Advanced Control of A Reverse Osmosis Desalination Unit. Assef. 1996. Jprocont
Advanced Control of A Reverse Osmosis Desalination Unit. Assef. 1996. Jprocont
Advanced Control of A Reverse Osmosis Desalination Unit. Assef. 1996. Jprocont
283-289, 1997
1997 Elsevier Science Ltd. All rights reserved
Printed in Great Britain.
0959-1524/97 $17.00 + 0.00
eLsevieR
PII: S0959-1524(97)00004-8
(1)
where Y, M, L, and D are in deviation form and K represents a matrix of steady-state gains. Note that Equa-
284
M~a
Disturbances,
L ~
~r~at~d
vmabl~M
Figure
u,,,,~v.a
r~mi~ l'-----1 Dt~r~ba.~,~D
_]1
+
'
~
~
~
-1 rW~i~ [ r ~/~.-~.,/
o.~.~v
Y2
b2
K2]ml
(2A)
(2B)
with
bl = Yts - Knm~s
K12m2s
K~3m3s
KLlls
(3A)
bz = Y2s - K2~ml~
K22m2s
g23m3s
gL2/s
(3B)
where y, m, l, and D are in absolute form and the subscript s denotes steady-state. With this background, an
optimization problem may be formulated. If linear programming (LP) is selected, the optimization index may
be written as:
subject to:
Kl~m~ + K n m 2 + K~3ms - y~ = - b~ - an - K L J
(5A)
(5B)
Yl + S1u - VIU = Y~
(5C)
S1L + VlL = yL
(5D)
Yl -
Y2 + $2U - V2U = Y~
(5E)
Y2 - SL + ~2L "~ y L
(5F)
--- mlU
2.
(5G)
~
~
Manipulated
< mzU
(5I)
_>
(s J)
<
(SK)
m3 - m3L
(5L)
_ _ [ ~ ~ . 1
- ~
C,p ~.~
~ Outputs,
Y
---~Y-~---~
'
, -Unme.asm'edD i s t u ~
,~u~r~
Figure 2
48
3.
= ~ hiMk+]_l
~,]
where:
i=l
(6)
49
285
(ii)
~ _ ~
~,=,~;
" ~
i(~)
(~
~"
Figure3 Processschematic
t-g~
~..~,~,.
>
286
AcidlnletValve
P~m~t~ *'t"
'l
raowP,~t~ ~ t - ~ .... =--i=-'i--]
Oh)
,
. . . . . .
~
,,
.
P-" ~ I
~'ty
0~)
* = " '~ '* '~
~
~
tr~ia)
FeedpH
.
"
"
~['
*'
,~*~"
. . . . . . .
,
!
~
I'
~
=-~==~- _ ._ _~
. . . . . . . . . . . . . .
,~
*~l
1
~k.
[
~ "
-"
4~
* . . . . . .
~ ~* ~*
discussion
Table
Prg~a~Con~lValw
brought to steady-state and a step change was introduced into the acid inlet valve and the responses of the
four outputs were recorded. Then, the step test was
repeated this time utilizing the pressure control valve.
The resulting step responses (see Figure 4) were fitted to
low order transfer functions with dead-time and the
results are shown in Table 1. Note that the time constants and deadtimes in Table 1 are in seconds.
These transfer functions were used to determine the tuning constants of the two PI controllers and in the
implementation of constrained model predictive control.
The PI control strategy was implemented using the
standard PID blocks provided by FIXDMACS. A sampiing interval of 0.25 seconds was used for the flow loop
and 10 seconds for the conductivity loop. The constrained model predictive controller communicates with
the F1XDMACS data base via an interface program.
-0.155 (0.375s + 1)
(0.22s +1)(2.51s + 1)
Conductivity,/.tS/cm
2.48
(114s + 1)(113s + 1)
Trans-membrane pressure,
psia
--4.74
(1.45s + 1)
Inlet pH
0
0.45
(104s + l)000s +1)
0
--0.077
(21.2s +1)
50
-------------
~;~.~
e
o
aT-~,
-8 -4
! ~
. j 4o0
._1i. . . . .
t35~
"~a
a0o !
4
2s0 ~
F--.
,
-- ,
~20o
0 4 8 12 16 ao 24
~m(,,m)
~ : 3o
t
2
li
"
_=
~=
~ 2
10
.8
l o
a n +6 2o 24
~m(,,i,)
c,~.~_~
s~
~
s~p.~
na',~s
v~
v~,~ v ~
~
Ln * - ~ , ~ )
8.m
7.02
7.0O 0.02 0 . 0 3
2. ~
0,tS/cm)
381.55
381.75
381.5
10.01
256.9
3. Tnm~mm~Pm~(.mia) 233.74 208.99 ~
4. ~ p a
4~0
6.~
~
~L v~,~
, ~ e,tt
, , - .w ~ , ~
49.0
54.4
2. /u:id Inlet Valve
53.6
22.3
4oo
.am
.aa0
~ ~ ~ i
"''ia~ ]
i~ ~
~'
4
'
- -,
3
-8
-2
10
16
--
Figure 5(a) Set point change in permeate flow rate under PI control
t
-~
22
28
.24o
220
Results with C M P C
200
34
Tlrae (rain)
i,o
-2
10
16
22
28
~.....___~
~ 420
,o o
20
.8
287
-I0
10
20
Tlme (mln)
34
30
1"Into (mln)
1.
2.
3.
4.
7.26
38321
215.08
5.78
7.25
389.37
217.42
4.56
53.3
36.8
51.6
51.8
7.25
389.0
---
r , (*x,~) t , ( ~ )
0.02
0,03
10.01
2.'6..9
N
c=
lo t
---
0
-10
--
lO
I
0
0
10
20
30
Tlm.lmin)
com~ll~lVtri~Un
1.
2.
3.
4.
51
3o
20
~.~,~.~v.~
1. ]hre~ O=tl. Vtlve
2. Aeld lalet Valve
P,m Rale 0 1 ~ )
Cam~'livity (l~S/cm)
Tnm~mem. Pre~. (mia)
l a l ~ l~'l
1. r,=,.c,ttv,~
Sttrti~ ~
Value
7.98
405.41
216.95
4.20
Vtlne
7.07
407,58
196,69
6.02
~o.o
39,5
~.4
16.2
o.o
O.O
Limit
7.05
407.0
250.0
6.0
Limit
6,95
405.0
160.0
4.0
70.o
~o.o
80.O
0.0
Figure 6 Set point change in permeate flow rate under CMPC (prediction horizon: 88, control horizon: 8, move suppression,0,0)
288
0'
-,~0
8 "---.r--,.r-~
" ~,
-35O'35O
9~ = 81
....
4.
- 240
-22o
', ~
-10
20
__
4 ~ ....
200
10
-10
20
30
40
190
50
Time (mln)
--1
..+
236
10
Time (lain)
5O
420
.374
3 I
30
5O
+.+
++
to
10
10
0 I
-10
10
20
30
: 0
-10
10
20
30
Time (mln)
40
50
Time (rain)
CemmlledVmSab~
Crm~r~ v ~ - -
~
V~l~
Val~
I. FlowIrate (gph)
2. Coadaclivity(FS/cm)
3. 'IYmss-asan.Press. (psia)
4. pH
7.97
386.44
220.08
4.3.1
7.30
388.18
203.87
5.98
47.5
44.5
53.1
24.9
r~
Wdd~
o':f
3
-10
,
0
14,9
1. l ~ w R ~ (m~)
6.95
386.0
160.0
4.0
2. Ceod=cfivity(j~S/cm)
3. Tram-"w,,, Press. (paia)
4. IuletpH
I. ~ O ~ V ~
47.5
.54.0
0.0
0.0
70.0
30.0
0.0
39.0
10.7
";
~
10
20
Time (rain)
7.08
7.05
388.0
250.0
6.0
80.0
,,~
_
7.95
Limit
14.9
1.75
1.00
12.0
'~= 67
wci~j~
Cost
Cost
,r 8
~,,,~+
va~
Low
vmi~t-,
L Press.CmLValve
2. Acid lalet V&lve
Stmi~
value
H~
,
30
~,~
l++
1.75
1.00
3.00
Low
7.05
6.95
4 0 5 . 0 403.0
250.0 160.0
6.0
4.0
Z~ntpu~:d V m s b ~
0.0
0,0
70.0
30.0
80.0
0.0
F i g u r e 9 Set p o i n t c h a n g e in t h e p e r m e a t e flow r a t e u n d e r C M P C
c o n t r o l - effect o f o u t p u t w e i g h t i n g ( p r e d i c t i o n h o r i z o n : 88, c o n t r o l
h o r i z o n : 8, m o v e s u p p r e s s i o n o f 6, 1.5)
,~" 7 - - - - - - , ~ - - - - - - - r
244
~ 5
200
4
3
40
404.15 405.6'7
218.69 196.19
4.17
6.25
nigh
~
-'~-
-385
~
'
~-~--'~["
-345 -
-265
-10
*
-5
~
0
:
5
10
15
.225
185
20
Time (rain)
60 --.
60
4 0 2050 ~
~ 5 O 2 0 _
12o
10
-10
t
10
[
20
40 N~
8050
5o
4o
30
40
2. ~
(p,.q/cm)
3. Tnms-mzgu.Press. (psis)
4. lal~pH
Mzai/~L_,,,~__Vmisbks
1. Pre~ Cad. Valve
2. Acid Inlet Valve
409.9
204.69
4.98
51.5
30.0
~
Ylllue
7.3O
Weight/
Cost
14.9
6,01
12.0
52.7
15.8
0.0
0.0
401..54
204.21
1.7';
L00
~
~
7.3O
402.0
250.0
6.0
70.0
80.0
3O ~.
2O 10
0
-10
, S ~
Value
7.7.8
10
Time (rain)
Coanu11~ Vadablcs
60
,to
3O
20
I0
0
1 %~ " - -
Low
l,msit
7.20
400.0
160.0
4.0
30.0
0.0
F i g u r e 8 Set p o i n t c h a n g e in p e r m e a t e c o n d u c t i v i t y u n d e r C M P C
c o n t r o l ( p r e d i c t i o n h o r i z o n : 88, c o n t r o l h o r i z o n : 8, m o v e s u p p r e s s i o n ,
6, 1.5)
-5
5
Tlme (rain)
10
15
Cmuollat Vss'isbl~
1. Flow PJt~ (sph)
2. ~
0~q/c~)
~
Value
6.99
411.78
~
W~
Vah~,- Cost
7.25
3. Trims-',',,*,,,Press. (1~)
4. Inletl:~I
411.04
191..96 195.22
5.21
4.00
Mmmds=d Vmables
L Ptms. Cad. Valve
2. +~:~txuk-tv,~
5G.0
21.5
52.3
44.v
0
20
I-ligh
Limit
14.9
9.00
414.0
0.155
0.0
70.0
80.0
1.75
1.00
12.0
law
Limit
6.00
412.0
2 5 0 . 0 160.0
6.0
4.0
30.0
0.0
F i g u r e 10 T h r o u g h p u t m a x i m i z a t i o n u n d e r C M P C ( p r e d i c t i o n h o r i z o n : 88, c o n t r o l h o r i z o n : 8, m o v e s u p p r e s s i o n o f 6, 1.5)
52
289
from distributed control systems. Engineers with training in advanced control and specially trained operators
would also be required for successful installations.
Test 3 is concerned with a step change in the conductivity set point. The results of this test are shown in Figure 8. T o demonstrate the role o f output weighting, test
75, 119-140.
2. Mindler, A. B. and Epstein, A. C., System identification and
control of reverse osmosis desalination. Desalination, 1986, 59,
4. Garcia, C. E. and Morari, M. Internal Model Control: A unifying review and some new results. Ind. Eng. Chem. Proc. Des.
To demonstrate the throughput maximization capabilities of CMPC, the costs on the manipulated variables were a p p r o p r i a t e l y c h a n g e d . T h e results in Figure
outputs.
A comparison of the results with PI control and
CMPC shows the excellent potential for CMPC in RO
plant operations, The features that CMPC offers are
unique and could lead to much improved plant operations, extend membrane life, and result in less down
time. Small plants can be controlled with 486-PCs while
CMPC implementation in large R O units would benefit
53
Conclusions
A successful experimental application of constrained
model predictive control on a RO unit has been presented. The results o f several tests are presented to bring
out the salient features of CMPC. For comparison purposes, the process was operated under traditional PI
control. A comparison of the results with PI control
and C M P C confirms the excellent potential of C M P C
for ROdesalination plants.
References
1. Alatiqi, I. M., Ghabris, A, H. and Ebrahim, S., Measurement
and control in reverse osmosis desalination. Desalination, 1989,
343-379.