An Adaptive Fuzzy Pid Control of Hydro-Turbine Governor: Xiao-Ying Zhang, Ming-Guang Zhang

Proceedings of the Fifth International Conference on Machine Learning and Cybernetics, Dalian, 13-16 August 2006
AN ADAPTIVE FUZZY PID CONTROL OF HYDRO-TURBINE GOVERNOR

XIAO-YING ZHANG, MING-GUANG ZHANG
School of Electrical and Information Engineering, Lanzhou University of Technology, Lanzhou 730050, China
E-MAIL: zhxy525@263.net
often regionally far from the consumer loads and

individually isolated. Consequently, classical methods were
used to adjust governor parameters for acceptable
performance by using a single-input-single-output control
structure [7]. But conventional governors did not respond
satisfactorily over the whole range of plant operation .In
addition, non-linearity exists in the models of the
hydro-turbine and the electric generator. A linear controller
designed to work optimally around a nominal operating
point of a non-linear system may be degraded at other
operating points. In order to overcome these difficulties, a
novel adaptive fuzzy control for hydro-turbine governing
system is proposed in this paper. The adaptive fuzzy
controller is applied to improve the overall performance of
hydro-turbine governing system and adjusts adaptively the
parameters as the operating conditions change. The
adaptive fuzzy system with a variable structure has the
advantage of not requiring an exact mathematical model of
the controlled plant and ensures that the overall system
remain asymptotically stable. Simulation results show that
the adaptive fuzzy governor has a satisfactory performance.
Abstract:
An adaptive fuzzy control strategy for hydro-turbine
governor is presented in this paper. Considering the complex
dynamic characteristic and uncertainty of the hydro-turbine
governor model and taking the static and dynamic
performance of the governing system as the ultimate goal, the
novel controller combined the classical PID control theory
with adaptive fuzzy control theory is designed. The presented
controller is a variable structure type; therefore its
parameters can be adaptively adjusted according to the
information of the control error. Simulation results show that
the proposed control strategy has adaptability, stability and
robustness and achieves good performance when applied to a
hydro-turbine governing system.
Keywords:
PID control; Adaptive fuzzy control; Hydro-turbine
governor; Simulation
1.
Introduction
In recent years, there has been considerable interest in

the application of modern control techniques to design
controllers for hydro-turbine governor [1], [2], [3].
Researches were shown that conventional controllers were
unable to perform optimally over the full range of operating
conditions and disturbance, due to the highly complex,
non-linear nature of hydro-turbine governing system [4],
[5]. The plant models are always inaccurate and the
governing system is a non-minimum phase system. The
unstable zeros can influence the stability and performance
of the system, so the design of a controller is known to be
quite difficult [6]. The early power systems were small,
2.
Hydro-turbine governing system models
The hydro-turbine governing system includes penstock

system, turbine, governor, electro-hydraulic servo system
and generator etc. The structure of the governing system is
shown in Figure.1.
2.1.
Model of turbine
Penstock
Q
r +
Governor
Electro-hydraulic
servo system
Mg0
_
Turbine
Mt
Figure 1. The hydro-turbine governing system

1-4244-0060-0/06/$20.00 2006 IEEE
325
Generator
Here, the Francis turbine is taken as example and the
flow and torque equations are described as:
Q = Q( H , n, )
(1)
M t = M t ( H , n, )
(2)
Where Q is flow rate, M t is turbine torque, H
is hydro-turbine water head, is gate opening, n is
hydro-turbine speed. When the parameters of turbine vary
in the small range at a stable operating point, the above two
equations can be linearized as:
q = eqy y + eqx x + qqh h
(3)
mt = e y y + eqx x + eqh h
inertia increases with Tw increasing.
flow to time. As Francis turbine, the simply transfer

function may be described as:
G (s) =
partial derivative of the torque to the water head of turbine.

q
, partial derivative of flow to gate opening.
y
q
, partial derivative of flow to speed of the turbine.
=
x
q
, partial derivative of flow to the head of turbine.
=
h
where
eqh
The plant time-vary parameters result in the transmission

coefficients of turbine ( e y , e x , eh , eqy , eqx , eqh ) variety
(5)
Where Tw is water inertia time constant, given as
(10)
coefficient.
T
h( s )
= 2hw th r s
q( s)
2
2.3.
(11)
Model of generator and load system
The mathematical model of generator and load is

given[4] as:
(Ta + Tb )
dx
+ eg x = mt mg 0
dt
(12)
dm g
dx
is derivative of load torque to
speed.
2.4.
[8]:
LQr
gFH r
Tw
= hw
Tr
where hw is penstock character

Substituting Eq. (10) into Eq. (8) gives:
frequency. e g =
The penstock system may be modeled as:
Tw =
(9)
Where Ta is unit inertia time constant, Tb is load inertia

time constant, mt is turbine torque, mg0 is load torque at rate
Model of penstock system
dq
h = Tw t
dt
2Tw Tr
( s +)
Tr 2 s 2
where Tr is penstock reflection time, is damp

coefficient. Regardless of any damp, then = 0 .result
in:
with the operating situation [7].

2.2.
(8)
is given as :
eqy =
eqx
(7)
T
h( s )
= 2 th( r s + )
2
q(s)
Where mt is the torque of turbine, q is the flow of
h is the hydro-turbine water head. e y = mt , partial

y
m
derivative of the torque to gate opening. e x = t , partial
x
m
derivative of the torque to speed of turbine. eh = t ,
h
mt ( s )
1 Tw s
=
1
y(s)
1 + Tw s
2
When the penstock is long and the elasticity of the

water can not be ignore, the equation accounting water
hammer is given as:
(4)
turbine, y is gate opening, x is hydro-turbine speed, and
dqt
is derivative of
dt
Model of electro-hydraulic servo system

The servo system model may be given as:
(6)
Where L and F are the length and cross-sectional area

of the penstock, Hr and Qr are the per-unit base values of
the water column head and that of the water flow rate,
respectively, g is the acceleration due to gravity. The water
Ty
dy
+ y=u
dt
(13)
where Ty is servo system time constant.

In practice, many components of a plant model are
precisely unknown. Furthermore, most plants are inherently
326
non-linear, and can be approximated by linear models only
in the near hood of the operating point. These and other
factors lead to uncertainty. An adaptive fuzzy controller
whose design is base on the models of governing system
should provide stability and performance requirements in
the presence of uncertainty.
3.
3.1.
labeled N.
In this layer compute the normalized firing strength of each
rule.
Layer 4: Every node i in this layer is a circle node with
an output:
zi = wi ( pi x + qi y + ri )
N4
Adaptive fuzzy control
z = wi ( pi x + qi y + ri )
A3
Layer2 Layer3
N z2
A4
y
A5
form as:
Layer4
z
C = ( wi x j , wi y j , wi : i = 1, N 4 ) : j = 1, N p
Figure 2. Adaptive Fuzzy Inference System

Layer 1: Every node i in this layer is a circle node with
a node function UAi(x) giving the membership of x to the
fuzzy set Ai which is in the form of a linguistic label (i.e.,
small, large). Usually, UAi(x) is given by:
1
bi
]T
C is the matrix of coefficients given by
Ai
U Ai ( x) =
parameters denoted by SC. The constructed AFIS will

update the parameters using Hybrid Learning Rule.
From the proposed AFIS architecture, given the values
of SP , the set of output (16), each corresponding to a
point ( x j , y j : j = 1, N p ) , can be expressed in matrix
such that s = pi , qi , ri : i = 1, N 4
z1
zi
the set of consequent
Z = Cs
(17)
Where Z is the vector of outputs of size NP each calculated
using (16). s is the vector of consequent parameter values
Layer1
A2
{pi , qi , ri : i = 1, N 4 } is
where
The design objective of the fuzzy controller is to learn

to achieve good performance of the nonlinear system in
presence of disturbances and uncertainties. The Adaptive
Fuzzy Inference System(AFIS) architecture is shown in
Figure 2, where node functions in the same layer are of the
same type, as described below.
(16)
i =1
Fuzzy controller design
A1
(15)
(14)
x ci 2
1 + (
)
ai
where {ai , bi , ci : i = 1, N1} is the set of premise
parameters which is denoted by Sp . As the values of these

parameters change, the bell-shaped functions vary
accordingly, thus exhibiting various forms of membership
functions on the linguistic label Ai.
Layer 2: Every node in this layer is a circle node
labeled that multiplies the incoming signals and sends the
product output, which represents the firing strength (or
weight) of a rule.
Layer 3: Every node in this layer is a circle node
Each epoch of the hybrid leaning procedure is

composed of a forward pass and a backward pass. In the
forward pass, input data is processed using available values
of SP to determine the elements of matrix C. Then the over
determined set of equations(17) is solved using a least
squares method to determine the values of SC. After
identifying these parameters the functional signals keep
going forward until the error measure is calculated. In the
backward pass, the error is propagated from the output end
toward the input end, and the parameters in SP are updated
by the gradient descent method.
3.2.
Principles of the control strategy
Corresponding to the operating situations of the

turbine governing system, the output of the fuzzy controller
needs seven states[5]:
{ N B , N M , N S , Z O, PS , P M , P B}
(18)
Where NB is Negative Small, NM is Negative
Medium, NS is Negative Small, ZO is Almost Zero, PS is
Positive Small, PM is Positive Medium, PB is Positive Big.
Since both input and output of the controller are fuzzy
variable, the controller is designed base on fuzzy logic
control theory. To generate this fuzzy output, it is necessary
to fuzzify the input of the controller. The fuzzy rules are
if-then rules, membership functions are simple triangular
327
ones. The fuzzy rule base consists of a collection of fuzzy
If-Then rules in the following form:
l
R (l ) : If x1 is F1 and and xn is Fn ,Then y is G l .

T
Where x = ( x1 , x2 ,..., xn ) U and y IR are the

input
and
output
l
of
the
fuzzy
control
system,
respectively, Fi and G are labels of fuzzy set in
U i(i = 1,2,...n) and IR, respectively, and l = 1,2,...M ,

where M is the total number of fuzzy If-Then rules in the
rule base. The fuzzifier maps a crisp point
x = ( x1 , x2 ,...xn )T U into a fuzzy set defined on U. It
From the Figure 3(a) with classical control, we can

find the setting time is 13s and the overshoot is 37.4%. But
in Figure 3(b) with adaptive fuzzy control, the setting time
is 8.9s and overshoot is nearly 0.
A 1% frequency disturbance is applied to the speed
reference and the response curves are obtained in
Figure4(a)-(b). The setting time is 12s and overshoot is 1%
in Figure.4(a), however, the setting time is 8s and overshoot
is nearly 0.
It is observed that the response with the proposed
controller is less oscillatory and has a smaller frequency
swing. The performance of the turbine governing system is
improved by the novel controller.
provides input fuzzy sets for the fuzzy inference engine.

The range of the speed error e and its derivative e/ is [-6,6].
The fuzzification control-rule and defuzzification steps are
all embedded in the final formulation of the fuzzy control
law.
4.
Simulation and results
Simulation studies have been performed on the system

shown in Figure.3 using the adaptive fuzzy controller.
yr(t)
+
PID
Adaptive
Fuzzy
controller
Hydroturbine
Governing
system
(a) classical control

(b) adaptive fuzzy control
Figure 4. The dynamic curves for 30% load rejection
y(t)
Figure 3. Adaptive fuzzy control system structure of

hydroturbine governing system
Where e=yr-y., The gains such an KI, KP, and KD have

to be carefully reconsidered when the system is subjected to
a new condition.
Several simulations are run after changing a number of
system parameters in order to ascertain the sensitivity of the
control to these changes. Assuming that the power system
was exposed to small changes in load during its normal
operation, the model would adequately represent its
dynamics. The parameters used in the simulation and their
values are:
Ta=6s,Tw=1s,Ty=0.1s,ex=0.8,ey=1.3,eh=1.1,
eqx=0.1,eqy=1,eqh=0.4.
The load test is an important design criterion and a
step load change may provide an indication about the
system stability. A step load change of 0.3 was applied to
the governing system and the time-domain responses for
classical PID controller and adaptive fuzzy controller were
shown in Figure 3(a) and (b), respectively.
(a) classical control

(b) adaptive fuzzy control
Figure 5. The response for 1% frequency disturbance
5.
Conclusions
In this paper, an efficient approach to the

hydro-turbine governor is presented. Taking the dynamic
and static performance as the ultimate goal, the control
strategy improves the level of intelligent decision making
and can adapt to various working requirements. A series
simulations for a hydro-turbine governor show that the
adaptive fuzzy control method not only has simple design
procedure and strong robustness, but also achieves static
and dynamic performances of the governing system under
different run condition. The satisfactory simulation results
prove that the control method proposed in the paper is
328
correct and practical.
[4]
Acknowledgements
This work supported by Gansu Provincial Natural
Science Foundation of China (No. ZS032-B25-027).
[5]
References
[1] O.P.Malik, G.S. Hope Ye Luqing, Wei Shouping and
Xu Haibo,An intelligent self-improving control
strategy with a variable structure and time-varying
parameters for water turbine governor Revue
Generate de 1Electricite, Pairs, No.4, pp.23-35, 1989.
[2] G.Orelind, Optimal PID gain schedule for
hydrogenerators design and application, IEEE Trans.
on Energy Conversion, Vol.4, No.3, Sept, pp.263-297,
1989.
[3] Mansoor SP, Jones DI, Bradley DA, Aris FC, Jones
GR,Reproducing oscillatory behaviour of a
hydroelectric
power
station
by
computer
[6]
[7]
[8]
329
simulation.Control Eng Practice 8, pp. 1261-1272,

2000.
Hannett LN, Fardanesh B,Field test to validate hydro
turbine-governor
model
struvture
and
parameters.IEEE Trans Power Systems 9,
pp1744-1751, 1994.
Armansyah
F,
Yorino
N,Sasaki
HRobust
synchronous voltage sources designed controller for
power system oscillation damping.Elect Power
Energy Systems 24, pp41-49, 2002.
Eilts LE, Schleif FR,Governing features and
performance of the first 600 MW hydrogenerating unit
at Grand Coulee.IEEE Trans PAS92, pp457-466,
2003.
Ye Luqing, Wei Shouping, Li Zhaohui, O.P. Malik
and G.S.Hope, Field tests and operation of a
duplicate microprocessor-based governor for water
turbine and its further development.,IEEE Trans on
Energy Conversion, Vol.5 No.2, pp.25-231, 1990.
I.Eker, M.Tumay, Robust multivariable-cascade
governors for hydroturbine controls.Electrical
Engineering 84, pp.229-237, 2002.

An Adaptive Fuzzy Pid Control of Hydro-Turbine Governor: Xiao-Ying Zhang, Ming-Guang Zhang

Uploaded by

Copyright:

Available Formats

An Adaptive Fuzzy Pid Control of Hydro-Turbine Governor: Xiao-Ying Zhang, Ming-Guang Zhang

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

An Adaptive Fuzzy Pid Control of Hydro-Turbine Governor: Xiao-Ying Zhang, Ming-Guang Zhang

Uploaded by

Copyright:

Available Formats

Proceedings of the Fifth International Conference on Machine Learning and Cybernetics, Dalian, 13-16 August 2006

AN ADAPTIVE FUZZY PID CONTROL OF HYDRO-TURBINE GOVERNOR

often regionally far from the consumer loads and

In recent years, there has been considerable interest in

Hydro-turbine governing system models

The hydro-turbine governing system includes penstock

Figure 1. The hydro-turbine governing system

inertia increases with Tw increasing.

flow to time. As Francis turbine, the simply transfer

partial derivative of the torque to the water head of turbine.

The plant time-vary parameters result in the transmission

Where Tw is water inertia time constant, given as

Model of generator and load system

The mathematical model of generator and load is

is derivative of load torque to

where hw is penstock character

The penstock system may be modeled as:

Where Ta is unit inertia time constant, Tb is load inertia

Model of penstock system

where Tr is penstock reflection time, is damp

with the operating situation [7].

Where mt is the torque of turbine, q is the flow of

h is the hydro-turbine water head. e y = mt , partial

When the penstock is long and the elasticity of the

turbine, y is gate opening, x is hydro-turbine speed, and

Model of electro-hydraulic servo system

Where L and F are the length and cross-sectional area

where Ty is servo system time constant.

Adaptive fuzzy control

Figure 2. Adaptive Fuzzy Inference System

C is the matrix of coefficients given by

parameters denoted by SC. The constructed AFIS will

the set of consequent

The design objective of the fuzzy controller is to learn

Fuzzy controller design

where {ai , bi , ci : i = 1, N1} is the set of premise

parameters which is denoted by Sp . As the values of these

Each epoch of the hybrid leaning procedure is

Principles of the control strategy

Corresponding to the operating situations of the

R (l ) : If x1 is F1 and and xn is Fn ,Then y is G l .

Where x = ( x1 , x2 ,..., xn ) U and y IR are the

respectively, Fi and G are labels of fuzzy set in

U i(i = 1,2,...n) and IR, respectively, and l = 1,2,...M ,

x = ( x1 , x2 ,...xn )T U into a fuzzy set defined on U. It

From the Figure 3(a) with classical control, we can

provides input fuzzy sets for the fuzzy inference engine.

Simulation and results

Simulation studies have been performed on the system

(a) classical control

Figure 3. Adaptive fuzzy control system structure of

Where e=yr-y., The gains such an KI, KP, and KD have

(a) classical control

In this paper, an efficient approach to the

simulation.Control Eng Practice 8, pp. 1261-1272,

You might also like