my knowledge, there isn’t a
Demethaniser’s number of stages 65 55
public available computational model that describes an
Demethaniser’s feed locations 15,20,25,33 15,20,25,37
Deethaniser’s number of stages 60 51 entire plant as complete as this one, despite the
Deethaniser’s feed location 27 27 assumptions that were made. Several articles and
C2 Splitter’s number of stages 120 117
patents were found, regarding single parts of the plant
C2 Splitter’s feed location 90 85
Depropaniser’s number of stages 60 37
like the cold box, the cracking furnace or the distillation
Depropaniser’s feed location 25 25 column train, but none connects all the parts in just one
C3 Splitter’s number of stages 230 230 flowsheet.
C3 Splitter’s feed location 120 120
The most detailed model found [10],
Compression train energy (MW) 40.8 39.7
Ethylene cycle compression energy (MW) 5.3 4.3 although making some assumptions, was the one that
Propylene cycle compression energy served to validate all the sub-models used in this work,
80.9 26.7
(MW)
apart from the cracking furnace.
Total annualized CAPEX (M$/yr) 6.80 5.31
Total annualized OPEX (M$/yr)
gPROMS can clearly be used to illustrate an
80.87 45.24
Total annualized cost (M$/yr) 87.67 50.55 industrial ethylene plant, since all the models from
PML-SS library were validated by reproducing an
The main source of optimisation was the earlier project [10] and the cracking model was
reduction of both propylene and ethylene refrigerants, validated as well with industrial data [4].
since they were over-specified, corresponding to a Some assumptions were made on the plant, like
reduction on the total annualized CAPEX+OPEX of the omission of the convection zone of the reactor, the
41.39%. On the other hand, after the change on the simplification of the molecular sieve, the simplification of
compression train pressure ratios and intercoolers’ outlet the acetylene hydrogenation reactor or even the omission
temperatures, the total annualized CAPEX+OPEX was of the acid gas removal unit. Regarding the omission of the
reduced by 42.19%. Finally, case 3 led to a final reduction convection zone of the reactor, the only inconvenient is that
of 42.34% of the base case total cost. the heat transfer integration there isn’t modelled. For
example, both steam and hydrocarbons are heated with
100 50%
42,34% upcoming flue gas, in different parts of the convection
Cost reduction from base case (%)
Total Annualized CAPEX+OPEX

87,67 41,39% 42,19%

80 40% zone. Apart from that, the reaction rates in the convection
60 30% zone aren’t null, as considered, but negligible when
(M$/yr)

51,38 50,69 50,55

40 20% compared with the radiation zone.
20 10% The molecular sieve was modelled as a simple

0 0%
0% component splitter. The simplification wasn’t that drastic,
Base Case 1 Case 2 Case 3 since that in several articles it’s stated that water is
Case
Case
completely removed from the cracked gas (water mass
fraction on the cracked gas around ppm) and all the other
Figure 6: Total cost and deviance of optimisation cases from the base case. components aren’t absorbed by the molecular sieve.

9
 About the acetylene hydrogenation reactor, both I’m also thankful to my PSE supervisor, Maarten
yields in ethane and ethylene were assumed, from [2]. Nauta, for being such a selfless and down to earth guy and
Although no kinetics were used and the reactor was for helping me every time I needed help.
modelled as a conversion reactor, it’s considered as well a Also, huge thanks to my family, especially my
good approximation, since the flowrate of acetylene is really parents and my brother for everything they’ve taught me in
small in comparison with ethane and ethylene’s flowrates. twenty three years of life and for investing in my education.
Lastly, the acid gas removal unit was omitted, Finally, I would like to thank to Mariana, for her
which was considered reasonable as well, since the mass patience, her belief in me and for keeping up with me in the
balance on the downstream equipment isn’t affected. last seven months we’ve been apart.
Regarding the optimisation cases, case 1 was the
one that led to the biggest decrease on the total annualized Main References
cost, from 87.67 M$/yr to 51.38 M$/yr, since the flowrates of
both refrigerants were over-specified, specially propylene’s [1] A. Chauvel and G. Lefebvre, Petrochemical Processes,

flowrate, that decreased from 400 kg/s to 218 kg/s, leading Paris: Éditions Technip, 1989.

to a decrease in the energy spent to compress it from 80.9 to [2] F. Ullmann, Ullmann's Encyclopedia Of Industrial

26.7 MW. So, it makes sense that this optimisation case was Chemistry, Weinheim: VCH, 1985.

the most important one, since the majority of the total [3] K. M. Sundaram and G. F. Froment, “Kinetics of Coke

annualized cost is the cost related to energy spent on the Deposition in the Thermal Cracking of Propane,” Chemical

compressors. Engineering Science, vol. 34, pp. 635-644, 1979.

In relation to the refrigeration cycles, although [4] M. Berreni and M. Wang, “Modelling and dynamic

there are numerous possibilities to design it, a similar design optimization of thermal cracking of propane for ethylene

to the ones stated on the articles consulted [8],[9] was used. manufacturing,” Computers and Chemical Engineering,

In order to find out if the results were realistic, the vol. 35, p. 2876– 2885, 2011.

ratio between refrigerants was compared to a ratio of an [5] K. M. Sundaram and G. F. Froment, “Modelling of

industrial plant [8]. A ratio of 7.9 kg propylene/kg ethylene, Thermal Cracking Kinetics - I,” Chemical Enginerring

was obtained, in comparison with the industrial one of 6.9 kg Science, vol. 32, pp. 601-608, 1977.

propylene/kg ethylene, a value 12% inferior. This deviance is [6] K. M. Sundaram and G. F. Froment, “Modelling of

not a surprise, given the possible differences in the heat Thermal Cracking Kinetics - II,” Chemical Engineering

requirements between cases. There are very possible Science, vol. 32, pp. 609-617, 1977.

causes for the deviance, for example, it might be a difference [7] A. A. Al-Rabiah, K. D. Timmerhaus and R. D. Noble,

in purities, perhaps on the C2 Splitter. The reflux and boil up “Membrane Technology for Hydrogen Separaton in

ratios were defined in the C2 Splitter to get 99.9% purity Ethylene Plants”.

ethylene (polymer grade) and maybe in the industrial plant [8] F. M. Fábrega, J. S. Rossi and J. V. d'Angelo, “Exergetic

the purity of ethylene is lower. analysis of the refrigeration system in ethylene and
propylene production process,” Energy, vol. 35, pp. 1224-

Acknowledgments 1231, 2010.

[9] M. Mafi, S. M. Naeynian and M. Amidpour, “Exergy
I would like to express my gratitude to Professor analysis of multistage cascade low temperature
Carla Pinheiro, Professor Henrique Matos, from Instituto refrigeration systems used in olefin plants,” International
Superior Técnico, and especially Professor Costas Journal of Refrigeration, vol. 32, pp. 279-294, 2009.
Pantelides, from Process Systems Enterprise, for giving me [10] K. Y. Cheung, Side-Wide and Supply Chain
the chance to work in Process Systems Enterprise for seven Optimisation for Continuous Chemical Processes, I. C.
months and learn a lot about gPROMS ModelBuilder. London, Ed., London: Centre for Process Systems
Thanks to my housemates Pedro, Vasco and Engineering, 2008.
Tiago for easing my homesickness and for playing an .
important role during the time we spent in London.
A word of appreciation goes also to all my friends
from Instituto Superior Técnico, especially my best friends
Bernardo and Inês for all they’ve done for me in the past five
years.

10