Electrical Power and Energy Systems 43 (2012) 996–1005

Contents lists available at SciVerse ScienceDirect

Electrical Power and Energy Systems

journal homepage: www.elsevier.com/locate/ijepes

Robust analysis of decentralized load frequency control for multi-area

power systems
Wen Tan ⇑, Hong Zhou
State Key Laboratory of Alternate Electrical Power System with Renewable Energy Sources, North China Electric Power University, Zhuxinzhuang, Dewai, Beijing 102206, PR China

a r t i c l e i n f o a b s t r a c t

Article history: Robust analysis for decentralized load frequency control (LFC) for multi-area power systems is studied in
Received 30 June 2010 this paper. It is observed that such an analysis can be decomposed into two steps considering the inher-
Received in revised form 26 May 2012 ent structure of a multi-area power system: robustness analysis against the parametric variations in
Accepted 29 May 2012
local-area power systems and robustness analysis against the structure and/or magnitude variations in
Available online 15 July 2012
the tie-line power flow network. A detailed structured singular value method is proposed for local-area
robustness analysis, and an eigenvalue method is derived for tie-line robustness analysis. The proposed
Keywords:
method is then applied to a four-area power system and the results show that the method is convenient
Load frequency control
Robustness
and useful for decentralized load frequency control of multi-area power systems.
Decentralized control Ó 2012 Elsevier Ltd. All rights reserved.
Multi-area power systems
Structured singular value

1. Introduction  Tertiary control is an economic dispatch that is used to drive

In power systems, changes in the load affect the frequency and
bus voltages in the systems. For small changes in the load the fre-
quency deviation problem can be separated or decoupled from the
voltage deviation. The problem of controlling the real power out-
put of generating units in response to changes in system frequency
and tie-line power interchange within specified limits, is known as
load frequency control (LFC) [1]. It is generally regarded as a part of
automatic generation control (AGC) and is very important in the
operation of power systems.
Usually AGC is organized in three levels:
 Primary control is performed by the speed governors of the
generating units, which provide immediate (automatic)
action to sudden change of load (or change of frequency).
With primary control, a variation in system frequency
greater than the dead band of the speed governor will result
in a change in unit power generation. Transients of primary
control are in the time-scale of seconds.
 Secondary control restores frequency to its nominal value
and maintains the power interchange among areas by
adjusting the output of selected generators. Transients of
secondary control are in the order of minutes.
⇑ Corresponding author. Tel.: +86 10 61772107; fax: +86 10 61772260.
E-mail address: wtan@ieee.org (W. Tan).
the system as economically as possible and restore security
levels if necessary. Tertiary control is usually performed every
5 min.
The speed governor on each generating unit provides the pri-
mary speed control function, and all generating units contribute
to the overall change in generation, irrespective of the location of
the load change, using their speed governing. However, primary
control action is usually not sufficient to restore the system fre-
quency, especially in an interconnected power system so the sec-
ondary control loop is required to adjust the load reference set
point through the speed-changer motor. Secondary control is com-
monly referred to as load–frequency control. See [2,3] for a com-
plete review of recent advance in LFC.
LFC becomes more significant today with the increasing size
and complexity of interconnected power systems. Multivariable
control techniques can be used to design centralized load fre-
quency controllers, however, due to the inherent structure of
large-scale power systems, decentralized load frequency control
is more appealing for its simplicity in design and implementation,
see, for example, [4–10] for different control methods to design
decentralized LFC.
Most of the methods uses decentralized controllers. Though
simulation results show that the decentralized control may achieve
good performance, no theoretical results have been found that can
be used to analyze the robustness of the decentralized control for
LFC. In this paper, robust analysis for decentralized load frequency
control of multi-area power systems will be investigated. The

0142-0615/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ijepes.2012.05.063
 W. Tan, H. Zhou / Electrical Power and Energy Systems 43 (2012) 996–1005 997

Nomenclature

Dfi frequency deviation of Area #i (Hz) DPti generator output in Area #i (p.u. MW)
Bi frequency bias setting of Area #i (p.u. MW/Hz) Ggi(s) transfer function of the governor in Area #i
Tij synchronizing coefficients between Area #i and Area #j Gti(s) transfer function of the turbine in Area #i
(p.u. MW/Hz) Gpi(s) transfer function of the rotor inertia and load in Area #i
DPdi load disturbance in Area #i (p.u. MW) Ri speed regulation of Area #i (Hz/p.u. MW)
DPtiei tie-line power between Area #i and other areas (p.u.
MW)

analysis is based on recent method ([11]) proposed to analyze the ACEi ¼ DPtiei þ Bi Dfi ð1Þ
stability of a multi-area power system under a decentralized LFC.
and the feedback control for Area #i takes the form
The method takes the inherent structure of the multi-area power
system into consideration, and makes it possible to check the im- ui ¼ K i ðsÞACEi ð2Þ
pacts of parametric variations in the local-area power systems
where Ki(s) is the local LFC controller.
and the structure and/or magnitude variations in the tie-line
According to [12,13], a decentralized controller can be designed
power flow network sequentially. Robustness analysis against the
assuming that there are no tie-line power flows, i.e.,
two groups of uncertainties will then be discussed. Finally, the pro-
DPtiei = 0(i = 1, . . . , n). In this case the local feedback control will be
posed method will be applied to analyze the decentralized LFC for a
four-area power system. It is shown that the method is convenient ui ¼ K i ðsÞBi Dfi ð3Þ
and useful for decentralized load frequency control of multi-area Denote the transfer functions of the governor, the turbine, and
power systems. the rotor inertia and load for Area #i by Ggi(s), Gti(s), and Gpi(s),
respectively, then the transfer function from ui to Dfi can be easily
2. Stability analysis of decentralized LFC found as

Consider the load frequency control problem for a multi-area

power system shown in Fig. 1. Each control area has the structure
shown in Fig. 2. The control area is a simplified power system con-
sisting of many generating units. According to [1], the collective
performance of all generators in the system is the interesting part
in the analysis of LFC. The intermachine oscillations and transmis-
sion system are not considered. It is assumed that the response of
all generators to changes in system load are coherent and can be
represented by an equivalent generator, which has an inertia con-
stant and a damping constant equal to the sum of the inertia con-
stants and damping constants of all the generating units. The
model is simple but captures the essential dynamics of a power
system and has been widely used for LFC design purpose.
The load frequency control problem for a multi-area power sys-
tem requires that not only the frequency deviation of each area
must return to its nominal value but also the tie-line power flows
must return to their scheduled values. So a composite variable, the
area control error (ACE), is used as the feedback variable to ensure
the two objectives. For Area #i, the area control error is defined as
Gpi Gti Ggi
Gi ðsÞ ¼ ð4Þ
1 þ Gpi Gti Ggi =Ri
So it is clear that to design a decentralized load frequency con-
troller, one just needs to design controllers for the following trans-
fer function for Area #i.
Pi ðsÞ ¼ Gi ðsÞBi ð5Þ
It is shown that the decentralized load frequency control of
multi-area power systems requires designing local controllers for
the model (5). Each local controller can be designed independently.
However, since tie-line power flows among areas are ignored in the
local load frequency control design, we need to check the stability
of the whole system to ensure the designed decentralized control-
ler works after the local controllers have been tuned. Many multi-
variable stability theories can be applied to check the stability of a
multi-area power system under a decentralized LFC controller.
However, the multi-area power system has its specific structure,
so a simple method can be derived for the stability analysis. The
following result is from [11].

Theorem 1. Given an n-area power system shown in Fig. 1 and

assume that each area has the structure as shown in Fig. 2. Then the
whole power system is stable if and only if the following transfer
function is stable.
hðsÞ :¼ detðI þ NðsÞT=sÞ ð6Þ

Tie-line where the ‘tie-line network matrix’ T is a constant matrix defined by

2X n 3
6 T 1j T 12  T 1n 7
6 j–1 7
6 7
6 X
n 7
6 7
6 T 21 T 1j    T 2n 7
6 7
T :¼ 6 j–2 7 ð7Þ
6 7
6 . .. .. .. 7
6 .. . . . 7
6 7
6 X
n 7
4 5
T n1 T n2  T 1j
j–n
Fig. 1. Simplified diagram of a multi-area interconnected power system.
998 W. Tan, H. Zhou / Electrical Power and Energy Systems 43 (2012) 996–1005

3.1. Local-area robustness analysis

Structured singular value is an efficient tool to analyze the

robustness of a control system subject to model uncertainties. For
LFC control, each part in Fig. 2 takes the following form (for simplic-
ity we will omit subscript i below when there is no ambiguity):

 Governor with dynamics:

1
Gg ðsÞ ¼ ð10Þ
T Gs þ 1
 Turbine with dynamics:
1 cT r s þ 1
Fig. 2. Block diagram of control area i. Gt ðsÞ ¼ ð11Þ
TT s þ 1 Trs þ 1
and the ‘local transfer matrix’ N(s) is a diagonal transfer matrix de-  Rotor inertia and load with dynamics:
fined by
KP
2 3 Gp ðsÞ ¼ ð12Þ
N1 ðsÞ 0  0 TPs þ 1
6 0 N2 ðsÞ    0 7
6 7
NðsÞ :¼ 6
6 .. .. .. ..
7
7 ð8Þ
4 5 For local-area robustness analysis, we need to test whether the
. . . .
local control system is still stable when there are uncertainties in
0 0    Nn ðsÞ the parameters of the models.
with Ni(s) given by The following result is useful for robustness analysis.

Gpi þ Gpi Gti Ggi K i Theorem 2. Given a plant with model

Ni ðsÞ :¼ ð9Þ
1 þ Gpi Gti Ggi =Ri þ Gpi Gti Ggi K i Bi
yðsÞ
From Theorem 1, the stability of a multi-area power system un- ¼ PðsÞ ¼ CðsI  AÞ1 B ð13Þ
uðsÞ
der decentralized LFC can be depicted as in Fig. 3. It is observed
that the local transfer matrix N(s) and the tie-line power flow net- and a stabilizing feedback controller
work T are separated from each other in h(s) due to the inherent uðsÞ ¼ KðsÞyðsÞ ð14Þ
structure of the multi-area power system, so it is easy to check Suppose there are uncertainties in the state-space data and the
the effects of local controllers (N(s)) and the tie-line interconnec- perturbed system can be described as
tion network (T) on the stability of the whole power system.
PD ðsÞ ¼ CðsI  ðA þ DA ÞÞ1 ðB þ DB Þ ð15Þ

3. Robustness analysis of decentralized LFC where the perturbations take the form
DA ¼ E DF a ; DB ¼ E DF b ð16Þ
Due to changes of operation points, the parameters of the power
systems are not fixed, so it is crucial for a designed LFC control to Here D is a set of unknown but bounded parameters, E, Fa and Fb are
achieve robust performance. According to Fig. 3, uncertainties for known constant matrices with compatible dimensions. Then the
multi-area power systems can be divided into two groups: uncer- closed-loop system remains stable for all D such that kDk1 6 1 if
tainties in local-area power systems and uncertainties in the tie- and only if
line power flow network. A detailed robustness analysis can be lD ðMÞ < 1 ð17Þ
done using structured singular value (SSV) method, however, since where lD(M) is the structured singular value with respect to struc-
there are correlations between the local-area power systems and ture D defined in [14] and
the tie-line network, it is very complicated to perform such a de-
tailed analysis. Instead, we can analyze system robustness for the M :¼ P11  P12 KðI þ P22 KÞ1 P 21 ð18Þ
two groups of uncertainties sequentially, i.e., we will first make with
sure that the designed local LFC is robust for local-area power sys-
tems, and then make sure the whole controller is robust for tie-line P11 ¼ F a ðsI  AÞ1 E; P 12 ¼ F a ðsI  AÞ1 B þ F b
ð19Þ
operation. P21 ¼ CðsI  AÞ1 E; P 22 ¼ CðsI  AÞ1 B

Proof. It can be verified that

PD ðsÞ ¼ CðsI  ðA þ DA ÞÞ1 ðB þ DB Þ

Fig. 3. Stability analysis of a multi-area power system under decentralized LFC.
¼ CðsI  A  EDF a Þ1 ðB þ EDF b Þ
¼ P22 þ P21 DðI  P 11 DÞ1 P 12 ð20Þ
So the feedback control system with open-loop model PD(s) (15)
and the feedback control (14) can be represented as Fig. 4a. To use
SSV technique to analyze robust stability, we need to transform the
system into Fig. 4b, and it can be verified that the transfer function
M in Fig. 4b is just (18). Theorem 2 follows by the well-known
small-l theorem [14] in robust control theory.
 W. Tan, H. Zhou / Electrical Power and Energy Systems 43 (2012) 996–1005 999

The procedure is a generalization of that reported in [15] for a

power system with a non-reheated turbine. For a power system
with a hydro-turbine, similar procedure can be applied. h

3.2. Tie-line robustness analysis

With the robustness analysis proposed in the previous subsec-

tion, local LFC can be checked to make sure that the stability of lo-
cal-area power systems under parameters variations is guaranteed.
What is left is then to check if the designed decentralized LFC is ro-
bust against tie-line operation.
As depicted in Fig. 3, various structures and magnitudes of the
(a) (b) tie-line network on the stability of the whole system can be
checked using the eigenvalues of h(s). For example, consider the
Fig. 4. Framework for robust analysis.
two-area case, we have
 
T 12 T 12
For a power system with a reheated turbine, the system can be T¼ ð27Þ
T 21 T 21
expressed into the state-space form (13) with
2 3 Suppose T12 = T21 = c, then
 T1P KP
0 0 3 2    
TP 0 c c 1
6 7
6 0  T1T c 1 7 607 T¼ ¼c ½ 1 1  ð28Þ
6 TT TT 7 6 7 c c 1
A¼6 1 7; B ¼ 6 1 7 ð21Þ
6 0  T1G 0 7 4T 5
4 RT G 5 G
Since det (I + AB) = det (I + BA) for any compatible matrices A
0 0 1c 1
 Tr 0 and B, we have
Tr

and C ¼ ½ 1 0 0 0 . e c=s
hðsÞ ¼ detðI þ NT=sÞ ¼ 1 þ NðsÞ ð29Þ
Suppose the uncertain parameters are
where
m1 :¼ 1=T P ; m2 :¼ K P =T P ; m3 :¼ 1=T T ;   
ð22Þ e N1 ðsÞ 0 1
m4 :¼ 1=RT G ; m5 :¼ 1=T G ; m6 :¼ 1=T r : NðsÞ ¼ ½ 1 1  ¼ N1 ðsÞ þ N2 ðsÞ ð30Þ
0 N2 ðsÞ 1
and that each parameter is varied within a bound, i.e.
So the maximum c such that the power system remains stable
mi ¼ mi0 ð1 þ wi di Þ ð23Þ e
is the largest gain which destabilizes NðsÞ=s. It can be checked by
e
using the root locus of NðsÞ=s.
where mi0 is the nominal value of the ith parameter, wi is the rela-
For interconnected power systems with more than two areas,
tive variation ratio of mi, and di is an uncertain parameter normal-
the tie-lie power flow network ‘T’ is much more complicated. For
ized as [1, 1].
instance, for a three-area power system several cases might be ta-
Define the set of uncertain parameters as
ken into consideration: operations with three tie-lines, with two
2 3
d1 tie-lines, and with one tie-line (the case without tie-line operation
6 d2 7 is trivial). Each case has its tie-line network matrix, so for a three-
6 7
6 7 area power system seven different matrices need to be considered
6 d3 7
D :¼ 6
6
7
7 ð24Þ for T. The number of combination increases with the number of
6 d4 7
6 7 areas. Moreover, the magnitude of each tie-line power flow is dif-
4 d5 5 ferent. So it is complicated to determine if the decentralized LFC is
d6 robust enough for structural uncertainties in tie-line operations.
However, once a the tie-lien topology is determined, the stability
It is noted that D has been normalized such that kDk1 < 1.
of the whole power system can be checked with Theorem 1. The
With the parameters of the local area power system varied as
complexity lies in enumerating the combination of tie-line opera-
(23), then the variations on state-space matrices A and B can be
tions, not the method in verifying the stability.
described as in (16) with
In summary, the stability of the whole power system relies on
2 3
m10 w1 m20 w2 0 0 0 0 the structure of the tie-line network, the magnitude of each tie-line
6 0 0 m30 w3 0 0 7 flow, and the local load frequency controllers. Checking the roots of
6 7
E¼6 7 ð25Þ h(s) under different tie-line network structure and magnitude can
4 0 0 0 m40 w4 m50 w5 0 5
verify the performance of the designed decentralized LFC.
0 0 0 0 0 m60 w6
2 3 2 3 4. Control area with different generation units
1 0 0 0 0
6 0 1 0 0 7 607
6 7 6 7 The model in Fig. 2 assumes that the individual control loops
6 7 6 7
6 0 1 c 1 7 607 and turbine-generators have the same regulation parameters and
F a :¼ 6
6 1 0
7; F b :¼ 6 7 ð26Þ
6 0 0 7
7
607
6 7 response characteristics. In practical power utilities, there are
6 7 6 7 many generators in each control area with different turbine-gover-
4 0 0 1 0 5 415
nor parameters and generation types. Furthermore, generators
0 0 1c 1 0
may or may not participate in the LFC task and participation rates
By Theorem 2, we can check the robust stability of a local-area are not the same for all participant generators [16]. In order to con-
power system using (17). sider the variety of generation dynamics and their sharing rate in
1000 W. Tan, H. Zhou / Electrical Power and Energy Systems 43 (2012) 996–1005

Fig. 5. Block diagram of control area i with different generation units and participating factors.

Bode Diagram
50

0
Magnitude (dB)

−50

−100

−150
0
Phase (deg)

−90

−180

−270
−1 0 1 2 3
10 10 10 10 10
Frequency (rad/sec)

Fig. 6. Bode diagrams of the dynamics of a multi-unit control area and its equivalence (solid: multi-unit area; dashed: lumped area).

the LFC, the dynamic model of Area #i in Fig. 2 can be modified to

that shown in Fig. 5, where m is the number of generation units in
Area #i,aij represents the participating factor; Rij is the speed regu-
lation; and Ggij(s), Gtij(s) are the transfer functions of governor and
Area #3 Area #2
turbine of generation unit j, respectively. The dynamics of rotor
inertia and load in a control area is lumped as in Fig. 2 and its
transfer function is denoted by Gpi(s).
It is easy to show that the transfer function for decentralized
LFC control (assuming Ptiei = 0) is still (5), with the transfer func-
tion from ui to Dfi in Fig. 5 becomes
Pm Area #1 Area #4
Gpi a
j¼1 ij Gtij Ggij
Gi ðsÞ ¼ P ð31Þ
1þ Gpi mj¼1 Gtij Ggij =Rij

Fig. 7. Simplified diagram of a four-area power system.

It is also easy to verify that the stability analysis for the decen-
tralized LFC for a multi-area power system in Fig. 1 with each con-
trol area having structure in Fig. 5 is similar to that with structure P
Gpi þ Gpi m j¼1 aij Gtij Ggij K i
in Fig. 2. Theorem 1 still holds, only the elements in the ‘local Ni ðsÞ ¼ Pm P ð32Þ
transfer matrix’ N(s) should be changed to 1 þ Gpi j¼1 Gtij Ggij =Rij þ Gpi m j¼1 Gtij Ggij K i Bi
 W. Tan, H. Zhou / Electrical Power and Energy Systems 43 (2012) 996–1005 1001

−3
x 10
0.01 5

0.005

0 0

Δ f1 (Hz)

Δ f2 (Hz)
−0.005

−0.01 −5

−0.015

−0.02 −10
0 20 40 60 80 100 0 20 40 60 80 100
Time (sec.) Time (sec.)
−3
x 10
5 0.01

0 0.005
Δ f3 (Hz)

Δ f4 (Hz)
−5 0

−10 −0.005

−15 −0.01

−20 −0.015
0 20 40 60 80 100 0 20 40 60 80 100
Time (sec.) Time (sec.)

−3 −3
x 10 x 10
4 3

2
2
(p.u.MW)
(p.u.MW)

0
1
−2
tie−2
tie−1

0
−4
ΔP
ΔP

−1
−6

−8 −2
0 20 40 60 80 100 0 20 40 60 80 100
Time (sec.) Time (sec.)

−3 −3
x 10 x 10
4 3

2 2
Δ Ptie−3 (p.u.MW)

(p.u.MW)

0 1
tie−4

−2 0
ΔP

−4 −1

−6 −2
0 20 40 60 80 100 0 20 40 60 80 100
Time (sec.) Time (sec.)

Fig. 8. Responses of the four-area power system (solid: Tij = 0.0707; dashed: Tij = 0.444; dotted: Tij = 0.73).

If the generation units in a control area shown in Fig. 5 are of the and
same type, then the area can be simplified to a control area in Fig. 2. R1 ¼ 3; R2 ¼ 3; R3 ¼ 3:3 ð34Þ
For example, suppose a control area has three generation units with
non-reheated turbines, and the parameters of the units are The rated powers of the three units are
P1 ¼ 300 MW; P2 ¼ 200 MW; P3 ¼ 250 MW ð35Þ
1 1 1
Gg1 ¼ ; Gt1 ¼ ; Gp1 ¼ and suppose the participating factors for the three units are
0:08s þ 1 0:4s þ 1 0:1667s þ 0:015
1 1 1 a1 ¼ 0:4; a2 ¼ 0:4; a3 ¼ 0:2 ð36Þ
Gg2 ¼ ; Gt2 ¼ ; Gp2 ¼ ð33Þ
0:06s þ 1 0:36s þ 1 0:12s þ 0:014 The first step in simplifying is to find an equivalent generator
1 1 1 for all the units in the area. The inertia and damping of the equiv-
Gg3 ¼ ; Gt3 ¼ ; Gp3 ¼
0:07s þ 1 0:42s þ 1 0:2s þ 0:015 alent generator is the weighted sum of all the inertia and damping
1002 W. Tan, H. Zhou / Electrical Power and Energy Systems 43 (2012) 996–1005

−3 −3
x 10 x 10
5 4

2
0
0

Δ f (Hz)

Δ f (Hz)
−5 −2

2
−4
−10
−6

−15 −8
0 20 40 60 80 100 0 20 40 60 80 100
Time (sec.) Time (sec.)

−3
x 10
5 0.01

0.005
0
0

Δ f (Hz)
Δ f (Hz)

−5 −0.005

4
3

−0.01
−10
−0.015

−15 −0.02
0 20 40 60 80 100 0 20 40 60 80 100
Time (sec.) Time (sec.)

−3 −3
x 10 x 10
4 2.5

2
2
(p.u.MW)

(p.u.MW)

1.5
0 1
tie−1

tie−2

−2 0.5
ΔP

ΔP

0
−4
−0.5

−6 −1
0 20 40 60 80 100 0 20 40 60 80 100
Time (sec.) Time (sec.)

−3 −3
x 10 x 10
4 3

2 2
(p.u.MW)
Δ Ptie−3 (p.u.MW)

1
0
0
tie−4

−2
−1
ΔP

−4
−2

−6 −3
0 20 40 60 80 100 0 20 40 60 80 100
Time (sec.) Time (sec.)

Fig. 9. Responses of the four-area power system under structure variation in the tie-line network (solid: current; dashed: Case 1; dashdotted: Case 2).

of the generators in the area, where the weights accounts for the 1 1
Gg ðsÞ ¼ ; Gt ðsÞ ¼ ð38Þ
portion of each unit in the total rated power. So the dynamics of 0:07133s þ 1 0:396s þ 1
the equivalent generator is

1
Gp ðsÞ ¼ ð37Þ and the speed regulation is lumped as R = 3.0938. The Bode diagram
0:1653s þ 0:01473 of the original local-area dynamics and the reduced local-area
Now the dynamics of the turbines and governors are lumped dynamics are shown in Fig. 6. They are very close so there is no loss
using the same idea as for the equivalent generator. They are found of generality to use the control area shown in Fig. 2 in LFC analysis
as and design if all the units are of the same type.
 W. Tan, H. Zhou / Electrical Power and Energy Systems 43 (2012) 996–1005 1003

A better approximation can be found by approximating the lo- decentralized LFC cannot guarantee the stability of the multi-area
cal-area dynamics (31) with a reduced-order transfer function in power system.
the frequency domain, especially when the number of generation To test the performance of the designed decentralized LFC
units in the control area is large. Moreover, even if the generation against tie-line network interaction, suppose a step load
units are not of the same type, the lumped area in Fig. 2 is still DPd1 = 0.01 is applied to Area #1 at t = 10 and a step load
applicable in analyzing the effects of the tie-line flow, with Gti(s)G- DPd3 = 0.01 is applied to Area #3 at t = 50. The system responses
gi(s) replaced by a single transfer function which is approximated are shown in Fig. 8 for the nominal tie-line network and the cases
in the frequency domain from the local-area dynamics (31). that Tij = 0.444 (increased by a factor of 2p) and Tij = 0.73 (near sta-
bility margin). From the simulation results, it is observed that the
5. Illustrative example decentralized LFC works very well for weak tie-line network, and
performance degrades for strong tie-line network, which com-
Consider a four-area power system discussed in [17,13]. A sim- pletely agrees with the analysis discussed above.
ple diagram of the four-area power system is shown in Fig. 7,
where Areas #1,#2,#3 are interconnected while Area #4 is only 5.2. Robustness against tie-line structure change
connected with Area #1.
It is assumed that Areas #1, #2, #3 are identical systems with Despite the possible magnitude change of the tie-line power
reheated turbines. The parameters are: flow network, its structure may also be changed. We consider
two circumstances under the current network structure with a
1 ci T ri s þ 1 1 K Pi
Ggi Gti Gpi ¼ ð39Þ strong interaction (Tij = 0.444):
T Gi s þ 1 T ri s þ 1 T Ti s þ 1 T Pi s þ 1
with Case 1. Area #4 is connected to Area #2 (T24 = T42 = 0.444).
Case 2. Area #2 is disconnected from Area #1 (T12 = T21 = 0).
T Gi ¼ 0:2; T Ti ¼ 0:3; T Pi ¼ 20; K Pi ¼ 120;
T ri ¼ 20; ci ¼ 0:333 ði ¼ 1; 2; 3Þ ð40Þ The system responses under the same load disturbances as be-
and Ri = 2.4. Area #4 is a power system with hydro turbine, with the fore are shown in Fig. 9. It is observed that the designed decentral-
open-loop plant model given by: ized LFC works well in Case 2, but performance degrades for Case 1.
In fact, the minimal damping ratio of h(s) is 0.0054 for Case 1 and
1 0:513s þ 1 1  s 80 0.1364 for Case 2. Compared with the nominal damping 0.0947, it
Gg Gt Gp ¼ ð41Þ
48:7s þ 1 10s þ 1 1=2s þ 1 13s þ 1 is observed that disconnecting one area from an existing power
network may increase the damping of the system, while connect-
and R = 2.4. The synchronizing constants are
ing more areas into an existing power network may decrease the
T 12 ¼ T 21 ¼ T 13 ¼ T 31 ¼ T 23 ¼ T 32 ¼ 0:0707 ð42Þ damping of the system. The large oscillation in Case 1 indicates
T 14 ¼ T 41 ¼ 0:0707; T 24 ¼ T 42 ¼ T 34 ¼ T 43 ¼ 0 ð43Þ that the decentralized LFC can only provide a very weak damping
to the multi-area power system, so for Case 1 it needs to be re-
and the frequency bias constants Bi(i = 1, 2, 3, 4) are 0.425 for each tuned to achieve good damping.
area.
In [13], the following PID controllers are tuned for Areas #1, #2
5.3. Robustness against local-area uncertainties
and #3,
1:5847 Another source of uncertainties in a multi-area power system
K i ðsÞ ¼ 2:4457 þ þ 1:5261s ði ¼ 1; 2; 3Þ ð44Þ
s lies in the parameters of the local area. We can utilize Theorem 2
to analyze the effects of parametric variations on the stability of
and the following modified PID controller is tuned for Area #4.
  the system. Assume that there are 40% variations in all the param-
0:0216 3:8787s þ 1 eters in Areas #1, #2, #3, and 18% in Area #4, the SSV plots (lD(M))
K 4 ðsÞ ¼ 0:1456 þ þ 18:7715s ð45Þ
s 0:03919s þ 1
We would like to test the robust performance of this decentral- 1
ized LFC under different situations.
0.9

5.1. Robustness against tie-line network interaction 0.8

Since the decentralized LFC is tuned by assuming there is no tie- 0.7

line connection, so first we need to check if the decentralized LFC 0.6

works when the tie-line power flow is in operation. By Theorem
μΔ(M)

1, the minimum damping ratio for h(s) is 0.4219, so the nominal 0.5

performance under current tie-line operation is very good. In fact,

0.4
we note that Tij is very small, so the interactions between each area
is weak, and the designed decentralized LFC works well for 0.3
weak tie-line network.
0.2
To check if the decentralized LFC works for strong tie-line net-
work, we increase the magnitude of the synchronizing coefficients 0.1
Tij by a factor of 2p(Tij = 0.444), then the minimum damping ratio
0
of h(s) becomes 0.0947. This means that as the tie-line network be- −2 −1 0 1 2 3
10 10 10 10 10 10
comes strong the performance of a decentralized LFC will degrade
Freq (rad/s)
due to the strong interaction of subsystems. In fact, if Tij is in-
creased to 0.77, then the minimum damping ratio of h(s) becomes Fig. 10. Robustness measures for local areas (solid: Areas #1, #2, #3; dashed: Area
negative, which means that under such strong connection, the #4).
1004 W. Tan, H. Zhou / Electrical Power and Energy Systems 43 (2012) 996–1005

−3
x 10
0.01 5

0.005
0

Δ f2 (Hz)
Δ f (Hz)
0
−5

1
−0.005

−10
−0.01

−0.015 −15
0 20 40 60 80 100 0 20 40 60 80 100
Time (sec.) Time (sec.)

−3
x 10
5 0.01

0 0.005

Δ f4 (Hz)
Δ f3 (Hz)

−5 0

−10 −0.005

−15 −0.01

−20 −0.015
0 20 40 60 80 100 0 20 40 60 80 100
Time (sec.) Time (sec.)

−3 −3
x 10 x 10
4 3

2
2
(p.u.MW)

Δ Ptie−2 (p.u.MW)

0
1
−2
tie−1

0
−4
ΔP

−6 −1

−8 −2
0 20 40 60 80 100 0 20 40 60 80 100
Time (sec.) Time (sec.)

−3
x 10
−3 x 10
4 3

2 2
Δ Ptie−3 (p.u.MW)

Δ Ptie−4 (p.u.MW)

0 1

−2 0

−4 −1

−6 −2
0 20 40 60 80 100 0 20 40 60 80 100
Time (sec.) Time (sec.)

Fig. 11. Responses of the four-area power system (solid: nominal; dashed: +40% parametric variations in Areas #1, #2, #3, and +18% in Area #4; dashdotted: 40% parametric
variations in Areas #1, #2, #3, and 18% in Area #4).

of each area are shown in Fig. 10. From the plot, if there are 40% To test the robust performance of the designed decentralized
parameter variations in Areas #1, #2, #3, the maximums of the LFC against uncertainties in local-area power systems, suppose a
SSV are just less than 1, so robustness is guaranteed by Theorem step load D Pd1 = 0.01 is applied to Area #1 at t = 10 and a step
2. It is the same with Area #4. If the variations are more than load DPd3 = 0.01 is applied to Area #3 at t = 50. The system
40% in Areas #1, #2, #3 (or 18% in Area #4), the SSV will be larger responses are shown in Fig. 11 when there are 40% simultaneous
than 1, and then robustness is not guaranteed. Since the open-loop variation in the models of Areas #1,#2 and #3 and 18% simulta-
model (5) of Area #4 is unstable, it is not easy to achieve similar neous variation in the model of Area #4, and the magnitude of
robustness in Area #4 without performance degradation as in the tie-line network is 0.444. They validate the statements
other areas, where the open-loop models are stable. above.
 W. Tan, H. Zhou / Electrical Power and Energy Systems 43 (2012) 996–1005
6. Conclusion
Robust analysis of decentralized load frequency controller for
multi-area power systems was discussed in this paper. Robust
analysis for the whole system was decomposed into two steps:
robustness analysis for local-area power systems and that for tie-
line operation. The method was shown to be useful for decentral-
ized LFC tuning. Further research on efficient robustness tests
against structure variations of tie-line power network for large
power systems is under investigation.
Acknowledgment
This work was supported by the National Natural Science Foun-
dation of China under Grant 61174096, the Natural Science Foun-
dation of Beijing under Grant 4122075, and the Natural Science
Foundation of Hebei under Grant F2011502069.
References
[1] Kundur P. Power system stability and control. McGraw-Hill; 1994.
[2] Ibrabeem PK, Kothari DP. Recent philosophies of automatic generation control
strategies in power systems. IEEE Trans Power Syst 2005;20(1):346–57.
[3] Shayeghi H, Shayanfar HA, Jalili A. Load frequency control strategies: a state-
of-the-art survey for the researcher. Energy Convers Manage 2009;50
(2):344–53.
[4] Lim KY, Wang Y, Zhou R. Robust decentralized load–frequency control of
multi-area power systems. IEE Proc Part C 1996;143:377–86.
1005
[5] Yang TC, Ding ZT, Yu H. Decentralized power system load–frequency control
beyond the limit of diagonal dominance. Int J Eelctr Power Energy Syst
2002;24(3):173–84.
[6] Bevrani H, Mitani Y, Tsuji K. Sequential design of decentralized load frequency
controllers using l synthesis and analysis. Energy Convers Manage
2004;45:865–81.
[7] Alrifai MT, Hassan MF, Zribi M. Decentralized load frequency controller for a
multi-area interconnected power system. Int J Electr Power Energy Syst
2011;33(2):198–209.
[8] Farhangi R, Boroushaki M, Hosseini SH. Load–frequency control of
interconnected power system using emotional learning-based intelligent
controller. Int J Electr Power Energy Syst 2012;36(1):76–83.
[9] Yazdizadeh A, Ramezani MH, Hamedrahmat E. Decentralized load frequency
control using a new robust optimal MISO PID controller. Int J Electr Power
Energy Syst 2012;35(1):57–65.
[10] Khodabakhshian A, Pour ME, Hooshmand R. Design of a robust load frequency
control using sequential quadratic programming technique. Int J Electr Power
Energy Syst 2012;40:1–8.
[11] Tan W. Decentralized load frequency controller analysis and tuning for multi-
area power systems. Energy Convers Manage 2011;52(5):2015–23.
[12] Tan W. Tuning of PID load frequency controller for power systems. Energy
Convers Manage 2009;50(6):1465–72.
[13] Tan W. Unified tuning of PID load frequency controller for power systems via
IMC. IEEE Trans Power Syst 2010;25(1):341–50.
[14] Doyle JC. Analysis of feedback systems with structured uncertainties. IEE Proc
Control Theory Appl 1982;129:242–50.
[15] Tan W, Xu Z. Robust analysis and design of load frequency controller for power
systems. Electric Power Syst Res 2009;79(5):846–53.
[16] Bevrani H, Mitani Y, Tsuji K. Robust decentralized load–frequency control
using an iterative linear matrix inequalities algorithm. IEE Proc Gener Transm
Distrib 2004;151(3):347–54.
[17] Chang CS, Fu WH. Area load frequency control using fuzzy gain scheduling of
PI controllers. Electr Power Syst Res 1997;42(2):145–52.