See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/224561310

Load Frequency Control for Multiple-Area Power Systems

Conference Paper  in  Proceedings of the American Control Conference · July 2009

DOI: 10.1109/ACC.2009.5160621 · Source: IEEE Xplore

CITATIONS READS
39 3,164

3 authors, including:

Zhiqiang Gao
Cleveland State University
184 PUBLICATIONS   7,729 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

ADRC: a new paradigm of the science of automatic control View project

Diesel Engine air system modeling and control View project

All content following this page was uploaded by Zhiqiang Gao on 16 May 2014.

The user has requested enhancement of the downloaded file.

 2009 American Control Conference ThB03.5
Hyatt Regency Riverfront, St. Louis, MO, USA
June 10-12, 2009

Load Frequency Control for Multiple-Area

Power Systems1
Yao Zhang2, Lili Dong2, *, Member, IEEE, Zhiqiang Gao, Member, IEEE

Abstract—This paper presents the development and application Control (AGC) through changing load reference inputs of
of an Active Disturbance Rejection Controller (ADRC) to governors of selected units in the control area, and then
regulate the frequency error for a three-area interconnected adjusting their outputs [1, 3-4] to track the reference inputs.
power system. As the interconnected power system transmits the
power from one area to another, the system frequency will
However, the reference signal is generally varying with
inevitably deviate from a scheduled frequency, resulting in a environmental variations that degrade the performance of the
frequency error. A control system is essential to correct the AGC greatly. In addition, multiple proportional and integral
deviation in the presences of external disturbances and structural parameters of the AGC make it very difficult to tune in the
uncertainties to ensure a safe and smooth operation of the power real world. Decentralizing the power system using modern
system. ADRC can extract the information of the disturbance control method such as Linear Quadratic Gaussian (LQG) [5],
from input and output data of the system and actively
compensate for the disturbance in real time. In addition, it has
state feedback [6] and H infinity [7] can possibly reduce the
the advantages of simple structure, few tuning parameters, and size of the inter-connected power system. But the structures of
robustness against parameter uncertainties over the traditional these controllers are too complicated to be implemented for
PID controllers used in current power industry. The effectives of industry practices. Recently, a process utilizing Linear Matrix
the controller are validated by both simulation results and a Inequalities (LMI) [8] or Genetic Algorithms LMI (GALMI)
frequency-domain analysis of the control system. [9] to tune the parameters of Proportional Integral Derivative
(PID) controllers has been developed for the multiple-area
Index Terms—load frequency error, active disturbance power systems. Although the GALMI based PID controllers
rejection based control, robustness, power system.
are simple and effective to some extent, they react passively to
the disturbances in the power system, such as temperature and
I. INTRODUCTION humidity change, machine wear and tear, and other
unpredictable frictional forces.
I nterconnected electric power generation systems utilize tie-
lines to transmit power from one area to another, either
scheduled via a contract or in support during a system
This paper aims to develop a practical Active Disturbance
Rejection Control (ADRC) solution to solve the LFC problem.
Compared to conventional PID controllers, ADRC is a
disturbance [1]. The “quality” of the power generating system technology that actively anticipates and fights disturbances
is defined by three factors: constancy of frequency, constancy before they affect system operation. The basic idea of ADRC
of voltage and level of reliability [2]. In actual power system is using an Extended State Observer (ESO) to estimate and
operations, the load is changing continuously and randomly, cancel the generalized disturbance (all the input efforts
resulting in deviations of the load frequency and the tie-line excluding the control effort) of the system in order to simplify
power between any two areas from scheduled generation the control problem. The designed controller based on this
quantities. Therefore, a Load Frequency Controller (LFC) is concept is easy to tune [10], does not need an accurate model
widely used to ensure a good quality of the power systems [10], is able to be decentralized [11] and has a notable
through regulating the deviations. robustness [12-13] against unexpected disturbances and
The tie-line deviations from scheduled values are defined as structural uncertainties. The ADRC has been applied to many
Area Control Error (ACE). It is a summation of the tie-line micro and mechanical systems [10-15]. It is the first time that
power deviation ΔPtie and the frequency deviation Δf we modified it and applied it to the LFC problem of a power
multiplied by a bias factor B [2]. The LFC is employed in each system. A decentralized robust LFC controller based on the
area of the interconnected power systems to drive the ACE to idea of the ADRC is introduced in the paper to reduce the
zero. ACE to zero.
A traditional LFC is performed by an Automatic Generation This paper is organized as follows. The dynamic model of
the generation power system is explicated in section II. The
1
The work was sponsored in part by the University Research Development design of LFC using the ADRC is introduced in section III.
Program of Cleveland State University under Grant 0210-0630-10 Simulation results of the control system are given in section
DONGL101.
2
The authors are with the Department of Electrical & Computer
IV. The performance analysis is in section V. Concluding
Engineering, Cleveland State University, Cleveland, OH 44115, USA. remarks are made in the last section of this paper.
*Corresponding author. Email: [email protected].

978-1-4244-4524-0/09/$25.00 ©2009 AACC 2773

 Fig. 1. Dynamic model of one-area power system

Num f ( s ) = −T ( H 2 H 3 R1 + H1 H 3 R2 + H1 H 2 R3 + H1 H 2 H 3 H s )
II. DYNAMIC MODEL OF THE MULTIPLE-AREA POWER (7)
SYSTEMS + TH1 H 2 H 3 ( Barea1 s + T ) s ,

We will design the LFC controller based on a three-area Den( s ) = H 2 H 3 s R1 + H1 H 3 s R2 + H1 H 2 s R3

, (8)
interconnected power system. A dynamic model of one-area + H1 H 2 H 3 H
power system is shown in Fig. 1. The model consists of three where
generation units, each of which is composed of three major T = T12 + T13 ;
parts: governor, turbine and generator. Governor catches load
change error and outputs valve position (PV). The physical H1 = (Tg1 s + 1)(Tch1 s + 1) ;
limitation of the valve position change is represented by a rate H 2 = (Tg 2 s + 1)(Tch 2 s + 1) ;
limiter in the figure. Turbine turns natural power into H 3 = (Tg 3 s + 1)(Tch 3 s + 1) ;
mechanical torque, which drives the generator to generate
electric power. A common non-reheat turbine unit is employed H = M area1 s 2 + Darea1 s + T .
in the paper. All generators in one area response coherently, so where
they are represented by an equivalent generator [2]. Droop Δ deviation from nominal values;
characteristic is provided to change the speed setting of the f area frequency;
governor so that several governors can operate in parallel. Tie- f D area interconnection signal;
line power deviation is proportional to the integral of the
D load damping constant;
frequency difference between the two areas connected with
(for one generation company);
the tie-line. The ACE output defined in previous section is M inertia constant;
equal to BΔf + ΔPtie as shown in the figure. (for one generation company);
For the conveniences of the controller design and the Tij tie-line synchronizing coefficient between area i & j;
performance analysis, we focus on Laplace transform Pl power load change;
representation of the system in this section. Let y(t) denote the
Ptie tie-line power flow;
ACE output of the system, and u(t) the control input to the
system. The Laplace transform of the system model is Pm mechanical power;
Tch turbine time constant;
Y ( s ) = GP ( s )U ( s ) + GD ( s )ΔPl ( s ) + G f ( s )Δf D ( s ) , (1) PV governor valve position;
where Tg governor time constant;
GP ( s ) = NumP ( s ) Den( s ) , (2)
R speed regulation characteristic;
GD ( s ) = NumD ( s ) Den( s ) , (3) α ramp rate factor;
G f ( s) = Num f ( s ) Den( s) , (4) PC load reference setpoint;
In (2), (3) and (4), B composite frequency response characteristic;
NumP ( s ) = (α1 H 2 H 3 + α 2 H1 H 3 + α 3 H1 H 2 )( Barea1 s + T ) , (5) (for one generation company);
Darea1 area load damping constant;
NumD ( s ) = (− H1 H 2 H 3 )( Barea1 s + T ) , (6)
M area1 area inertia constant;
Barea1 area composite frequency response characteristic;

2774
 3
Darea1 = ∑ Di ; L -1 ⎡⎣Gleft ( s)Y ( s) ⎤⎦ and external disturbances including the
i =1
3
terms of ΔPl and Δf D , and b=b0. Substituting the plant
M area1 = ∑ M i ;
i =1 parameters into b, we have
3 3 3 3
Barea1 = 1 ∑ (1 R ) + D
i =1
i area1
. b = ∑ αi ∑ Tg j Tch j Barea1 ΠT
i =1
gi Tchi M area1 , (15)
i =1 j =1, j ≠ i

In the next section, a LFC controller will be designed based

It is assumed that d is locally Lipschitz in the argument and
on the system model given in (1-8). The design objective is to
bounded within the domain of interests. Let x1 = y , x2 = y& ,
reduce the ACE to zero in the presences of disturbances and
parameter variations. x3 = &&
y , x4 = d , the state space equations of the model
represented by (13) and (14) are:
III. CONTROLLER DESIGN
In general, an nth order plant can be represented by (9): x& = Ax + Bu + Eh
(16)
y ( n ) = d + bu (9) y = Cx
where y is the output of the system, u is the control signal, b is where
the controller gain. The generalized disturbance, denoted as d, ⎡ x1 ⎤ ⎡0 1 0 0⎤ ⎡0⎤ ⎡0 ⎤
includes all the input efforts of the system excluding the ⎢x ⎥ ⎢0 ⎥
0 1 0⎥ ⎢ 0 ⎥ ⎢0 ⎥
control effort [15]. If this generalized disturbance can be x = ⎢ 2⎥ , A = ⎢ , B = ⎢ ⎥ , E = ⎢ ⎥ , h = d& ,
⎢ x3 ⎥ ⎢0 0 0 1⎥ ⎢b ⎥ ⎢0 ⎥
estimated and cancelled, the system will be simplified to an nth ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
order-integral plant, which is easy to control. A practical way ⎣ x4 ⎦ ⎣0 0 0 0⎦ ⎣0⎦ ⎣1 ⎦
to estimate d in real time is to use the ESO. The idea of ESO is C = [1 0 0 0 ] .
to treat the d as an extended state of a state space model of the From (16), a state space model of the system with the
system and to use an augmented Luenberger observer to extended state can be written as:
estimate the state.
In detail, the Laplace transform of the plant (1) can be z& = Az + Bu + L( y − yˆ )
(17)
rewritten as: yˆ = Cz

Den( s) NumD ( s) Num f ( s)

where ŷ is the estimated y , z = [ z1 , z2 , z3 , z4 ] is the
T
Y (s) = U ( s) + ΔPl ( s) + Δf D ( s) , (10)
NumP ( s) NumP ( s) NumP ( s)
estimated state vector of x, L is the observer gain vector and
L = [ β1 , β 2 , β 3 , β 4 ] . For simplifying the tuning process, the
T
where NumP ( s ) and Den( s ) are fifth-order and eighth-order
polynomials respectively. Conducting a polynomial long parameters in L are chosen as
division on Den( s ) NumP ( s ) yields ⎛ 4⎞
βi = ⎜ ⎟ ωOi , i = 1,L , 4 , (18)
⎝i⎠
b0 s + b1 s + b2 s + b3 + Gleft ( s ) ,
3 2
(11) so as to make all the eigenvalues of the ESO equal to –ωO
[15]. With accurate estimations of the ESO, z4 will track d
where b0, b1, b2, b3 are the coefficients of the quotient, and
closely, that is, z4 = dˆ ≈ d .
Gleft(s) is a remainder that is given by
For the LFC problem, the control goal is to reduce the ACE
Gleft ( s ) = ( a0 s 4 + a1 s 3 + a2 s 2 + a3 s + a4 ) NumP ( s ) , (12) to zero. Therefore, the reference signal to the ACE output is
chosen as r = 0 . Then the feedback control law based on the
estimation results of the ESO is:
Replacing the left side of (10) with (11) and (12) gives
NumD ( s )
b0 s 3Y ( s ) = ( −b1 s 2 − b2 s − b3 − Gleft ( s ))Y ( s ) + ΔPl ( s ) u = (u0 − z4 ) / b (19)
Num p ( s )
(13)
Num f ( s ) where u0 is a PDD controller and
+ ΔFD ( s ) + U ( s ),
Num p ( s )
Taking all the terms excluding U(s) on the right side of (13) as u0 = k1 (r − z1 ) − k2 z2 − k3 z3 , (20)
d(s), and transforming (13) from frequency domain to time
domain produces where the difference between the reference signal r and the
estimated ACE output z1 is defined as the tracking error e. In
y = d + bu ,
&&& (14) (20), the controller gains are selected as
⎛ 3 ⎞ 4 −i
where d includes all the information of &y& , y& , y , ki = ⎜ ⎟ ωC , i = 1,L ,3 , (21)
⎝4−i⎠

2775
In this way, the only tuning parameter of the controller is the ACE and Δf than that of the GALMI tuned PI controller.
controller bandwidth ωC and all the closed-loop poles are set However, the control effort of the ADRC shows an overshoot
to -ωC [15]. at the switching edge of the load change since it takes a short
Replacing u in (14) with (19) will yield time for the ESO to approximate the disturbance. Nevertheless
the overshoot magnitude of the ADRC is reasonable. So it will
&y&& = d + b((u0 − dˆ ) / b) ≈ u0 , (22)
not affect the implementation of the controller in practice.
Then the original complicated system model is simplified as a Random Load Changes
0.02
triple integrator as shown in (22). Using the ADRC controller 0.01

ΔPl1 (pu)
consisting of (19), (20), and (21), the tracking error e will be 0

driven to zero. -0.01 Area1

0 50 100 150 200 250 300
0.02
IV. SIMULATION RESULT

ΔPl2 (pu)
0.01

The ADRC is implemented on the three-area power system 0

-0.01 Area2
in Fig.1 with the parameters listed in Appendix. In this 0 50 100 150 200 250 300
section, the simulation results of the ADRC controller are
0.01
compared with the results of the GALMI tuned PI controller in

ΔPl3 (pu)
0
[9] when three different power load changes are applied to the -0.01 Area3
three areas of the power system respectively. The load 0 50 100 150 200 250 300
changes are functioning as external disturbances to the system. Time (Sec.)

They are ideal to test the robustness of the two controllers Fig.2. Random load changes for case one
against disturbances. The controller parameters for both the -3
x 10 Area 2
ADRC and the GALMI are listed in Table I and Table II. ACE (pu) 2
0
TABLE I: ADRC PARAMETERS -2
0 50 100 150 200 250 300
Order of ESO ωC ωO b x 10
-3
ADRC
Area 1 3 5 20 228 2 GALMI
Δf (Hz)

Area 2 3 5 20 234 0
Area 3 3 5 20 282
-2
0 50 100 150 200 250 300
TABLE II: PI CONTROL PARAMETERS FROM GALMI DESIGN [9] x 10
-3

2
ΔPe (pu)

Area 1 Area 2 Area 3 0

Kp -3.27E-04 -6.96E-04 -1.60E-04
Ki -0.3334 -0.3435 -0.3398 -2
0 50 100 150 200 250 300
Time (Sec.)

In case one, a random load change is added to each area of x 10

-3 Area 3
2
the power systems as shown in Fig.2, where ΔPl1 denotes the
ACE (pu)

0
random load change for area 1, ΔPl2 the random load change -2
0 50 100 150 200 250 300
for area 2, and ΔPl3 the random load change for area 3. Fig.3 x 10
-3
ADRC
2
shows the ACE, load frequency deviation Δf and the GALMI
Δf (Hz)

0
difference between control effort and load disturbance, which
-2
is ΔPe ( ΔPe = ΔPC − ΔPl ). In Fig.3, both the ADRC and the 0 50 100 150 200 250 300
-3
x 10
GALMI tuned PI controller compensate the load fluctuations 2
ΔPe (pu)

rapidly, that is, the ΔPe s are driven to zeroes under the control 0

efforts. However, the ACE, Δf , and ΔPe of the ADRC -2

0 50 100 150 200 250 300
controller have less magnitudes (the peak errors of the ACE Time (Sec.)

and Δf for the ADRC are no more than 0.05%) than the Fig. 3. System response for case 1
GALMI tuned PI controller.
In case three, two large, step-input load changes are added
In case two, a step load change with large amplitude is
to areas 2 and 3 with the magnitudes of ΔPl2 = 100MW and
added to each area. The amplitudes of the load changes for the
three areas are ΔPl1 = 100 MW , ΔPl2 = 80MW , and ΔPl3 = 50 MW , while the controllers of areas 2 and 3 are
assumed to be out of services. This case is to test the reliability
ΔPl3 = 50 MW respectively. The ACE, Δf and control effort
of the controlled system in tough situation. From the results
for both controllers are shown in Fig. 4. The ADRC shown in Fig. 5, we can see that the ACE and the Δf in areas
demonstrates smaller oscillations and faster responses in the

2776
2 and 3 can not converge to zeroes but only be bounded under Area 1

this situation. Though the ACE and Δf are still driven to 0.02

ACE (pu)
0
zeroes in area 1 under both the ADRC and the PI controller. -0.02 ADRC

Similarly, ADRC produces smaller oscillations and faster -0.04 GALMI

0 2 4 6 8 10 12 14 16 18 20
response in the results of the ACE and Δf than the PI
0
controller. However, the control effort of the ADRC is a little -0.02

Δf (Hz)
-0.04
bit bigger than the GALMI tuned PI controller. -0.06
-0.08
Area 1 0 2 4 6 8 10 12 14 16 18 20

0 0.02
ACE (pu)

ΔPc (pu)
0
-0.05
ADRC -0.02
-0.1 -0.04
GALMI
-0.06
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Time (Sec.)
0
Area 3
Δf (Hz)

-0.05 0 ADRC

ACE (pu)
-0.02 GALMI
-0.1 -0.04
0 2 4 6 8 10 12 14 16 18 20 -0.06
0 2 4 6 8 10 12 14 16 18 20
ΔPc (pu)

0.4
0.2 0

Δf (Hz)
-0.02
0
-0.04
0 2 4 6 8 10 12 14 16 18 20 -0.06
Time (Sec.)
0 2 4 6 8 10 12 14 16 18 20
Area 2 -3
x 10
0 1
ACE (pu)

ΔPc (pu)
-0.05 0
ADRC
-0.1 GALMI
-1
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Time (Sec.)
0 Fig. 5. System response for case 3
Δf (Hz)

-0.05

-0.1 Replacing the Z ( s ) in (24) with (25) gives

0 2 4 6 8 10 12 14 16 18 20

0.4
U ( s ) = GPF ( s )GEC ( s ) R ( s ) − GEC ( s )Y ( s ) , (26)
ΔPc (pu)

0.2

0
0 2 4 6 8 10 12 14 16 18 20
where GEC ( s ) is represented by
Time (Sec.)
−1
Fig. 4. System response for case 2 GEC ( s ) = K [T ( s) ] L b , (27)

V. STABILITY ANALYSIS and GPF ( s ) is a pre-filter represented by

In this section, the stability analysis of the closed-loop
power control system in one area is presented.
The transfer function between load reference signal and
⎣ ( −1
GPF ( s ) = ⎡k1 b − K [T ( s ) ] B ⎤
⎦ ) (bK [T (s)] L) .
−1
(28)

ACE has been given by (2). The Laplace transform of the ESO The closed-loop control system for the one-area power
in (17) is system is constructed in Fig. 6.
sZ ( s ) = ( A − LC ) Z ( s ) + BU ( s ) + LY ( s ) (23)

The Laplace transform of the PDD controller in (20) is

U ( s ) = k1 R ( s ) b − KZ ( s ) b , (24)
Fig. 6. Block diagram of the closed-loop control system
where K = [k1 k2 k3 1] .
The transfer function of the open-loop system is:
Substituting the U ( s ) in (23) with (24) yields
GO ( s ) = GEC ( s )GP ( s ) (29)
Z ( s) = [T ( s)] [ Bk1 R(s) b + LY (s)] ,
−1
(25)
The frequency responses of the open-loop transfer function
where T ( s ) = sI − A + LC + BK b . GO ( s ) with varying area inertial constants ( M area1 ) are shown
in Fig. 7 and the corresponding stability margins are listed in

2777
Table III. In Fig.7 and Table III, the inertia constant is Ramp rate × 5 min , in which the regulation requirement
α=
changing from 0.5M to 2M while the controller gains and regulation requirement
observer gains are remained unchanged. Such a big change of for each area is 100MW, and the Ramp rate is given in the
the inertia constant seldom happens in power systems and can Table IV. The tie-line synchronizing coefficients between any
be taken as an extreme condition. However, the ADRC can two areas are T12 = 0.2pu./ rad , T23 = 0.12 pu./ rad , and T31 = 0.25 pu./ rad .
still stabilize the power system with positive stability gains. The other parameters of the power plant are given in Table IV.
This demonstrates not only the stability of the control system
but also the strong robustness of the ADRC against parameter TABLE IV
GENERATING UNIT PARAMETERS
uncertainties.
Parameters Generation Companies
VI. CONCLUSION REMARKS MVA base
1 2 3 4 5 6 7 8 9
(1000MW)
This paper presents the development and application of the D (pu/Hz) 0.01500.01400.0150 0.0160 0.0140 0.01400.01500.01600.0150
ADRC to a complicated three-area interconnected power M (pu.sec) 0.16670.12000.2000 0.2017 0.1500 0.19600.12470.16670.1870
system. Such a system has zeros in its transfer function, Tch (sec) 0.4 0.36 0.42 0.44 0.32 0.40 0.30 0.40 0.41
making the controller design different from other systems in Tg (sec) 0.08 0.06 0.07 0.06 0.06 0.08 0.07 0.07 0.08
[10-15]. For the first time, the generalized disturbance of the R (Hz/pu) 3.00 3.00 3.30 2.7273 2.6667 2.50 2.8235 3.00 2.9412
B (pu/Hz) 0.34830.34730.3180 0.3827 0.3890 0.41400.36920.34930.3550
ADRC is extended to include the derivatives of the control α 0.4 0.4 0.2 0.6 0 0.4 0 0.5 0.5
input in the system model. Our simulation results successfully Ramp rate
8 8 4 12 0 8 0 10 10
demonstrate that the ADRC not only compensates the external (MW/min)
load changes, but also regulates both the ACE and frequency
deviations in the three areas. In addition, the ADRC is shown
REFERENCES
to be superior to the existing GALMI tuned PI controller in
[1] A. G. Morinec and F. E. Villaseca, “Continuous-Mode Automatic
smaller ACE and Δf and faster responses of the closed-loop Generation Control of a Three-Area Power System,” in Proc. of the
system. The frequency-domain analysis further validates the 33rd North American Control Symposium, pp. 63-70, 2001.
[2] P. Kundur, Power System Stability and Control (Book style). New York:
strong robustness of the controller against parameter variations McGraw-Hill, pp. 581-621, 1994.
and external disturbances. [3] N. Jaleeli, L. VanSlyck, D. Ewart, L. Fink, and A. Hoffmann,
“Understanding Automatic Generation Control,” IEEE Trans. Power
Bode Diagram Syst., vol. 7, NO. 3, pp. 1106-1122, Aug. 1992.
Marea1=0.4867
50
Marea1=0.2134
[4] V. Donde, M. A. Pai, and I. A. Hiskens, “Simulation and Optimization
Marea1=0.9734 in an AGC System After Deregulation,” IEEE Trans. Power Syst., vol.
0
16, pp. 481–489, Aug. 2001.
Magnitude (dB)

-50
[5] X. B. Chen, and S. Stankovic, “Overlapping Decomposition and
Decentralized LQG Control for Interconnected Power Systems,” Proc.
-100
IEEE Int. Conf. Syst., Man, Cybern., vol. 3, pp. 1904-1909, Oct. 1996.
[6] K. Y. Lim, Y. Wang, G. Gua, and R. Zhou, “A New Decentralized
-150 Robust Controller Design for Multi-Area Load-Frequency Control via in
90
Complete State Feedback,” Optimal Control Applications & Methods,
0 vol. 19, pp. 345-361, 1998.
Phase (deg)

-90
[7] M. H. Rahi and A. Feliachi, “ H ∞ Robust Decentralized Controller for
-180
Nonlinear Power Systems,” in Proc. of 30th Southeastern Symposium of
-270 Syst. Theory, Mar. 1998, pp. 268–270.
-360 [8] X. Yu and K. Tomsovic, “Application of Linear Matrix Inequalities for
10
-1 0
10 10
1
10
2
10
3
Load Frequency Control with Communication Delays,” IEEE Trans. on
Frequency (rad/sec)
Power Syst., vol. 19, NO. 3, Aug. 2004.
Fig. 7. Frequency response of GO ( s ) with different M area1 [9] D. Rerkpreedapong, A. Hasanovic, and A. Feliachi, “Robust Load
Frequency Control Using Genetic Algorithms and Linear Matrix
Inequalities,” IEEE Trans. on Power Syst., vol. 18, NO. 2, May 2003.
TABLE III [10] Z. Gao, Y. Huang, and J. Han, “An Alternative Paradigm for Control
System Design,” in Proc.s of IEEE conference on Decision and Control,
STABILITY MARGINS WITH DIFFERENT M area1
vol. 5, NO. 4-7, pp. 4578-4585, Dec. 2001.
[11] Z. Gao, “Scaling and Parameterization Based Controller Tuning,” in
Gain Margin (dB) Phase Margin (dB) Proc. of the 2003 American Control Conference, vol. 6, NO. 4-6, pp.
4989-4996, June 2003.
M area1 = 0.4867 11.2 77.4 [12] W. Zhou and Z. Gao, “An Active Disturbance Rejection Approach to
Tension and Velocity Regulations in Web Processing Lines,” IEEE
′′ 1 = M area1
M area 2= 0.2434 5.14 53.8
Trans. on Power Syst., vol. 19, NO. 3, Aug. 2004.
′ 1 = 2 M area1 = 0.9734
M area 17.3 40.2 [13] G. Tian and Z. Gao, “Frequency Response Analysis of Active
Disturbance Rejection Based Control System,” in Proc. of 16th IEEE Int.
Conf. on Ctrl. Apps. Part of IEEE Multi-conf. on Syst. and Ctrl.,
APPENDIX Singapore, 1–3, Oct. 2007.
[14] L. Dong, D. Avanesian, “Drive-Mode Control for Vibrational MEMS
The parameter values of the power system from [9] are Gyroscopes,” to appear in IEEE Trans. on Industrial Electronics.
given as follows. The ramp rate factor [15] Z. Gao, “Active Disturbance Rejection Control: A Paradigm Shift in
Feedback Control System Design,” in Proc. of American Control
Conference, Minneapolis, MN, June 2006.

2778

View publication stats