Professional Documents
Culture Documents
2
ISSN 1827- 6660 March-April 2012
Abstract – Under-frequency load shedding is a common technique to maintain the power system
stability by removing the overload in some part of the system. The existing under-frequency load
shedding schemes operate successfully in the interconnected grid system. A particular approach
of under-frequency load shedding scheme is required for an islanded system which is known to be
not as strong as the grid. This paper presents a new approach of load shedding scheme in an
islanded distribution system incorporating distribution state estimation. The scheme is based on
an intelligent technique to reduce the effect of power deficiency in an islanded network. The
disturbance magnitude is calculated based on rate of change of frequency by using the simplest
expression of the generator swing equation. Moreover, the process for assessing the power
consumption of the customer can be achieved by state estimation method. In this paper, state
estimation for a distribution network is developed and enhanced by composite load model to
obtain the most practical value of the system state variables. Finally, the calculated value of the
disturbance magnitude will be shed according to the estimated value of loads in each bus.
Copyright © 2012 Praise Worthy Prize S.r.l. - All rights reserved.
Nomenclature I. Introduction
H Generator inertia constant Nowadays, islanding operation of a distribution
fc Frequency of the center of inertia network is a topic of interests to improve security and
Pi Real power injection reliability of a network. Implementing an intentional
Qi Reactive power injection islanding operation of Distributed Generation (DG) will
Pij Real power flow establish continuity of supply in the customer side [1]-
Qij Reactive power flow [2]. Accordingly, IEEE 1547 group has produced a
hi(x) Measurement matrix of the nonlinear standard which serve as a guide for practicing an
function intentional islanding operation in an electric power
H(x) Jacobian matrix of the measurement system [3]. However, the operation of a DG in an
function islanded brings up some concerns.
P Imbalance power The most critical concern is to maintain the power
DG Distributed generation quality of the islanded system. A sudden change in
UFLS Under-frequency load shedding generation and load immediately after system is islanded
SE State estimation poses a challenge in controlling the frequency response.
EMS Energy management system A large excess of load over local generation in the
DSE Distribution state estimation islanded system could result in a rapid frequency drop.
WLS Weighted least square The smaller system inertia tends to cause a quick
AUFLS Adaptive Under-frequency load shedding frequency drop. It also has a limited spinning reserve to
ROCOF Rate of change of frequency cover for the shortfall of the generation. For such an
FCM Frequency calculator module imbalance power, DG would not be able to stabilize the
LSCM Load shedding controller module frequency response. The only way to stabilize the
DSEM Distribution state estimation module frequency to its nominal value is to reject several loads.
COI Center of inertia It is a common practice among the power utilities to
employ a conventional Under-Frequency Load Shedding
(UFLS) scheme. The UFLS relay is set to shed certain
amount of loads based on a specified frequency
threshold. Although the scheme is simple, it is widely
Manuscript received and revised March 2012, accepted April 2012 Copyright © 2012 Praise Worthy Prize S.r.l. - All rights reserved
4155
M. Karimi, H. Mohamad, H. Mokhlis, A. H. A. Bakar
known to be unreliable in shedding the right amount of an islanded distribution system. The strategy utilizes an
load [4]-[7]. In an adaptive UFLS scheme, the amount of adaptive UFLS scheme incorporating an intelligent
load to be shed is determined based on the estimated technique using state estimation for distribution network
magnitude of the disturbance (power imbalance). The application. A composite load model is utilized in state
magnitude is normally estimated using the swing estimation to compute an accurate value of active and
equation which requires an estimated value of the rate of reactive power of various types of loads such as
change of frequency (df/dt) in the system [8]-[10]. industrial, commercial and domestic. The proposed
Different aspects of intelligent load shedding scheme Distribution State Estimation (DSE) is based on
have been reported in the literature [11]-[14]. Intelligent Weighted Least Square (WLS) method to find and
scheme utilized power system automation, two-way high capture a network operation point. It would be able to
speed communication system, fast and accurate intelligently shed the load in accordance to the load
measurement units and other modern facilities to acquire priority. The proposed scheme is implemented to an
knowledge and information regarding the network for intentional islanding operation considering Malaysia’s
intelligently and optimally shedding the load [15]-[18]. existing distribution network that consists of two small
Therefore, the intelligent load shedding scheme for an units of hydro generation connected in parallel to the
islanded distribution network can be achieved by grid.
implementing a real-time monitoring of the network.
However in practice, measurement devices are very
few due to high cost of these devices. Only selected II. Proposed Under-Frequency Load
locations are being monitored, mainly at the critical area Shedding Scheme
and main substations. The problem of limitation number
In this paper a new approach of UFLS Scheme for an
of measurement equipment can be found in many
islanded distribution network is proposed. The proposed
conventional distribution systems. Very often distribution
method combined an Adaptive Under-Frequency Load
network is monitored only at the primary substation. One
Shedding (AUFLS) and DSE for islanded distribution
way to address this issue is by determining the state
network.
condition of a distribution network using State
As reported in the literature, AUFLS technique
Estimation (SE). State estimator could approximate and
executes the load curtailment in two steps. In the first
estimate the power flow (active and reactive power) and
step, it clarifies and calculates the rate of change of
voltage at a bus which measurement are not available.
frequency (ROCOF) in the system immediately after the
SE is a data processing scheme which is used to find
occurrence of the disturbance, at t = 0+ s. Then, it
the best possible values of the state. In the modern
determines and computes the power imbalance (that
Energy Management System (EMS), SE program
causes the frequency drop) in the system based on simple
processes a set of raw measurement data and provides a
expression of the generator swing equation. The scheme
real-time solution which is the basis of the advanced
will always try to shed right amount of load which is
functions for system security monitoring and control
close to the calculated power imbalance.
[19]. The ability of the state estimator to achieve a high
In general, the proposed scheme adopted the AUFLS
level of efficiency and numerical robustness is very
strategy which using ROCOF to determine the amount of
importance in today’s electric utility industry that heavily
load to be shed. It is designed to compute the value of
interconnected with loads. Pseudo measurement is
unbalanced power based on AUFLS. Therefore, to shed
commonly applied in SE due to address the limitation
the right amount of load, the proposed method utilizes
number of measurements. Customer load pseudo
DSE to estimate the online value of power for load at
measurements are obtained from the historical data, load
each bus. DSE program perform as a real-time
forecasting or other similar approximation methods in
monitoring device in the distribution network. This is
customer side. Hence, a few numbers of real-time
therefore solved the lacking of real time measurement in
measurements in a system is acceptable when using SE.
the distribution network.
Various techniques have been used to obtain an SE
Another presetting step of the load shedding scheme is
solution. Excellent surveys on SE algorithms can be
to clarify the load priority. The load shedding is
found in [20]-[22]. Two types of data are required for
performed according to the load priority in the network.
SE; the network data and measurement data. A real-time
The loads are classified into 3 categories; vital, semi-
central control system can be used to improve the
vital and non-vital. The non-vital load will be the first
security and reliability of electrical power system [23]. A
priority to be shed, followed by the semi-vital and lastly
SE algorithm based on forward-backward propagation
the vital load. Fig. 1 illustrates the overall concept of
with higher R/X ratio of lines has been presented in [24]-
load shedding scheme for an islanded distribution
[25]. A branch estimation based on SE method, suitable
network connected with Mini Hydro power plants.
for real-time monitoring of distribution network is
The scheme consists of three main modules; (a)
presented in [26].
Frequency Calculator Module (FCM), (b) Load
This paper proposes a new conceptual strategy of a
Shedding Controller Module (LSCM) and (c)
load shedding scheme which will provide a solution for
Distribution State Estimation Module (DSEM).
Copyright © 2012 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 7, N. 2
4156
M. Karimi, H. Mohamad, H. Mokhlis, A. H. A. Bakar
where:
II.2. Load Shedding Controller Module
- Zi is the vector of measurements
In the LSCM, the amount of loads that are needed to - X is the state vector
be shed is determined and calculated. A certain delay - hi(x) is the measurement matrix of the nonlinear
time equal to 100ms which comprises communication, function
calculation and circuit breaker opening time is being - ri is the measurement errors vector.
considered in this module. The LSCM will have to Then, the measurement residual vector can be defined
decide the power imbalanced value based on the and formulated in (3) as follow:
ROCOF. This strategy is to address the changes of
frequency due to sudden increment of load in an islanded ri Zi hi x , i 1, 2,...,m (3)
system. In this strategy, the amount of load to be shed
depends on the disturbance magnitude. It is estimated By applying the WLS method, it will minimize the
based on the amount of power imbalance following objective function as describe in (4):
frequency deviation in the system. The amount of power
imbalance is estimated by using the ith generators swing m 2
Zi hi x
equation which is expressed in (1): J x (4)
i 1 Ri
N
2 Hi df c
P i 1
(1) - Ri diagonal matrix of [1/ 12, 1/ 22, …, 1/ m2]
fn dt 2
- m is the covariance of the mth measurement error.
The best solution of the state vector X can be achieved
where: by the solution of the following equation. It is calculated
Copyright © 2012 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 7, N. 2
4157
M. Karimi, H. Mohamad, H. Mokhlis, A. H. A. Bakar
to reflect the expected accuracy of the corresponding estimator cannot attain convergence. Therefore, pseudo
meter used: measurements are needed to assist convergence of the
state estimator for distribution network.
T 1
G x xk 1 H xk R Z h xk (5)
III. Validation of the Proposed Load
where: Shedding
- H(x) is the Jacobian matrix of the measurement
function The test system consists of 30 buses, 27 lumped loads
- h(x) which is mentioned in (2) with 2 mini hydro generators. The mini hydro generators
- G(x) is called the Gain matrix; it can be defined and rated at 2MVA with 2.5 s of constant inertia are operated
formulated in (6) as follow: at 3.3kV. Two parallel units of 2MVA generator
transformer were connected to the DGs to step up the
T 1
voltage level to 11kV. The transmission grid is
G x H xk R H xk (6) connected to the distribution network via steps down
transformers (132kV/11kV). As indicated in Fig. 2, a
By solving the (5) to minimize such errors and find an Normal off Point (NOP) for 11kV network is located at
optimal estimate, system state will be solved, iteratively. the intersection of two buses 21 and 22. Each load is
In this paper, WLS state estimation, as a mathematical connected with motorized switchgear that can be
analysis tools and an iterative solution, is presented to remotely controlled for load shedding purposes. Non-
filter out such errors and find an optimal estimate of Elastic Water Column without Surge Tank and PID
system states in distribution network. The measurement Controls including Pilot and Servo Dynamics model are
matrix can be of a variety of most commonly used used as a hydro turbine and governor model,
measurements in the practical system. The line power respectively. IEEE Alternator Supplied Rectifier
flows, bus power injections, bus voltage magnitudes and Excitation System is also chosen as an excitation model.
line current flows are common measurements in real The model parameters are shown in Table I, II and III
power system. The proposed DSEM is used the line respectively. The system state variable measurement (i.e.
power flows, bus power injections and bus voltage active power, reactive power and voltage) are monitored
magnitudes in order to construct the measurement by the DSEM. The system frequency is also calculated
matrix. Real and reactive power injection, Pi and Qi, by FCM to prepare the system information for LSCM.
equations at bus i can be expressed by below: The LSCM collect the data from FCM and DSEM to
determine or estimate the amount of power imbalance
m and accordance load need to be shed based on DSE
Pi Vi Gij cos ij Bij sin ij Vj (7) values.
j 1
TABLE I
HYDRAULIC TURBINE PARAMETERS
m
Parameter Value Parameter Value
Qi Vi Gij sin ij Bij cos ij Vj (8)
TW 1.2 Initial output power 0.7
j 1
fP 0.02 Initial operating head 1.0
D 0.5 Rated output power 1.0
where:
Gij is a line conductance from bus i to bus j in Siemens TABLE II
GOVERNOR PARAMETERS
(real part of admittance matrix),
Parameter Value Parameter Value
Bij is a line susceptance from bus i to bus j in Siemens KP 2.25 TC 0.2
(imaginary part of admittance matrix). KI 0.37 TD 0.2
Real and reactive power flow, Pij and Qij, equations KD 0.9 Max gate opening 0.16
from bus i to bus j can be expressed as follow: TA 0.05 Max gate closing 0.16
RP 0.04 Dead band value 0
Max gate position 1.0 Min gate position 0
Pij Vi 2 Gij Gi 0 ViV j Gij cos ij Bij sin ij (9)
TABLE III
EXCITATION PARAMETERS
2
Qij Vi Bij Bi 0 ViV j Gij sin ij Bij cos ij (10) Parameter Value Parameter Value
TC 0 KF 0.03
TB 0 TF 1
The proposed DSEM is based on voltage KA 400 TE 0.8
measurements. The power injection and power flow is TA 0.02 KE 1
computed by using (7), (8), (9) and (10) at each buses of VAMAX 14.5 KC 0.2
the distribution network. Finally, the calculated values VAMIN -14.5 KD 0.38
VRMAX 6.03 VRMIN -5.43
are used to construct the measurement matrix (hi) in SE(VE1) 0.1 SE(VE2) 0.03
order to solve (5). Since real-time information is very VE1 4.18 VE2 3.14
limited in a conventional distribution system, the state
Copyright © 2012 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 7, N. 2
4158
M. Karimi, H. Mohamad, H. Mokhlis, A. H. A. Bakar
Copyright © 2012 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 7, N. 2
4159
M. Karimi, H. Mohamad, H. Mokhlis, A. H. A. Bakar
TABLE V
AMOUNT OF LOAD SHEDDING FOR BASE LOAD IN CASE II
Load Case Power Imbalance (MW) Load Shed (MW)
Base Load 0.30 0.227
Base Load 0.45 0.290
TABLE VI
AMOUNT OF LOAD SHEDDING FOR PEAK LOAD IN CASE II
Load Case Power Imbalance (MW) Load Shed (MW)
Peak Load 0.30 0.113
Peak Load 0.45 0.284
2
Peak Load
1 Base Load
-1
-2
-3
-4
-5
-6
47.5 48 48.5 49 49.5 50 50.5 Fig. 6. Frequency response of the overload for case II
Frequency (Hz)
0.4
Fig. 5. df/dt versus frequency response in case I 0.45MW Overload
0.2 0.30MW Overload
0
IV.2. Case II: Sudden Connection of a Load
-0.2
load shedding.
49.4
For peak load case, the LSCM is also compute the
amount of load need to be shed based on AUFLS and 49.2
DSE. Fig. 8 shows the frequency response of the system
49
in this case. The amount values of load shedding are 0 5 10 15 20 25 30 35 40 45
Time (s)
presented in Table VI. Fig. 9 is illustrated the df/dt
versus the frequency response of the network. It is Fig. 8. Frequency response of peak load for case II
observed that the system frequency could be converged
and restored back to the normal value after applying the
load shedding.
Copyright © 2012 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 7, N. 2
4160
M. Karimi, H. Mohamad, H. Mokhlis, A. H. A. Bakar
Copyright © 2012 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 7, N. 2
4161
M. Karimi, H. Mohamad, H. Mokhlis, A. H. A. Bakar
[26] Y. Deng, et al., "A branch-estimation-based state estimation H. Mokhlis received his B. Eng in Electrical
method for radial distribution systems," IEEE Transactions on Engineering in 1999 and M. Eng. Sc in 2002
Power Delivery, vol. 17, pp. 1057-1062, 2002. from University of Malaya, Malaysia. His
[27] J. J. Ford, et al., "Adaptive load shedding and regional obtained PhD degree from the University of
protection," International Journal of Electrical Power & Energy Manchester, UK in 2009. Currently he is a
Systems, vol. 31, pp. 611-618, 2009. Senior Lecturer in the Department of Electrical
[28] M. Karimi, H. Mohamad, H Mokhlis, A. H. A. Bakar, " Under- Engineering, University of Malaya. His main
Frequency Load Shedding Scheme for Islanded Distribution research interest is in distribution automation
Network Connected with Mini Hydro," International Journal of area and power system protection.
Electrical Power & Energy Systems, vol. 42, pp. 127-138, 2012.
[29] U. Rudez and R. Mihalic, "Monitoring the First Frequency A. H. A. Bakar received his B.Sc. in Electrical
Derivative to Improve Adaptive Underfrequency Load-Shedding Engineering in 1976 from Southampton
Schemes," IEEE Transactions on Power Systems, pp. 1-1, 2011. University UK and M.Eng. and PhD from
[30] M. Giroletti, et al., "A hybrid frequency/power based method for University Technology Malaysia in 1996 and
industrial load shedding," International Journal of Electrical 2003. He has 30 years of utility experience in
Power & Energy Systems, 2011. Malaysia before joining academia. Currently he
is Consultant in the Department of Electrical
Engineering, University of Malaya, Malaysia.
Authors’ information
1
Departement of Electrical Engineering, Faculty of Engineering,
University of Malaya, 50603 Kuala Lumpur, Malaysia
2
University of Malaya Power Energy Dedicated Advanced Centre
(UMPEDAC), Level 4, Wisma R&D UM, University of Malaya, 59990
Kuala Lumpur, Malaysia.
3
Faculty of Electrical Engineering, University of Technology MARA,
Malaysia.
Copyright © 2012 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 7, N. 2
4162