Professional Documents
Culture Documents
Indexing terms: Load and voltage regulation, Power systems and plant, Modelling
Abstract: The paper presents a new method of designing decentralised load-frequency regulators for inter-
connected power systems. Within the framework of this method, the interconnected mutliarea power
system is decomposed into several subsystems, each of which is controlled separately by a decentralised
regulator. Each subsystem consists of one area and its external equivalent in a simplified form. A decentra-
lised control law for the study area is introduced by using a quadratic performance index. Feedback gains of
the decentralised regulator, which minimise the index, are determined by a Newton-Raphson iterative
algorithm. The proposed method is applied to an interconnected longitudinal 4-area system, and the effects
of the proposed regulator are examined by digital simulations and associated sensitivity analysis of the
system. Furthermore, a suitable means for preventing excessive control action is also considered involving a
significant system nonlinearity, i.e. some generation rate constraint.
1 Introduction
part and additional proportional and derivative control parts
The load-frequency control (LFC) problem has been one of with the area control error (ACE) used as the feedback signal.
the major subjects concerning power-system engineers, and is By including the conventional integral control part, the
becoming much more significant today in accordance with proposed regulator guarantees the zero steady-state error in
increasing size and complexity of interconnected power both frequency and tie-line power.
systems. In the analysis of large interconnected systems, the The proposed method is applied to an interconnected
growth in size and complexity of such systems presents a longitudinal 4-area system, and the effects of the regulator are
number of difficulties from the viewpoint of required com- examined under various conditions of the system. In order to
putation time and computer memory. make the analysis more realistic, the effects of the significant
Until now much research [13] has been carried out into system nonlinearities, for example the generation rate con-
the LFC problem, using modern control theory, but futher straint [7], are taken into account. Furthermore, a practical
consideration may be required because, formerly, oversim- means of preventing the excessive and unnecessary control
plified models have been used for the analysis, i.e. only action is also investigated involving the constraint.
nonreheated steam plants have been taken into account, and The results prove that the proposed method is very useful
only two areas have been considered. Also significant system for designing decentralised regulators of large interconnected
nonlinearities, i.e. generation rate constraints and limitations power systems.
of generation for different sorts of plants used for load-
frequency regulation, have been ignored, and nonmeasurable 2 Representation of mathematical model
quantities have been used for feedback signals. In this study, an interconnected longitudinal 4-area system,
Furthermore, it is also recognised that the implementation shown in Fig. 1, is considered to be a sample interconnected
of a centralised load-frequency control poses many difficulties, system.
for example difficulties in telemetering required data to the
centralised regulator, when the size and complexity of the
interconnected systems increase. In recent years, significant
efforts [4, 5] have been made to establish suitable decentra-
lised regulators for such large interconnected systems. tie 12 tie 23 tie 34
The purpose of this paper is to present a new method of
designing decentralised load-frequency regulators for large AP . = 7 AP,. ..
ei L tie ik
interconnected power systems. Within the framework of this
method, the interconnected multiarea power system is k
decomposed into several subsystems, each of which is con- Fig. 1 Sample 4-area system
trolled separately by a decentralised regulator. Each subsystem
contains one area and its external equivalent in a simplified 2.1 Model 1
form. The parameters of the equivalent are estimated by a well In model 1, all the areas consist of the nonreheated steam
known estimation technique [6]. plant with the same characteristics. The constants are indicated
Decentralised control law for a study area is introduced by in Table 1 [1]. An external equivalent in a simplified form is
a quadratic performance index for the mathematical model of selected for the external area model of the study area, as
the study area. Feedback gains of the regulator, which mini- shown in Fig. 2. The parameter values of the equivalent are
mise the index, are determined by a Newton-Raphson iterative estimated by a well known estimation technique [6].
algorithm based on the 1st- and 2nd-order sensitivities of the In a power system having steam plants, power generation
index with respect to the feedback gains. The regulator con- can change, only at a specified maximum rate [7]. The gener-
sidered in this paper is based on the conventional tie-line bias ation rate constraint for the study area is considered by adding
control (TBC), which is used by most utilities of the present a limiter to the governor, as shown in Fig. 2, i.e. the maximum
day. The regulator consists of a conventional integral control rate of valve opening or closing speed is restricted by the
limiter. Here, it must be noted that the generation rate con-
straints, and the control actions, for the external areas are
Paper 1715C (P9, Pll), first received 2nd March and in revised form ignored for simplicity throughout the present work.
7th October 1981
The author is with the Department of Electrical Engineering, Kumamoto The decentralised regulator for the study area is also
University, Kumamoto 860, Japan illustrated in Fig. 2. The regulator is constructed of a con-
IEEPROC, Vol. 129, Pt. C, No. 1, JANUARY 1982 0143- 7046/82/010017+ 07 $01.50/0 17
Table 1 : Constants for model 1 Table 2: Constants of study area for model 2
Tv( 0.1 s = governor time constant Khi = 0.5 = reheat coefficient, Kti 1 ACW = 0.5
Tfi 0.3 s = turbine time constant Tn ~ 6-0 s = reheat time constant
H* = 5.0 s, H, = 2Hf/f0 (p.u.MWs/Hz) = inertia constant a,- =1.8 Hz/p.u.MW = regulation rate of hydro plant
Di = 0.00833 p.u.MW/Hz = load-frequency constant Mi = 24.0 Hz/p.u.MW = hydro governor temporary droop
/?, = 2.4 Hz/p.u.MW = regulation characteristic
f0 = 60.0 Hz = nominal frequency 7"di = 4.0 s = time constant for temporary droop
Tjj = 0.5 p.u.MW/Hz s = synchronising coefficient between /th Tgi = 0.6 = hydro governor time constant
and/th area T~wi = 1.0 s = water starting time
fl,. = i//? { + Dt = 0.425 p.u.MW/Hz = frequency bias kl = 0.7, k2 = 0.3 = load allocation for steam and hydro plants
P^IPfj = 1.0 = capacity ratio between /th and/th areas Bi = MRi + Mai + Dt - 0.981 p.u.MW/Hz = frequency bias
Parameters of external equivalent
kx + k2 = 1.0, on considering the economic load dispatch and
H*s, He - 2He/f0 p.u.MWs/Hz = inertia constant
the generation capacity of each sort of plant.
r,-e p.u.MW/Hz s = synchronising coefficient between study area and
its external equivalent The constants for the study area are indicated in Table 2.
De p.u.MW/Hz = load-frequency constant involving regulation rate As shown in this Table, the 50% reheated steam plant is used
for the steam plant model, and the load allocation to each
plant is assumed as follows: 70% to the steam plant and 30%
to the hydro plant, by considering the recent trends in power
system, i.e. the decreasing percentages of the capacity in hydro
plants, and the increasing percentages of the capacity in base
loaded nuclear plants.
governor
-h,
s
B. At,
-h2
9?
j
-sh3 hydrogovernor
ATK + KA = -Q (10) In this work, area 2 in the sample system, shown in Fig. 1, is
considered to be a study area.
where Tr [ ] is the trace operator, and K denotes the positive
definite solution matrix of eqn. 10. The matrix Xo in eqn. 9 4.1 Example for model 1
is stated by
4.1.1 Parameter estimation of external equivalent: The para-
(11) meters of the external equivalent for area 2 have been esti-
mated through a well known estimation technique [6] by
x0 = initial value of state vector*.
using simulated values of the sample 4-area system for a pulsed
The necessary conditions for optimality are given by load change in area 2. The results are as follows:
H* = 2.7355 s, De = 0.8926 p.u.MW/Hz,
= 0 (12)
8ht Tie = 1.1677 p.u.MW/Hs
for all /. where the regulator action and the generation rate constraints
The regulator gain h can be modified recursively by a for all areas are neglected here.
Newton-Raphson algorithm along the direction of correction Figs. Aa and 4b indicate the responses following a pulsed
Ah, determined by the following equation until the optimality load change, and also following a step-load change, in area 2,
conditions in eqn. 12 are satisfied: respectively. As shown in these Figures, the external equiv-
Ah = -M~ld alent with these parameter values can be used as an external
(13)
area model for area 2, when various types of load change may
2 2
8J 8K be taken into account.
= Tr -
[ShjShj
4.1.2 Determination of regulator gains: The regulator gains for
= (*,/) element of Jacobian matrix M area 2 have been determined by minimising the index in eqn.
14. The results are indicated in Table 3. These gains have been
obtained through five to seven times iterations. In the calcu-
lation, the vector x0 in eqn. 11 is selected by considering a
step-load change AL in area 2 as follows:
= / th element of column vector d
* 0 = [- AL,0,0,- AL,- AL,0]T
3.2 Con trol purposes and performance indices
In order to determine the optimal gains for the regulator, the where the value of AL is specified to 1.0 for simplicity, because
following specifications for load-frequency regulation are the system considered is linear.
taken into account [7, 10]: The global optimality of these gains has been checked
(a) The zero steady-state error in both frequency and tieiine through the numerical experiments, by initiating the iterative
J 0.005
* - 0 01
-0.02 6 8 10
time, s
0 01 0.010r
i j
0 V 2 A 6 fi _ .19
K, -0 01 < -0.010 time, s
time, s
\ J
-0.C2 - 0.020 L
-0.03
Fig. 4 Responses obtained using external equivalent for area 2 0.015
Generation rate constraints and control action for all areas are not q = 1.0
considered 0.010
using detailed 4-area model >=
using external equivalent
a Pulsed load change APj2 = 0.01 p.u.MW for 0.5 s
0.005
7/*- q = 100
b Step load change APj2 = 0.01 p.u.MW
1/ 1 i i i
^ 15 H* = 2.7355 s
Tje= 1.1677 p.u.MW/Hz s so that the rate of the output from the integral control part
should not exceed the maximum rate of the governor output.
Figs. 9 and 10 illustrate the effects of the limiter added to
10 that of the integral control. The regulator for q=\.O acts
satisfactorily on adding the limiter, i.e. the excessive control
0.02
0.8926
1 - 0.01
1.0 2.0
De,p.u.MW/Hz 10 12 14 16 18 20
b
25
Tie = 1.1677 p.u.MW/Hz s
n =0.8926 p.u.MW/Hz
-0.04 L
20
Fig. 7 Control effects with a generation rate constraint under step
load change APh = 0.01 p.u.MW
Area 2: generation rate constraint of 10% per minute
a With conventional regulator
S 15 b With optimal regulator for q = 1.0
c With optimal regulator for q = 10.0
Areas 1, 3 and 4: without constraints and regulators
using detailed 4-area model
using external equivalent
10
amax
5- 2.7355
t L_
h
1
A ACE-
10 20
/ s
-AP
a max
Fig. 6 Variation of index J associated with parameter value changes L
of external equivalent for area 2
limiter
regulator with optimal gains
regulator with initial gains Fig. 8 Limiter for preventing excessive control action
detailed 4-area model, as shown in Figs. 7 and 9. Initial gains 0. 1 0.0 0.0 0.0 0.0 12.2553
Optimal gains 0. 5269 0.0634 0.0655 - 0 . 9 4 7 1 - 0 . 8 0 3 7 5.9100
5 Conclusion
In this paper a new method of designing a decentralised
regulator has been proposed, and the effects of the proposed
6 Acknowledgment
The author is grateful to Prof. T. Suyama for his kind discussion
about this work.
and for the 2nd-order sensitivity terms
7 References (17)
5a5/3
1 FOSHA, C.E., and ELGARD, O.I.: 'The megawatt-frequency con-
trol problem: A new approach via optimal control theory', IEEE
Trans., 1970, PAS-89, pp. 563-567 Qap =
2 CALOVIC, M.: 'Linear regulator design for a load and frequency 5a5|3 5a50
control', ibid., 1972, PAS-91, pp. 2271-2285
3 MOORTHI, V.R., and AGGARWAL, R.P.: 'Suboptimal and near- 8AT 8K 8K 8A 8AT 8K 8K 8A
j j j J
optimal control of a load-frequency-control system', Proc. IEE,
1972, 119, (11), pp. 1653-1660 5a 5/3 5/3 5a 5/3 5a 5a 5/3
4 VENKATESWARLU, K., and MAHALANABIS, A.K.: 'Design of where the terms a and /3 denote the regulator gains hit and also
decentralised load-frequency regulators', ibid., 1977, 124, (9),
pp. 817-821 present the parameters of the external equivalent as described
5 BENGIAMIN, N.N., and CHAN, W.C.: 'Multilevel load-frequency in Section 4.1.4.
control of interconnected power systems', ibid., 1978, 125, (6), In the numerical calculation [11], first the Lyapunov
pp. 521-526 eqn. 10 is solved, then eqns. 16 and 17 are solved successively
6 YU, Y.N., EL-SHARKAWI, M.A., and WVONG, M.D.: 'Estimation using the solution matrices K, 8K/8oc, and 5AT/5/3.
of unknown large power system dynamics', IEEE Trans., 1979,
PAS-98, pp. 279-289 The computation time required for calculating the 1st- and
7 NANDA, J., and KAUL, B.L.: 'Automatic generation control of an 2nd-order sensitivities is much reduced, because all the
interconnected power system', Proc. IEE, 1978, 125, (9), pp. equations have the same coefficient matrix A. Moreover, the
385-390 matrices 8A/8oc, 5/4/5/3, and 52/4/5a5/3 are sparse, so the sparse
8 CUNO, B.: 'A new design technique for automatic generation
control'. Proceedings of the 7th PSCC, 1981^pp. 677-681 matrix techniques can be used for the storage and for the
9 ANDERSON, J.H., HUTCHISON, M.H., WILSON, W.J., ZOHDY, computation.