DRPT2008 6-9 April 2008 Nanjing China

Design of Additional Mix Robust Governor

Control for Hydraulic Turbine Generator
CHEN Jian, LIANG Gui-shu, DONG Qing

Considered the nonlinear and time-varying property of

Abstract-- In order to reduce the adverse effects from the hydraulic turbine regulating system, a robust pole assignment
uncertainty of Tw, the hydraulic turbine generator additional controller was proposed in [2]. Reference [3] has proposed a
robust mix governor control (HGMRC) is designed, in which a nonlinear adaptive control strategy on the basis of
measurable active power disturbance is used as the feedback,
comprehensive nonlinear model with inaccurate parameter by
and the robust control theory is used to deal with the uncertain
system. Considering the difficulties in the process of solving the utilizing the differential geometry method and adaptive
robust control problem, a new approach is presented in the control theory.
paper based on linear matrix inequalities (LMI) on the robust A novel nonlinear robust governor control for hydro-
control system. turbine generator sets is suggested in [4] and the nonlinear
The simulating tests of HGMRC used on hydraulic turbine robust control law for the speed governor of hydro-turbine
generator had been done by three-phase short circuit in single generators has been derived. But the study for Tw is relatively
machine and infinite system. The results of the simulating test
show that: the rotor oscillation of hydraulic turbine generator
less, the uncertainty of parameter Tw is serious for the
can be restrained by HGMRC; HGMRC has robustness for Tw stability of power system.
parameter perturbation. In this paper, the Hydraulic turbine generator additional
mix robust governor control (HGMRC) is designed. And the
Index Terms-- Hydraulic turbine generator, LMI, Parameter performance of HGMRC in simulating test is verified.
uncertainty, Robust control theory.
II. ROBUST CONTROL THEORY
I. INTRODUCTION

W ITH the increasing capacity of single hydraulic turbine

generator unit and the extensive use of hydraulic
turbine generators, more and more stability issues of
A. Robust Control Theory to Restrain the Measurable
Disturbance
Consider the following uncertain system with uncertainty
hydraulic turbine generator attract people’s attention. For parameter:
instance, the water hammer effects on the stability of system  x = ( A + ∆ A) x + B ω + ( B + ∆ B )u
and the way to improve the stability of hydraulic turbine 1 2 (1)

generator. And with the expanding of the power system, the  z = Cx + Du
low frequency oscillation (LFO) issue caused by hydraulic where x∈Rn, w∈Rr, z∈Rm, u∈Rp, A, B1, B2, C and D are
turbine generator has become more and more evident in some known constant matrices, ∆A and ∆B are uncertain matrices
areas. So, it is necessary to research the control method to and satisfied:
improve the stability of hydraulic turbine generator. [∆A ∆B ] = E Σ [ Fa Fb ] (2)
In the conventional power system transient stability
In equation (2), E,Fa and Fb are known constant matrices. Σ
calculation, a simplified linear model is often used as the time
constants Tw is unchanged, and a hydraulic turbine generator is unknown matrix with Σ∈Ω, and Ω={Σ |ΣTΣ≤I }
is considered to be able to maintain the same linear In the control system above, a state feedback H∞ control
characteristics near the operating mode. But in large or long- law is shown in (3).
term fluctuations in transition process, the uncertainty of Tw u = Kx (3)
will have a greater impact on the stability of the power The system satisfies the following H∞ performance conditions:
system. To solve this problem, many theories and methods (1) For any Σ∈Ω, the closed-loop system is internal stable.
were used, but the results are not satisfied. The rapid (2) For any Σ∈Ω, the transfer function between ω and z of
development of the robust control theory in recent years the closed-loop system, Tzω(s), is satisfied in (4).
makes it possible to solve this problem. Tzω ( s) < 1 (4)
A general robust control theory was used to enhance the
If the condition, DT[D C]=[I O], is satisfied, the system
capabilities of anti-interference with control system [1].
shown in (1) could be deduced in (5).
CHEN Jian, LIANG Gui-shu and DONG Qing are with School of  x = ( Ak + ∆ Ak ) x + B1ω
Electrical and Electronic Engineering, North China Electric Power  (5)
University, Baoding AL 071003 China (e-mail:[email protected];  z = Ck x
[email protected]; [email protected]).

978-7-900714-13-8/08/ ©2008DRPT 2840

 DRPT2008 6-9 April 2008 Nanjing China

where Ak=A+B2K, Ck=C+DK, ∆A=EΣFk and Fk=Fa+FbK. positive λ, the positive definite result P and λ is obtained
The solving procedure is shown in the following theorem. by the Riccati inequality in (10). The controller K is
Theorem 1: It is assumed that G(s)=C(sI-A)-1B is strict conceived by equation (11).
canonical rational transfer function ,and matrix A is stability, K + R −2 ( B2T P + λ −2 FbT Fa ) = 0
The necessary and sufficient condition of G( s) ∞ < 1 is that The inequality in (10) could be rewritten as below.
there is the positive definite matrix P>0 satisfying the Riccati 1
AT P + PA + P( B1 B1T + λ 2 EE T ) P + C T C + 2 FaT Fa −
inequality: λ
PA + AT P + PBB T P + C T C < 0 (6) 1 1
( PB2 + FaT Fb ) R −2 ( B2T P + FbT Fa ) +
The procedure of prove was given in [5]. λ2 λ2
1 1
According to theorem 1, if an advisable positive definite {K + R −2 ( B2T P + FbT Fa )}T R 2 {K + R −2 ( B2T P + FbT Fa )} < 0
matrix P∈Rn×n is existed and satisfies the condition in (7).
λ 2
λ2
The above inequality is (13), so inequality of (10) is
( Ak + ∆ Ak )T P + P( Ak + ∆ Ak ) + PB1 B1T P + CkT Ck < 0 (7) established for any Σ∈Ω.
It could be rewritten in (8).
B. The Robust Control for Restraining Measurable
( A + E ΣFa )T P + P( A + E ΣFa ) + PBBT P + C T C < 0 (8)
Disturbance and Parameter Uncertain
The necessary and sufficient condition is that a positive
From the above analysis, the inequality of (10) could be
λ is existed and satisfies the condition in (9).
rewritten as it in (14).
AT P + PA + P( BBT + λ 2 EE T ) P + C T C + FaT Fa < 0 (9) 1 1
Theorem 2: If the feedback matrix K and positive definite ( AT − 2 FaT Fb R −2 B2T ) P + P( A − 2 B2 R −2 FbT Fa ) +
λ λ
matrix P is existed with any Σ∈Ω, the necessary and P ( B1 B1T + λ EE T − B2 R −2 B2T ) P + (14)
sufficient condition is that there is a positive λ satisfied the T 1 T 1 T −2 T
Riccati inequality in (10). C C + 2 Fa Fa + 4 Fa Fb R Fb Fa < 0
λ λ
1 The parameters are defined in (15).
AT P + PA + P( B1 B1T + λ 2 EE T ) P + C T C + 2 FaT Fa −
λ 1
1 1 A1 = A − 2 B2 R −2 FbT Fa
( PB2 + 2 FaT Fb ) R −2 ( B2T P + 2 FbT Fa ) < (10) λ
λ λ B = B1 B1T + λ EE T − B2 R −2 B2T = B3 B3T (15)
1 1 1 T 1 T
−{K + R −2 ( B2T P + 2 FbT Fa )}T R 2 {K + R −2 ( B2T P + 2 FbT Fa )} T
Q = C C + 2 Fa Fa + 4 Fa Fb R Fb F −2 T
λ λ λ λ
In order to obtained a positive and definite result P, where The inequality of (10) is rewritten in (16).
R 2 = I + λ −2 FbT Fb . The sate feedback controller of the A1T P + PA1 + PB3 B3T P + Q < 0 (16)
closed-loop system is given in (11). The above inequality is given by (17).
1  A1T P + PA1 + Q PB3 
K = − R −2 ( B2T P + 2 FbT Fa ) (11)
λ  <0 (17)
 B3T P −I 
Prove: The necessity is proved at first.
Above all, if there exist K and P>0 satisfies (7), there exist It is easy to solve the result of (17) by LMI tool-box in
a positive λ satisfied the condition in (12) MATLAB. The function of feasp could be used to solve it.
1
AkT P + PAk + P ( B1 B1T + λ 2 EE T ) P + Ck T Ck + F T F < 0 (12) III. THE MATHEMATICAL MODEL OF HYDRAULIC TURBINE
λ2 k k GOVERNOR SYSTEM
. If Ak, Ck and Fk are substituted, the above inequality could
be rewritten as: The hydraulic turbine governor system is internally stable,
because a PI controller is adopted for the normal operation
1
AT P + PA + P( B1 B1T + λ 2 EE T ) P + C T C + FaT Fa + and the purposed of the additional controller is to restrain the
λ2 disturbance. So the condition, DT[D C]=[I O], is satisfied in
1 1
( PB2 + FaT Fb ) K + K T ( B2T P + FbT Fa ) + the system. Hence the hydraulic turbine governor system
λ 2
λ2
1 satisfies the robust H∞ performance.
K T (I + FbT Fb ) K < 0 If we analyze the stability of power system and extrude
λ2
So the important point of issue, the simplified model of
hydraulic turbine generator is expressed in (18).
1 T
AT P + PA + P( B1 B1T + l 2 EE T ) P + C T C + Fa Fa − 1 − sTw
l2 G ( s) = (18)
1 T 1 1 + 0.5sTw
( PB2 + 2
Fa Fb ) R -2 ( B2T P + 2 FbT Fa ) < (13)
l l Tw is concern about the following parameter in (19).
1 1
−{K + R -2 ( B2T P + 2 FbT Fa )}T R 2 {K + R -2 ( B2T P + 2 FbT Fa )} LV
l l Tw = , Tw ∈ [0.5,8]s (19)
gHT
It is very clearly that P satisfies the Riccati inequality.
Then the sufficiency is proved. If there existed appropriate where L is the length of aqueduct (unit is m), V is the speed

2841
 DRPT2008 6-9 April 2008 Nanjing China

of stream current in aqueduct (unit is m/s), g is gravitational Pm = Pm1 + Pm 2

acceleration of the upright position of aqueduct (unit is m/s2), Pm1 = −2U g + 2kω ∆ω + 2u (23)
HT is height of water head (unit is m).
In the report of IEEE [6], the typical value of Tw is equal to 3
Pm 2 = (U g − 3kω ∆ω − u )
1s and the range is between 0.5s to 5s. Because of the 1 + 0.5sTw
hydraulic turbine time constant Tw is equal to 1s, it is bigger And (23) simplified to
than the inertia time constant, (the typical value is equal to ∆ Pm1 = 2kω ∆ω + 2u
0.1s), of hydraulic pressure servo transmission system in
speed governing system. In order to simplify analysis, we −2 (24)
Pm 2 = (∆ Pm 2 + 3kω ∆ω + 3u )
neglect the dead space and inertia time constant of hydraulic Tw
pressure servo transmission system. Hydraulic turbine is the With (21) and (24), the sate function of hydraulic turbine
mainly part of the model of hydraulic turbine governor control system is given by
system. The simplified diagram of governor system is shown
 D − 2 kω 1   2 
in Fig.1.  −   1 
 ∆ω 
 H H  ∆ω   −  H 
∆P  =    + H ∆ Pe + 
2   ∆ Pm 2   
 u (25)
 m 2   − 6kω −   −
6
 Tw Tw   0   Tw 
1 − sTw
  
1 + 0.5 sTw If the time constant of hydraulic turbine transfer function
(Tw) has some perturbation, it will be simplified into Taylor
series in (26).
Fig. 1. Simplified diagram of hydraulic turbine governor system 1 1 1
= − ∆Tw (26)
Tw (t ) Tw 0 Tw0 2
In Fig. 1, Ug is the given normal rated mechanical power,
KW is the magnification of speed feedback, u is the additional It is assumed that ∆Tw ≤ 0.1 . The proportion of perturbation
control, Pm is the output mechanical power. is posed in (27).
The operating function of the hydraulic turbine generator is  0.6 kω 0.2  0.6
given in (20). ∆ A(t ) x =  x+ u (27)
δ = ω − ω0  Tw0 Tw0  Tw0
According to the matrix of (25), all parameter matrix of
ω0 D ω
ω= Pm − (ω − ω0 ) − 0 Pe (20) uncertain system is given in (28).
H H H
 D − 2kω 1   2 
Eq′Vs − H  1
Pe = sin δ H  −  H 
xd′ Σ A=  , B1 =  H  , B2 =   (28)
 − 6kω 2    6
In equation (20), δ is power angle; ω is rotational speed; H −  0  −
 Tw Tw   T 
is inertia of rotor; D is coefficient damping; Pm is the   w
mechanical power; Pe is output electric active power E‘q The coefficient matrix of performance evaluation is shown as:
express the transient voltage; Vs is conference voltage of the W 0 0 
infinite system; x‘dΣ is express the integrated transient C1 =  1  ,D =   (29)
 0 W2  1 
reactance.
The fluctuation of active power is used as the measurable
The coefficient matrix of perturbation is denoted in (30).
key disturbance. The nonlinear problem of hydraulic turbine  0.1 
 , Fa = [ 6kω 2 ] , Fb = [ 0 6 ]
T
E= (30)
generator equation will be avoided when the measurable T
 w0 
disturbance is used as a feedback in the controller. By using a When equation of (28), (29) and (30) is substituted into
special filter to obtain the vibration of the disturbance, the (10), we will obtain (11). Equation of (11) is substituted into
hydraulic turbine generator additional controller will have (3) and the hydraulic turbine governor additional mixed
nothing to do with the setting output of generator active robust governor control law is shown in (31).
power. The design procedure of the controller is simple and
u = k1∆ω + k 2 ∆ Pm 2 + W3Q∆ Pe (31)
the complex relation with the system is all included in the
measurable disturbance [7], [8]. So (20) is simplified in (21). In equation of (31), W3 is weighting coefficient and Q is
H ω = ∆Pm − D∆ω − ∆Pe (21) equal to 0.6. Equation of (31) is the HGMRC.
The equation of (18) is equal with the equation of (22).
IV. SIMULATIONS
1 − sTw 3
= −2 + (22) In order to verify the effectiveness of the performance of
1 + 0.5sTw 1 + 0.5sTw
closed-loop system with HGMRC, simulations were
Then the output mechanical power of hydraulic turbine is performed on SIMLINK platform in MATLAB. The single
obtained in (23). machine and infinite system is shown in Fig. 2. The fault
tested in the system is three-phase short circuit.

2842
 DRPT2008 6-9 April 2008 Nanjing China

In Fig. 2, Sn=210MVA ； Un= 1.38KV, the load of port

with generator is P =0.4 ， Q =0.3. The Π-parameter of
transmission line is:XL0=2×10-3H/km,CL0=8.6×10-9F/km and
RL0=0.25Ω/km. The length of transmission line is 300km.
(P,Q)
fn(3)
∞
(a)
Fig. 2. Diagram of the single machine and infinite system

The time constant and rated point operational parameter of

generator and hydraulic turbine generator is given in Table I.
TABLE I
OPERATIONAL PARAMETER OF HYDRAULIC TURBINE AND GENERATOR

Xd X‘d H D Kω Td’ Tw Pm0

1.3 0.3 3.2 0.1 1 4.49 1 0.75

(b)
The information of TABLE.I is substituted into
(28),(29),(30), and select the weighting coefficient W1 and W2 Fig. 3. The results of simulation without HGMRC
in (29). All coefficient matrices are shown in (32) and (33).

-0.5938 0.3125  −0.3125  0.625

A=   , B1 =   , B2 =   (32)
 -1.875 −2   0   0 

T T
E = [ 0.1] , Fa = [ 6 2 ] , Fb = [ 0 6] (33)
We specified λ=0.2, the LMI will be founded. So we can (a)
use the function feasp in LMI tool-box. We obtain the
purposed controller.
K = [−0.9855 −0.3269] (34)
Equation (34) is substituted into (31). The simulation
model of single machine and infinite system will be
established in SIMLINK. We assume that the style of fault is
three-phase short-circuit, the fault position is the output-port
of generator, the beginning time of fault is the 1st second, the
period of fault is equal to 0.2s. After 0.2s, the fault will be cut
(b)
off.
The curve in Fig. 3 shows the system performance Fig. 4. The results of simulation with HGMRC
corresponding to three-phase short-circuit fault without the
purposed controller. It is easy to seen that the curve of rotate In the next simulation, we will verify the closed-loop
speed and active power have very larger oscillation. system performance with HGMRC when Tw perturbed. We
When we auto-add the HGMRC, in Fig. 4 shows the assume that the range of perturbation is [0.9, 1.1], because Tw
waveform of rotate speed and active power, it can be seen that is specified with 10%~-10% perturbations. We select the tow
the closed-loop system has good performance. The rotate external Tw=0.9 and Tw=1.1. In Fig.5 shows the result of
speed and the active power will be rapidly come back the simulation, HGMRC has the best control effect at the rated
rated operational point, once the three-phase short-circuit operational point, Tw=1. If Tw is equal to 0.9 or 1.1, HGMRC
fault clear. From the process that we compare the system also restrain the oscillation of speed and power. HGMRC is
responses without purposed controller with the system insensitive to parameter perturbation. In the deigned range,
responses with purposed controller, it can be seen that the the HGMRC has good performance.
purposed controller can restrain the oscillation of rotate speed
and reduce the oscillation time of speed. The HGMRC
improve the anti-disturbance ability and stability of power
system.

2843
 DRPT2008 6-9 April 2008 Nanjing China

is verified by the simulation test in single machine and

infinite system with three-phase short circuit fault. The result
of simulation shows that HGMRC can restrain the oscillation
of rotate speed and reduce the oscillation period of speed.
HGMRC is insensitive to parameter perturbation.
Hence, HGMRC will be helpful to improve the ability to
resist disturbance of hydraulic turbine generator and improve
power system stability.
（a）
VI. REFERENCES
[1] LIU Hui-juan, ZOU Mu-chun and YU Zhao-hui, "Research on H-
infinity Robust Control for the Hydraulic Turbine Governor with
the Electro-hydraulic Servo System," Chinese hydraulics
&Pneumatics, No.8, pp.36-39, Aug.2006.
[2] MENG Zuo-hong, CAI Wei-you, CHEN Guang-da and CHENG Yuan-
chu, "The Design of Optimal Robust Pole-placement Governor for
Hydraulic Turbine Regulationg System," Large Electric Machine and
Hydraulic Turbine, No. 4, pp.60-64, Aug. 2000.
[3] GUI Xiao-yang, MEI Sheng-wei, LIU Feng and LU Qiang, "Adaptive
Nonlinear Control for Hydraulic Turbine Governor," IEEE Trans.
(b) Proceeding of the CSEE, Vol.26, No.8, pp.66-71, Apr.2006.
[4] SUN Yu-song, SUN Yuan-zhang, LU Qing and SHAO Yi-xiang,
Fig. 5. Comparison of control result when Tw=1.1、1 and 0.9 "Resech on Nonlinear Robust Control Strategy for Hydroelectric
Generator’s Valve," IEEE Trans. Proceeding of the CSEE,
Vol.21,No.2,pp56-,59,65,Feb.2006.
[5] YU Li, Robust control-the way to deal with linearity matrix inequality,
Beijing, Tsinghua University Press, 2002, p.13
[6] IEEE Committee Report, Dynamic Models for Steam and Hydro
Turbines in Power System Studies, IEEE Trans, PAS-73, 1973.
[7] DONG Qing, YAN Xiang-wu, HOU Zi-li and CAO De-ling, "Analysis
effect of synchronous generator governor control on stability by using
transient energy," Electric Power, Vol.38, No.11, pp. 9-13, Nov.2005
[8] DONG Qing, GAO Shu and BAO Hai, "Governor Additional H ∞
Robust Decentralized Control of Synchronous Generator," IEEE Trans.
Proceeding of the CSEE, Vol.22, No.2, pp.47-51, Feb.2002.
(a)

VII. BIOGRAPHIES
CHEN Jian was born in Shandong, China,
in 1983. He received bachelor degree in
Shandong institute of architecture and
engineering, Jinan, Shandong, in 2005. And now
he is studying in the Department of Electrical
and Electronic Engineering (NCEPU) for
Master’s degree.
His current research interests include
analysis, operation and control in power system
(b)
Fig. 6. The control resulting of general lead-lag controller when parameter
perturbation
LIANG Gui-shu was born in Hebei, China, in 1961. He received the
Master’s degree from the North China Electric
In order to oversight the superiority of the purposed Power University (NCEPU), Baoding, China, in
controller. We compare the HGMRC with the conventional 1987. He is currently a Professor at NCEPU.
feedback controller. Fig. 6 shows the control effect of His current research interests include electrical
conventional feedback controller with Tw perturbation. It is network theory and its application in power
systems, electric power information processing,
easy to seen that the waveform of the rotate speed and the and electromagnetic compatibility in power
active power need longer period to comeback stabilize than system.
HGMRC. When Tw=0.9 and Tw=1.1, rotate speed and active
power have more oscillations, and longer period oscillations. DONG Qing was born
It is very clearly, the control effect of HGMRC is better than in Sichuan, China, in 1970. He received the
conventional feedback controller. Doctor’s degree from the North China
Electric Power University (NCEPU),
Baoding, China, in 2003.
V. CONCLUSIONS He is currently Associate Professor at
NCEPU. His current research interests
In this paper, an uncertainty system robust theory has been include analysis and control in power
applied to design the hydraulic turbine generator additional system.
mixed robust governor control. The control effect of HGMRC

2844