Grey Prediction Based Control of a Non-Linear

Liquid Level System Using PID Type Fuzzy

ontro er
Erdal Kayacan Okyay Kaynak
Electrical and Electronic Engineering Department Electrical and Electronic Engineering Department
Bogazici University Bogazici University
Bebek, Istanbul 34342 Bebek, Istanbul 34342
Turkey Turkey
erdal. kayacan(@fboun. edu. tr okyay. kaynak(@fboun. edu. tr

Abstract - Product-sum type fuzzy controllers are known to
have similar characteristics with PD type controllers. In the
case of type-O control systems, PID type fuzzy controllers
have been proposed in the literature in order to eliminate the
steady-state error. However, these control methods,
essentially based on conventional PID theory, have no
predictive capabilities. This paper proposes a grey prediction
based fuzzy PID controller that can overcome the stated
shortcoming. A non-linear liquid level system is taken as a
test bed. The grey model developed is examined under several
different conditions and it is shown that the proposed grey
fuzzy PID controller can predict the future output value of
the system. The simulation results indicate that the proposed
controller has the ability to control the non-linear system
accurately with little overshoot with no steady-state error.
INTRODUCTION
I considered. It is obvious that the liquid level in a tank is
In control theory, a system can be defined with a color that
represents the amount of clear information about that
system. For instance, a system can be called as "black box"
if its internal characteristics or mathematical equations that
describe its dynamics are completely unknown. On the
other hand if the description of the system is, completely
known, it can be named as white system. Similarly, a
system that has both known and unknown information is
defined as a grey system. In real life, every system can be
considered as a grey system because there are always some
uncertainties.
In real life, due to noise from both inside and outside of the
system of our concern (and the limitations of our cognitive
abilities!), the information we can reach about that system,
is always uncertain and limited in scope [1]. For instance,
speed measurement data collected through a motor control the predicted values of original data.
system always contains some grey characteristics due to the Consider a single input and single output system:
time-varying parameters of the sensor, not to mention the
measurement difficulties.
11 CONCEPTS OF GREY SYSTEM THEORY AND GM(1,1)
MODEL
A GM(n, m) Mode I
Grey models can predict the future outputs of systems with
high accuracy without knowing the mathematical model of
the actual system. In grey systems theory, GM(n,m) denotes
a grey model, where n is the order of the difference
equation and m is the number of variables. GM(1, 1) type of
. . . ' .
grey
r
am odel idel used ine itatue,
B GM(1, 1) Model
The GM(1,1) model can only be used in positive data
sequences [2]. In this paper, a non-linear liquid level tank is
always positive, so that GM(1,1) model can be used to
forecast the liquid level in this tank.
In grey systems theory, the main task is to extract realistic
governing laws from the available data. This process is
known as the generation of the grey sequence [3]. It is
argued that if the randomness of the data obtained from a
grey system is somehow smoothed, it is easier to derive the
any special characteristics of that system. In order to
smooth the randomness, the primitive data obtained from
the system to form the GM(1,1) is subjected to an operator,
named Accumulating Generation Operation (AGO)[2],
described below. The differential equation (i.e. GM(1,1))
thus evolved is solved to obtain the n-step ahead predicted
value of the system. Finally, using the predicted value, the
inverse accumulating operation (IAGO) is applied to find
Xo() (x ) (1), x( ) (2) ...... ,x)(n) n > 4 (1)
=

Grey system theory was first introduced by Professor Deng

Ju-long from China in the international journal "Systems whreX0).sannngtv euec n stesml
and Coto Letes in 1982 [2]., Th thor is..size of the data. When this sequence iS subjected to the
ditnushdwt it abiit to deal wit th
have partially unknown parameters. With 1 the use of grey rytesta
Accumulated Generation Operation (AGO), the following
1
system mteais(for instance grey equations and grey mteais . sequence X~' iS obtained.
matrixes, etc.) it is possible to generate meaningful
information using little poor data. Grey predictors have ,xl(n),J n>.4 ....
IX(l'=((1)(),x l(2)., ....... (2)
ability to predict the future outputs of a system by using
recently obtained data. where

1-4244-971 3-4/06/$20.OO ©2006 IEEE 292

Authorized licensed use limited to: ULAKBIM UASL - BOGAZICI UNIVERSITESI. Downloaded on February 19, 2009 at 07:59 from IEEE Xplore. Restrictions apply.
 E. Kayacan, 0. Kaynak Grey Prediction Based Control of a Non-Linear Liquid Level System Using PID Type Fuzzy Controller
k It is well known that if the controlled system is type "O", a
x(l) (k) x(°) (i),
x k = 1,2...., n (3) P or PD type controller cannot eliminate the steady-state
error. Although the use of an integral term in the controller
The generated mean sequence Z (1) of X (1) is defined as (such a PI controller) can take care of the steady-state error,
it can deteriorate the transient characteristics by slowing the
z (z(') (1) z (2), z--'Z(n)) (4) response. However, with a PID-type fuzzy controller fast
rise times and small overshoots as well as short settling
where z (1) (k) is the mean value of adjacent data, i.e. times can be achieved with no steady-state error.
I
B PID Type Fuzzy Controller Structure
z(')(k)Px(1)(k) + x()(k1), k=2,3...,n (5) A conventional PID controller uses three input variables;
2 2 the error, the change rate of error and the integral of error.
The least square estimate sequence of the grey difference The PID type fuzzy controller described in this paper can
equation of GM(l1,1) is defined as follows [2]: achieve a similar output response, using only two
x(°) (k) + az(1) (k) = b (6) variables, the error and the change rate of error. Allowing a
PD and PI type fuzzy controller to work in parallel makes
The whitening equation is therefore as follows: the controller more practical and requires less
d (1) (t) computational effort [4].
+ ax(l) (t) = b (7) An equivalent structure is shown in Fig. 1, where f and oc
dt are the weights of PI and PD type controllers, respectively.
In above, [a, b]T is a sequence of parameters that can be
found as follows: r+ eFuzzy u + u r
[a b]T = (BTB)I BTy (8) eController
Kd
where
[x
Y= (°) (2), x (°) (3), .......
x (°)
(n)] T
. .(9) Figure 1
PID type fuzzy controller structure

2-
(3) 1 The output of the controller can be expressed as:

B= . (10) uc= aru+ ,Ifudt (13)

z () (n) 1 As the uc/f ratio becomes larger, the effect of the derivative
control increases with respect to integral control [5]. This
According to (7), the solution of x(1) (t) at time k: controller is called as PID type fuzzy controller (PIDFC).

o b+-
b IV GREY PID TYPE FuzzY CONTROL
xl(k+1)= x() (1)
b
-ak (11
a _J a
L_
A Rule Base and Membership Functions
To obtain the predicted value of the primitive data at time In a conventional fuzzy inference system, an expert, who is
(k+H), the IAGO is used to establish the following grey familiar with the system to be controlled, formulates the
model. rules. The fuzzy PID type control rule base employed in
this work is shown in Table 1. The membership functions
x(4)(k + H)= x(°)(1) -- e-a(k+H1) (I - ea) (12) of normalized error, change rate of error and control signal
a are chosen as triangular membership functions as shown in
Fig. 2.
III COMBINING FUZZY AND PID TYPE CONTROL
Table 1
A general fuzzy PID type rule base
A Analysis of the Fuzzy Controller
Consider a product-sum type fuzzy controller with two NL NM NS ZR PS PM PL
inputs and one crisp output (MISO). Let the inputs to the N N
fuzzy controller be the error e and the rate of change of the Pb ZR PS PM PL PL PL PL
error e, and the output of the fuzzy controller (that is the PM NS ZR PS PM PL PL PL
input to the controlled process) be u. If an analysis of this PS NM NS ZR PS PM PL PL
controller is made, it can be seen that it behaves ZR NL NM NS ZR PS PM PL
approximately like a PD controller. We can therefore NS NL NL NM NS ZR PS PM
consider it as a time-varying parameter PD controller [4]. NM NL NL NL NM NS ZR PS
Such a controller is named as a PD type fuzzy controller Nb NL NL NL NL NM NS ZR
(PDFC) in the literature.

293

Authorized licensed use limited to: ULAKBIM UASL - BOGAZICI UNIVERSITESI. Downloaded on February 19, 2009 at 07:59 from IEEE Xplore. Restrictions apply.
 ICM 2006 * IEEE 3rd International Conference on Mechatronics

1 NL NM NS ZR PS PM PL
ff ~~~~control valve
for flap angle ~p

-1
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 discharge
valve
Figure 2
The membership functions of e, and u
-t qout

B Design of Grey PID Type Fuzzy Controller Figure 4

In most control applications, the control signal is a function A simple liquid-level system
of the error present in the system at a prior time. This
methodology is called as "delay control". In grey systems VI SIMULATION RESULTS
theory, prediction error is used instead of current measured A number of simulation studies have been carried out the
error [6]. Similarly, during the development of a grey PID
type fuzzy controller, the pla escried
' prediction error is considered the ~~~~~~~~~~values in thpaper
used in this pre K=1 andsection.
e are Kd=O. 1.TheFig.numeica
5 shows
error of the system. Fig.3 shows the general structure of a the response of the model to PDFC and PIFC. As can be
grey PID type fuzzy controller. grey PID type fuzzy controller. seen the
system response is very fast but there is a steady-
state error with PDFC. With PIFC, the system does not
r + e
'
Yout have a steady-state error but a big overshoot and a slow
- Fuzzy PID u Non-linear response. Fig. 6 shows the response of the system to PIDFC
de/dt Controller System and conventional PID controller. As can be seen, PIDFC
has a better capability, when compared to a conventional
PID controller in controlling such a non-linear system.
ypredicted Greydt
Predictor
Fig. 7 compares the unit step response of the system with a
PIDFC and a grey controller with different prediction
horizons, i.e. H in (12). As can be seen, when the step size
Figure 3 of the grey controller is large, it will cause over
The structure of a Grey PID type fuzzy controller copnain reutn in a lw sse epne
compensation, resulting in a slow system response.
Conversely, a smaller step size will make the system
V DESCRIPTION OF CONTROLLED OBJECT respond faster but cause larger overshoots [8]. The response
Fig. 4 shows a simple system [7], the objective of which is with H=20 is better than the one obtained with the fuzzy
to control the level of the liquid in a tank. PID type controller. Further simulations, shown in Fig. 8
and Fig. 9 are carried out with this value of H to determine
athis s in
rates in m3/s a3sfor ndpQut
are outlet
for input and the respectively. floThe maximumvl. the best inparameters
shown Fig. 9 hasof the
a fast Controller.
Greyrise time andThereasonable
response
controlled input liquid flow rate qin iS given by: overshoot.
qi Qin .sin (9 (t)) 9
(t) E [0,2T /2] (14) 1_4

The output liquid flow rate qo01t (that equals Qo01t since no 1.2
control is applied) is defined as:
qout aout 12gh(t)
0.8
where aout is the surface area of the outlet and g is the >

gravitational constant. -0.6

LT
_AXXA
The output variable h, which is the level of the liquid, is 0
calculated as:
-

h(t)=h(O) +-XJ(q1,,(r) -q01,(r))dr)d(15)5)IFO

0 0 50 100 150 200 250 300 350 400 450 500
Time(s)
where A is the surface area of the tan:k. The numerical
values used in this paper are A=1 in2, a011t=O.O1 in2, Figure 5
Qi,=O. 12m3/s and h(O) o. Step responses of the system to PDFC (oc=30, ,B=0) and PIFC (oc=0,
f3O. 1)

294

Authorized licensed use limited to: ULAKBIM UASL - BOGAZICI UNIVERSITESI. Downloaded on February 19, 2009 at 07:59 from IEEE Xplore. Restrictions apply.
 1A,X~ ~ VI
E. Kayacan, 0. Kaynak Grey Prediction Based Control of a Non-Linear Liquid Level System Using PID Type Fuzzy Controller

1.4 1.2

E 08
0.2

e. -_ -e- --

ie()Ties
T
0.6
-j
- Figure-6 Figure9 ----_

01

0.2 - (c-5- 02-.5--) -f 0.2 c- .4 H20)-

l.A Figure 7 PIF 9

PIDC with he bandlimitedwhite nise atCheNoutpt.PItFi
Step responses oft conventional an Grey PIDFtypeCfuzzyocontrollerscwithwit_difGrey cofO
(oc=5, P=0.5) PseItatwhlDteFreCPDC it ge p redito shows
(,B =0.2, oc=2.4, H=20)
t0.8~~~~~~~~__ PIOFO 0.8
n~~~~~~~~~~Ge PIFO H=0 Fig. 10 shows the unit step response of the system to grey
U) 0

1.4 Grey with the band-limited white noise at the output. It is

PIDFC
ll ~~~~~~~ ~~ ~ ~ ~~ ~~~~~~~~~~~0.2
,-\ robust characteristics compared to a classical PIDFC.
~~~~~~~~~~more
0 20 40 60 80 100 120 140 160 180 200 0.4 0 6 0 10 12 4 6 8 0
Time(s)Ties
60801010.5
difrn prdcto hoizn 140 C6 80200 20 40 60 80 100 120 140 160 180 200
Time(s) ~~~~~~~~~~~~~~~~~~~~~Time(s)

Figur 6_ _ _ Figure 10

1.2~~~~~~~~~~~~~~~~~~~~~~~~~~~.

Grey PIDFCO
02~~~~~~~~~~~~~~~~~~~~~otu mesrmn
Grey.2PIDFC,, H=2
208)
~0.6 20 * 40 60 80 100
Time(s)
120 140Gre 160
PIFC180
H=3 200 _ sut 0.6this-difficulty
handle __ results p wouldvery-much
e ic t be-welcome.
g In-thi
prediction
model can forecast the future outputs of a grey system toibe
_ VI CON LUS ON

0.4
-08 In life,,the0
paper, it-is-show-that
real
grey predicion approach is-a
mahmaia moe o hyiasse

Figure 8 10
Grey.P4Dtypefuzzy controller structure with differt c
0.2
s GreefiietyayofcotolinIhghyConliea,
mGrey
sytes.Th ppoc cntolerdecibd s- cmbnaio-o
ncrti
PID type fuzzy controller with the band-limited white noise at the
ge
Grey PIDFCOH3 prdcto .
with---
X.output
measurement
a-PD-tpe-uzz cotolrThe-
(,=0.2, oc=2.4, H=20)

1,
F
O 0.8 - 1/ r T -
~~~~~~~~~~~~~~In
real life, the mathematical model of a physical system
~~~cannot be defined exactly; there are always some
l5 06 _ / : : _ ~~~~~uncertainties. A control method that has the ability to
this difficulty would very much be welcome. In this
~~~~~~~~~~~~handle
0.4 ---------- paper, it is shown that grey prediction approach is an
way of controlling highly non-linear, uncertain
~~~~~~~~~~~~efficient
-

l
0.2 - -_ systems. The controller described is a combination of grey
~
i~
~~~~~~~~~-Gr_,ryPDF prediction approach with a PID type fuzzy controller. The
/ \~~~ ~ ~~~~~~~~9

Authorized licensed use limited to: ULAKBIM UASL - BOGAZICI UNIVERSITESI. Downloaded on February 19, 2009 at 07:59 from IEEE Xplore. Restrictions apply.
 ICM 2006 * IEEE 3rd International Conference on Mechatronics
REFERENCES [5] N. Engin, J. Kuvulmaz, and E. Omurlu, "Fuzzy Control of an
[1] Y. Lin and S. Liu, "A Historical Introduction to Grey Systems ANFIS model representing a nonlinear liquid level system," Neural
Theory," IEEE International Conference on Systems, Man and Computing andApplications, Vol. 13, pp. 202-210, 2004
Cybernetics, The Netherlands, Vol. 1, pp. 2403-2408, 2004 [6] D. F. Wang, P. Han, W. Han, and H. J. Liu, "Typical Grey
[2] J. L. Deng, "Introduction to Grey System Theory," The Journal of Prediction Control Methods and Simulation Studies," International
GreySystem, Vol. 1, pp. 1 G24,
1989. Conference on Machine Learning and Cybernetics, Xi'an, pp. 513-
[3] S. Liu and Y. Lin, "An Introduction to Grey Systems: Foundations, 518, 2003
Methodology
Methoolog Applications", IIGSS AcaemicPublsher
and andApplcatins",IIGS Academic Publisher, the [7] E. 0. Doebelin,
theDesign, Systems Dynamics: Modeling, Analysis Simulation,
Marcel Dekker, 1998
International Institute for General Systems Studies, Inc., 1998 Design, 1998
"Design
[4] W. Qiau and M. Muzimoto, PID type fuzzy controller and [8] 5. Y. Wang, "Design of a Linear-Like Fuzzy Grey Prediction
parameter
parametr adaptive method, Fuzzy Sets and Systems, Vol. 78, pp. Controller for a Three-Phase Induction Motor Speed Control
System," Conference on Grey Theory and its Applications, pp. 457-
23-25, 1996 463, 1998.

296

Authorized licensed use limited to: ULAKBIM UASL - BOGAZICI UNIVERSITESI. Downloaded on February 19, 2009 at 07:59 from IEEE Xplore. Restrictions apply.