Nonlinear Analytical Rules Based Fuzzy Control for

the Hydro Turbine Governing System

Sha Liu, Yuanchu Cheng Luqing Ye
Dept. of Power Engineering School of Hydropower and Digital Engineering
Wuhan University Huazhong University of Science and Technology
Wuhan, China Wuhan, China
[email protected] [email protected]

Abstract—In this paper, a new fuzzy controller for hydro turbine The control output surface of a fuzzy controller with linear
governing systems is proposed by introducing nonlinear analytical rules (Į=0.5) is given in Fig. 1.
analytical rules into the fuzzy controller. The approach to tuning
the rules of the fuzzy controller online is discussed and the 1 − e− x
Let, U = U max (2)
parameters tuning method is proposed by comparing the 1 + e− x
proposed controller to conventional PID controllers. The
simulation results show that the proposed fuzzy controller can where: x = α e E + α c EC (3)
obviously improve the dynamic performance of hydro turbine
governing systems, and strengthen their robustness. Then a fuzzy controller (2) with nonlinear analytical rules
is established, and its control rules are adjusted by Įe and Įc.
Keywords-hydro turbine governing system; analytical rule; The control output surface of the proposed fuzzy controller
fuzzy control; nonlinear; PID algorithm with nonlinear analytical rules is given in Fig. 2. The
introducing of the nonlinear symmetrical function is arbitrary.
I. INTRODUCTION And the new controller is closer to the manual control than the
linear analytical controller. The schematic diagram of a hydro
Hydro turbine governing systems are the time-varying,
turbine governing system with the proposed nonlinear
nonlinear and nonminimal phase systems. Their dynamic
analytical fuzzy logic controller is given in Fig. 3. In the figure,
performance and stability are always concerned. In recent
Kr is a proportional factor, and Ku is a integrate factor.
years, with the appearance of the intelligent control technology,
many novel control strategies for hydro turbine governors have
been developed. III. THE RULE ADAPTIVE AND PARAMETERS TUNING
METHOD
Among these intelligent control methods, the fuzzy logic
control is easy to be realized and has a good prospective, due to A. Control Characteristic
its strong adaptability and high reliability. Today, various fuzzy
logic controllers have been developed based on fuzzy set When triangular membership functions are used, the control
theory for different applications[1]-[4]. To simplify the output of the proposed fuzzy controller is the algebraic
designing of the fuzzy controllers of hydro turbine governing presentation of input variables.
systems and guarantee the high adaptability of governing U = K e eα e k1 + K ecα c ek
 2 (4)
systems under different operation environments, a new fuzzy
control strategy is proposed by introducing nonlinear analytical where, k1 and k2 are the action coefficients related to E and
rules in Section II, and the method of on-line adjusting rules of EC respectively.
the proposed controller is presented in Section III of this paper.
Section IV presents the results of the simulation studies, while ∂U 1 2e − x
k1 = = U max (5)
Section V gives the conclusions. ∂E α e (1 + e − x )2

II. AN ADAPTIVE NONLINEAR ANALYTICAL FUZZY ∂U 1 2e − x

k2 = = U max (6)
CONTROLLER FOR HYDRO TURBINE GOVERNOR ∂EC α ec (1 + e− x ) 2
The control rule in the analytical fuzzy controller with an
As showing in Fig. 3, the crisp control output acts on the
adjustable factor is given below [5]:
electro-hydraulic mechanism can be written as(T is a small
U = [α E + (1 − α ) EC ] α ∈ (0,1) (1) time constant and omitted):

where, input E (error) and EC (error change) determine u = ( K r K e k1α e + K u K ec k2α ec )e

control output; Į is a weighting factor of E (or EC). (7)
+ Ku K e k1α e ³ edt + K r K ec k2α ec e

This work was supported in part by Ninth five-year science plan of China
government.

978-1-4244-2487-0/09/$25.00 ©2009 IEEE

 Figure 1. Control output surface of fuzzy controller with analytical rules Figure 2. Control output surface of the fuzzy controller with nonlinear
(Į=0.5). analytical rules (Įe=Įc=0.5).

rt

Figure 3. The schematic diagram of the hydro turbine governing system with nonlinear analytical fuzzy logic controller

The proposed nonlinear analytical fuzzy controller behaves 3) The decreasing of Įe and the increasing of Įc will
like a PID type controller with time-varying parameters. Its strengthen the damping of control system, when the
equivalent proportional gain, integral coefficient and derivative amplitude of EC increased and E keeps a constant as
factor (omitted T) are given as follows: well as the amplitude of both increased.
K P = Kuα c K ec k2 + K rα e K e k1 (8)
C. Parameters Tuning
K I = Kuα e K e k1 (9) Referring to human emulating intelligence rules [6], the
tuning rules of integration factor are as follows:
K D = K rα c K ec k2 (10)
­0 where E × EC < 0 , δ1 ≤ E ≤ δ 2 and EC ≥ δ 3
°
B. The Rule Adaptive °K e < δ1
Ku = ® u1 (13)
Įe and Įc can be defined as follows: e > δ4 , δ4 > δ2
° Ku 2
E + α e0 °K others
αe = (11) ¯ u3
2
E + EC +1 0 ≤ K u 2 < K u1 < K u 3 (14)
E
The proportional constant can be written as:
EC 2 + α c0
E ( K u 0 − K u )α c
αc = 2
(12) Kr = + Kr0 (15)
E + EC +1 1 − αc
E
where, α c 0 + α e0 = 1 where: Ku0, Kr0 are the initial values of fuzzy controller, Įe
and Įc are decided by (11) and (12). Form (5) and (6), we can
From (4) and (5), we know: see k1=k2=Umax/2 when the system operates at steady state, and
the k1 (k2) will decrease with the increase of E or EC.
1) Įe=Įe0 and Įc=Įc0 respectively as E=0 and EC=0. Therefore, taking k1=k2=Umax/3 and (16), (17) are reasonable.
2) The decreasing of Įc and the increasing of Įe will 3K I
speed the system response, during the amplitude of E Ku 3 = Ku 0 = (16)
increased and EC is constant. K eα e0U max
 3K D According to the parameters tuning rules mentioned above,
Kr0 = (17) the parameters of the proposed controller are Įe=0.6, Kr=1.3
K ecα c 0U max
and Ku3=0.63 under 10% load disturbance condition at 50%
Generally, KP equals the conventional PID’s or is smaller rated load and 80.6m head as initial operating point. Further,
slightly, KD equals the conventional PID’s or more, and KI is taking Ku1=0.42, Ku2=0.21, T=0.05. Simulation studies at
about 1.5 times as that of conventional PID on account of the various operation points for both the proposed algorithm and
integral action is canceled during E × EC < 0 . conventional PID control algorithm have been performed on
hydro turbine governing system shown in Fig. 3 in MATLAB
environment. The simulation results of typical operating point
IV. SIMULATION STUDIES ON NONLINEAR ANALYTICAL under load disturbance are shown in Fig. 4 to Fig. 6 in which
FUZZY HYDRO TURBINE GOVERNING SYSTEM f=50(x+1) is frequency of the hydroelectric generating unit and
The mathematical models and the parameters used in the their dynamic performance indexes are given in TablG I.
simulations are described as follows.
The response curves under frequency set point step with the
1) The simplified electro-hydraulic mechanism same parameters are shown in Fig. 7 to Fig. 9, and their
performance indexes are given in TablGII.
dy
y + Ty =u (18)
dt TABLE I. THE INDEXES OF RESPONSE UNDER VARIOUS LOAD
DISTURBANCE AND HEAD CONDITIONS
where, Ty=0.2s is servomotor time constant, and the
segmented closure law is adopted to relieve the water hammer Load
Disturbance Over-
in penstock. Head
(%)
Control
shoot
Settling Wave
(m) Law Time (s) Numbers
(Hz)
2) The dynamic process of a generator and its load Initial Final

dx PID 3. 32 14 0. 5
61 50 20
Ta = mt − mg − en x (19) NAFUZ 3. 23 16 0. 5
dt
PID 0. 81 11 0. 5
80. 6 50 60
where, x is unit speed deviation, y is servomotor position, NAFUZ 0. 72 5 0. 5
mt is turbine torque, mg is load torque, Ta=9.46s is unit inertia PID 3. 25 20 0. 5
80. 6 100 80
time constant, and en=0.2 is load characteristic coefficient. NAFUZ 3. 08 13 0. 5
3) The rigid water hammer equation for single-penstock PID 1. 53 13 0. 5
113 50 20
single-turbine system NAFUZ 1. 48 15 0. 5
PID 1. 11 13 0. 5
113 100 80
dq NAFUZ 1. 12 13 0. 5
h = −Tw (20)
dt
where, h is head, q is discharge, and Tw=2.673s is flow
inertial time constant.
4) The turbine’s characteristic is often expressed by hill
chart.
The specific discharge Q11 and the specific torque M11 can
be calculated with table interpolation with input y and n11
respectively, and
nD1 1+ x Figure 4. 10% load rejection at 61(m) head.
n11 = = n11r (21)
H h0 + h

Q11
Then, Q = Q11 D12 H q=Q = h0 + h (22)
Qr Q11r

Mt M11 (23)
M t = M11 D13 H mt = = (h + h)
M tr M 11r 0

where, n11r is rated specific speed, Q11r is rated specific Figure 5. 20% load rejection at 80.6(m) head.
discharge, M11r is rated specific torque, and h0=H0/Hr is initial
head.
In the simulation, Ke=Kec=7, Umax=1, Q11r=1.12m3/s,
M11r=1160Nxm, n11r=83, Hr=80.6m, the gate opening is 95%
relative to rated model opening 40.8mm.
 TABLE II. THE INDEXES OF NO-LOAD FREQUENCY RESPONSE AT VARIOUS HEAD
Up Down
Head Control Over- Over-
Disturbance Settling Wave Settling Wave
(m) Law shoot shoot
time (s) numbers time (s) numbers
(%) (%)
PID 3. 5 6 0 20 17 0. 5
2Hz
NAFUZ 3. 5 5. 5 0 5 12 0. 5
113
PID 32 23 1 30 21 0. 5
4Hz
NAFUZ 0 11 0 2 16 0
PID 17. 5 17 0. 5 31. 5 18 0. 5
2Hz
NAFUZ 5 12. 5 0. 5 6 13 0. 5
80. 6
PID 21 19 0. 5 43 24 0. 5
4Hz
NAFUZ 4. 25 15 0. 5 11. 5 17. 5 0. 5
PID 16 18. 5 0. 5 27 21 0. 5
2Hz
NAFUZ 4. 5 14 0. 5 6 15. 5 0. 5
61
PID 19. 5 21 0. 5 41 24 0. 5
4Hz
NAFUZ 5. 2 17. 5 0. 5 9. 5 20 0. 5

V. CONCLUSIONS
A large number of simulation studies on hydro turbine
governing system show that the proposed controller has the
advantages of shorter settling time, smaller overshoot and
faster response, and can improve the dynamic performance and
strengthen the robustness of hydro turbine governing system.

Figure 6. 30% load rejection at 113(m) head. During process of control, the proposed controller can
adjust the fuzzy control rules according to the error and the
error change. Therefore, it can well solve the contradictory
requirement between the overshoot and the response time. In
addition, it doesn’t need to design the control rules that is
difficult for the designers of the ordinary fuzzy controllers, and
only needs to choose the initial control parameters according to
the experience of the process control with conventional PID
controllers. Hence, it is easy to implement in engineering
practice and has a good perspective for real-time control
Figure 7. 8% frequency reference step at 113(m) head.
systems.

REFERENCES
[1] T. Takagi and M. Sugeno, Derivation of Fuzzy Control Rules from
Human Operator’s Control Actions, Proceeding of the IFAC Conf. on
Fuzzy Information, Vol.1 Marseille, France, 1983: 55-60.
[2] T. Takagi and M. Sugeno, Fuzzy Identification of Fuzzy Systems and its
Application to Modelling and Control, IEEE Trans. on Systems Man
Cybernet, 15, 1985: 116-132.
Figure 8. 8% frequency reference step at 80.6(m) head. [3] Ying H., The Simplest Fuzzy Control using Different Inference Methods
are Different Nonlinear Proportional-integral Controllers with Variable
Gains. Automatica, 29(2), 1993: 499-505.
[4] Ying H., “General analytical structure of typical fuzzy controllers and
their limiting Structure Theorems,” Automatica, 28(4), pp. 1139-1143,
Apr. 1993.
[5] Minyou Chen and D. A. Linkens, “A hybrid neuro-fuzzy PID
controller,” Fuzzy Sets and Systems, 99(1), pp. 27-36, Oct. 1998.
[6] Zhou Qi Jian and Bai Jian KuoAn Intelligent Controller of Novel
Design. MICONEX’83 Proceedings of Multinational Instrumentation
Figure 9. 8% frequency reference step at 61(m) head. Conference. Shanghai, 1983: 137-149.