13 November 2021
POLITECNICO DI TORINO Repository ISTITUZIONALE
On the optimisation and control of Pressure Swing Distillation unit / Fissore D.; Barresi A. A.. - STAMPA. - (2005).
((Intervento presentato al convegno Computer Aided Process Engineering Forum - CAPE 2005 tenutosi a Cluj-Napoca (Romania) nel 25-26 April 2005.
Original
On the optimisation and control of Pressure Swing Distillation unit
Publisher:
Published DOI:
Terms of use:
openAccess
Publisher copyright
(Article begins on next page)
This article is made available under terms and conditions as specified in the corresponding bibliographic description in the repository
Availability:
This version is available at: 11583/1863847 since: 2017-09-15T12:25:55Z
CAPE-WP
On the optimisation and control of Pressure Swing Distillation Unit
Davide Fissore, Antonello A. Barresi
Dipartimento di Scienza dei Materiali ed Ingegneria Chimica, Politecnico di Torino,
corso Duca degli Abruzzi 24, 10129 Torino (Italy)
CAPE Forum 2005 ROMANIA February 25-26 Cluj-Napoca
Introduction
Recycling of process solvents is receiving a good deal of attention lately as
engineers and chemists strive to reduce long-term costs and minimize wastes.
Distillation can be used for handling these recycle streams because it is a
familiar and robust unit operation and because the required unit equipment may already exist on-site.
The presence of azeotropes, however, may severely limit the use of standard distillation. This work deals with the modelling of a Pressure Swing
Distillation (PSD) Unit for the recovery of ethyl-acetate from a mixture
containing also ethanol, water and other organic compounds in small amounts, being these components responsible for the formation of pressure-sensitive
azeotropes and of liquid-liquid splits; moreover in a PSD unit the sequence of the columns can be readily thermally integrated, thus allowing further savings.
Pressure Swing Distillation concept
C-201
E-107
S-201
E-106
E-201 C-202
C-203
E-203
E-204 E-202
The control problem
Constraints: Ethyl-acetate recovery > 90 %;
Purity of the ethyl-acetate recovered > 99.5 % Water concentration in the product < 0.1 %;
Energy integration between the low pressure and the high pressure column.
The Disturbances: Feed flow rate and composition are variable with time;
Pressure in the decanter and in the low pressure column may be variable.
The objective: Minimise the cost of the operation, i.e. the duty in the high pressure column
Model Predictive Control is the usual framework for optimising plant performances, but the
number of the optimised variables may render the on-line optimisation particularly difficult, thus requiring the use of simplified model (even black-box model).
Because of the complexity of the system, a detailed model should be used; this choice allows also to use the MPC algorithm on plants of different size or different operating conditions, without the need of re-evaluating the parameters of the (simplified or black-box) model.
Process optimisation
The main drawback of working with a detailed model in a MPC algorithm is the complexity of the on-line optimisation; as a consequence steady- state optimisation was used to point out the influence on the objective function of the variables that can be manipulated, showing that the
optimisation of one variable properly chosen may be enough, being poor the effect of the others on the plant optimisation.
Sensitivity analysis
0.88 0.9 0.92 0.94 0.96 0.98 1
3 4 5 6 7 8
RRhigh pressure column
0 50 100 150 200 250
Q, kW
Product purity Product Recovery Heat duty, kW
0.88 0.9 0.92 0.94 0.96 0.98 1
5 6 7 8 9 10
BRhigh pressure solumn
0 50 100 150 200 250 300
Q, kW
Product purity Product Recovery Heat duty, kW
0 0.2 0.4 0.6 0.8 1
0 0.2 0.4 0.6 0.8 1
D:Flow pressure column
0 50 100 150 200 250 300
Q, kW
Product purity Product Recovery Heat duty, kW
0.88 0.9 0.92 0.94 0.96 0.98 1
3 4 5 6 7
chigh pressure column, bar
150 170 190 210 230 250
Q, kW
Product purity Product Recovery Heat duty, kW
In order to point out the role of the various manipulated variables on the cost function we compare the results obtained when all the variables are optimisated and when just one is optimisated in presence of various disturbances.
130 150 170 190 210
-20% -10% 0% 10% 20%
Fin, km ol/ h
J, kW
All variables RR_CH P BR_CH P D:F_CLP
P_CH P P_CLP
The optimisation of the reflux ratio of the high pressure column is enough to fulfill the constraints and to minimise the cost function
Process optimisation
Influence of the parameters of the MPC algorithm
0.80 0.85 0.90 0.95 1.00
0 0.1 0.2 0.3 0.4 0.5
t, h
Molar fraction of the ethyl-acetate produced
Hp = 4 Hp = 3
no control action Hp = 5
0.82 0.84 0.86 0.88 0.90
0 0.1 0.2 0.3 0.4 0.5
t, h
Recovery of the ethyl-acetate in the product
Hp = 4 Hp = 3
no control action Hp = 5
Case study: step disturbance (-20%) in the feed flow rate
Control horizon = 2
120 121 122 123 124 125
0 0.1 0.2 0.3 0.4 0.5
t, h
Heat Duty in the high pressure column kW
Hp = 4 Hp = 3
no control action Hp = 5
Influence of the parameters of the MPC algorithm
Case study: step disturbance (-20%) in the feed flow rate
Prediction horizon = 2
0.80 0.85 0.90 0.95 1.00
0 0.1 0.2 0.3 0.4 0.5
t, h
Molar fraction of the ethyl-acetate produced
Hc = 2 Hc = 3
no control action
Hc = 4 0.82
0.84 0.86 0.88 0.90
0 0.1 0.2 0.3 0.4 0.5
t, h
Recovery of the ethyl-acetate in the product
Hc = 2 Hp = 3
no control action Hc = 4
120 121 122 123 124 125
0 0.1 0.2 0.3 0.4 0.5
t, h
Heat Duty in the high pressure column kW
Hc = 2 Hc = 3
no control action Hc = 4
Model Predictive Control Algorithm
minu set i
h
y i y i
p
2
1
min ( ( )) ( ( ))
( ).... ( )
u k u k h y y
j k k h
u i k k h
c
p p
e j u i
1
2 1
2 1
Constraints handling: penalty functions
k hp
k t
nc
i i SOFT
SOFT i i
SOFT
i y
y t F y
F
1
2
1 ,
,
,
Process
Model
Model Optimisation
+
+
+
+ - - +
- u
u u
Constraints
+
Process
Model
Model Optimisation
+
+
+
+ - - +
- u
u u
Constraints
+
Model error handling:
Model Predictive Control Algorithm
minu set i
h
y i y i
p
2
1
min ( ( )) ( ( ))
( ).... ( )
u k u k h y y
j k k h
u i k k h
c
p p
e j u i
1
2 1
2 1
Constraints handling: penalty functions
k hp
k t
nc
i i SOFT
SOFT i i
SOFT
i y
y t F y
F
1
2
1 ,
,
,
Process
Model
Model Optimisation
+
+
+
+ - - +
- u
u u
Constraints
+
Process
Model
Model Optimisation
+
+
+
+ - - +
- u
u u
Constraints
+
