International Review of Electrical Engineering (I.R.E.E.), Vol. 7, N.

4
ISSN 1827- 6660 July - August 2012

A Fuzzy Based Under-Frequency Load Shedding Scheme

for Islanded Distribution Network Connected with DG

H. Mokhlis1, 2, J. A. Laghari1, 3, A. H. A. Bakar2, M. Karimi1, 2

Abstract The frequency of power system is very sensitive to load changing when operating in
islanded mode. This may causes overloading or loss of generation cases. Under-Frequency Load
Shedding (UFLS) Scheme is commonly applied to stabilize the frequency during these cases.
Conventional UFLS scheme operates successfully in interconnected grid system and may not work
well when applied to DG based system operating in islanded mode. This paper presents a new
fuzzy logic based under- frequency load shedding scheme implemented on mini hydro type-DG
operating in islanded mode.
The proposed strategy is based on frequency, rate of change of frequency and load prioritization.
In proposed UFLS scheme, a fuzzy logic load shedding controller (FLLSC) with Load Shed
Controller Module (LSCM) is modelled. FLLSC measures amount of load to be shed and LSCM
shed the respective load to stabilize frequency. The proposed scheme is validated on different
event-based and response-based cases. Simulation results show that proposed scheme is effective
in shedding optimal number of loads while stabilizing the frequency. Copyright 2012 Praise
Worthy Prize S.r.l. - All rights reserved.

Keywords: Distributed Generation (DG), Fuzzy Logic Load Shedding Controller (FLLSC), Load
Shed Controller Module (LSCM), Islanded Distribution Network

Nomenclature The use of DG has provided benefits to the end user,

power utilities and DGs owner in terms of reliability,
fc Frequency of the center of inertia efficiency of power and economics [3]-[5]. However,
f Frequency deviation penetration of DG causes various problems to the
df/dt Rate of change of frequency existing network and power system need to be modified.
H Generator inertia constant One of the constraints is operation of DG during
P Active power at new voltage/frequency islanded mode, in which DG is electrically isolated from
P0 Active power at base voltage/frequency the main grid [6], [7]. This causes overloading or loss of
Q Reactive power at new voltage/ frequency generation cases and requires load shedding technique to
Q0 Reactive power at base voltage/ frequency stabilize the frequency within acceptable range.
P Imbalance power The conventional load shedding scheme employs
DG Distributed Generation frequency relay to stabilize frequency under abnormal
FLLSC Fuzzy logic load shedding controller conditions. In this scheme, the under frequency relay
RCB Remote Circuit Breaker operates when system frequency falls below a certain
UFLS Under-frequency load shedding threshold value. Conventional UFLS scheme shed a fixed
LSCM Load shedding controller module amount of electrical power in fixed steps.
This scheme is unreliable in shedding the optimal
number of loads [6]-[9]. Efforts have been made to
I. Introduction
improve the existing conventional UFLS scheme by
Distributed generation (DG) refers to small type of combining the generator tripping with UFLS scheme in
electric power generation having capacity less than 10 order to stabilize power system during unstable
MW [1]. DG units are located near load centers in order faults[10]. Another effort for improving the conventional
to avoid expansion of the existing network and supply load shedding scheme is presented by simultaneously
new load areas. studying the transmission and distribution network [11].
Due to market deregulation and environmental Since, conventional load shedding scheme shed the
constraints, the use of DG resources has been widely load in fixed steps without estimating the amount of load
employed in power industry[2]. to shed. Hence, it often shed more load than required.
The increasing trend of DG penetration in power This problem is solved by employing power swing
system network has opened the new challenging issues in equation to estimate the amount of load to be shed. The
the field of power system. technique is known as an adaptive UFLS scheme.

Manuscript received and revised July 2012, accepted August 2012 Copyright 2012 Praise Worthy Prize S.r.l. - All rights reserved

4992
 H. Mokhlis, J. A. Laghari, A. H. A. Bakar, M. Karimi

The literature on adaptive UFLS scheme has reported monitors system status at every instant of time, and
that the most schemes employ voltage variation to determines frequency of the equivalent inertial center fc
identify and shed the sensitive load buses [12], a as below [22]:
combination of frequency, df/dt and voltage changes [13] N
and initial slope of df/dt for setting the under-frequency H i fi
relays [14]. These schemes can operate successfully, fc i 1
(1)
N
provided that grid is facilitated with high speed
Hi
communication technology. This leads to intelligent
i 1
UFLS schemes. The intelligent UFLS schemes employ
two-way high speed communication system, power
where, Hi is inertia constant of ith generator in s and fi is
system automation. The intelligent UFLS schemes
frequency of ith generator in Hz and N shows number of
require fast and accurate measurements units to acquire
generators. Fig. 1 illustrates layout of load shedding
knowledge and information regarding the network for
scheme. Standard frequency pick value to begin load
optimally shedding the load [15]-[17].
shedding scheme is set to 49.5 Hz as practised in TNB,
These UFLS scheme may not ensure system security
Malaysia [23].
when applied in islanded distribution network connected
When DG system encounter load disturbance, FLLSC
with DG. This is due to the fact that system frequency
check frequency limit of 49.5 Hz. If frequency goes
severely disturbed during islanding mode. Also, DG
below 49.5Hz, FLLSC investigate about the type of
system has smaller inertia which causes to drop
disturbance (Event-based or Response-based) occur on
frequency quickly. Hence, a DG system operating in
DG and estimates amount of load to be shed.
islanded mode requires an appropriate adaptive load
shedding scheme. The research on UFLS schemes based
on islanded system has shown that not much work has
been conducted in this area. The developed load
shedding schemes by various authors are based on
frequency and df/dt information, customers willingness
to pay and load histories [18] and best time to shed the
loads [19].
This paper proposes a new fuzzy based under-
frequency load shedding scheme for islanded distribution
network connected with DG. The proposed strategy uses
frequency, rate of change of frequency and load
prioritization to shed the optimal number of loads. The
proposed UFLS scheme is tested at event-based and
response-based load disturbance as proposed in [20],
[21].
It consists of fuzzy logic load shedding controller
(FLLSC) with load shedding controller module (LSCM).
FLLSC estimates the power imbalance and LSCM shed
the required load to stabilize frequency.

II. Methodology
II.1. Proposed Load Shedding Scheme Description
The scheme is proposed to operate and monitor
distribution network after the occurrence of grid
disconnected. The proposed load shedding scheme uses
fuzzy logic control approach to stabilize frequency by
shedding some amount of load. The scheme has two
parts. Fuzzy logic load shedding controller (FLLSC) and
load shed controller module (LSCM). The FLLSC uses
frequency and rate of change of frequency as input, and
determines type of load disturbance whether Event or
Response-based and estimates the power imbalance.
FLLSC sends this value to LSCM, which shed the
required load according to load priority. The loads are
prioritized into three categories; vital, semi-vital and non-
vital. Non-vital loads have the lowest priority and will be
shed first followed by semi-vital and vital loads. FLLSC
Fig. 1. Proposed fuzzy based UFLS scheme layout
In order to avoid unnecessary shedding of load due to
smaller disturbances, the absolute value of Pmin is fixed
to 50 kW as this is minimum load value in the
distribution network. If estimated value is greater than
Pmin, FLLSC sends estimated P to LSCM for shedding
respective loads. FLLSC sends estimated value to LSCM
via communication link.
The delay time which includes calculation time,
communication time and circuit breaker operation time is
assumed as 100 ms, which is according to practical
considerations [6], [12].
The co-ordination of under-frequency protection of
generator with under-frequency load shedding scheme is
very important. If system frequency goes below certain
threshold value, under frequency protection relay of

Copyright 2012 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 7, N. 4

4993
 H. Mokhlis, J. A. Laghari, A. H. A. Bakar, M. Karimi
generator will operate and system will collapse
unnecessarily.
Hence, under-frequency load shedding technique
should be applied in such a way that frequency recovers
without going below threshold value.
The minimum allowed operating frequency usually
specified by the manufacturer according to the type of
turbine is 47.5 Hz [24].
In this paper, distribution network is assumed to have
reliable monitoring devices and fast communication
system for transmitting data.
Real-time measurement and Remote Circuit Breaker
(RCB) are facilitated at each of the load feeder. The
system state variable measurement (i.e. active power,
frequency and voltage) are monitored by FLLSC whereas
breakers status of load feeders is monitored by LSCM.
The flow chart of proposed scheme is shown in Fig. 2.
In FLLSC, there are two strategies; (1) Event-based
and (2) Response-based scheme. FLLSC decides right
strategies based on frequency, df/dt and breaker status at
the DG units. The description of these strategies is as
follows:
Fig. 2. Flow chart of proposed FLLSC controller and LSCM module
Copyright 2012 Praise Worthy Prize S.r.l. - All rights reserved
4994
II.2. FLLSC for Event-Based Case
Event-based case may occur when one of DG unit is
tripped during islanded mode. This tripping incident may
be initiated by the failure operation or malfunction of
generator differential protection. It may also happen due
to transmission line failure in power system. FLLSC will
estimate the amount of load to be shed thus, preventing
system from blackouts. When Event-based occurs,
FLLSC will estimate total power imbalance as given in
Equation (2):
P PDG PLoad (2)
where, P is power imbalance, PDG is DG dispatching
power, PLoad is total load demand.
FLLSC sends estimated value of P to LSCM which
sheds the load according to load priority defined in load
look-up table.
II.3. FLLSC for Response-Based Case
Response-based case occurs due to sudden increment
of load in an islanded system. In this case, number of
load to be shed depends on the disturbance magnitude.
FLLSC estimates the power imbalance by using
frequency and df/dt as input signals. After estimating
power imbalance, FLLSC sends estimated value to
LSCM for shedding the required load.
II.4. Fuzzy logic Load Shedding Controller (FLLSC)
Modelling
Fuzzy logic load shedding controller is modelled in
PSCAD software. Since, PSCAD does not provide fuzzy
logic tool box, FLLSC is modelled by writing C-coding.
Fuzzy logic load shedding controller for UFLS
scheme consists of two inputs and one output. The inputs
of fuzzy logic load shedding controller are frequency (f)
and rate of change of frequency (df/dt). The output is
amount of load shed (Lshed). Depending upon the input
values, fuzzy logic will estimates the amount of load
required to be shed. The FLLSC comprises of
fuzzification, rule base, inference mechanism and
defuzzification steps as shown in Fig. 3.
Fig. 3. Fuzzy logic load shedding controller block diagram
The respective membership function of frequency and
rate of change of frequency (df/dt) and Lshed are shown
in Figs. 4-6. The linguistic variables of input frequency
are Low (Low), Vlow (Very Low), Extlow (Extremely
Low), Vextlow (Very Extremely Low) and input rate of
change of frequency (df/dt) membership functions are
HN (High Negative), LN (Low Negative), LP (Low
Positive), HP (High Positive).
International Review of Electrical Engineering, Vol. 7, N. 4
 H. Mokhlis, J. A. Laghari, A. H. A. Bakar, M. Karimi

The linguistic variables of output Lshed are Vsshed defuzzification is carried out through weighted average
(Very Small Shed), Sshed (Small Shed), Bshed (Big to convert the fuzzy linguistic variable into real crisp
Shed), Vbshed (Very Big Shed). values.
TABLE I
RULE TABLE FOR FUZZY LOGIC LOAD SHEDDING CONTROLLER
(FLLSC)
Frequency
Low Vlow Extlow Vextlow
HN Sshed Bshed Bshed Vbshed
LN Sshed Sshed Bshed Vbshed
LP Vsshed Vsshed Ssshed Sshed
HP Vsshed Vsshed Vsshed Vsshed

III. Test System for Under-Frequency

Load Shedding Scheme (UFLS)
Fig. 4. Frequency membership functions
The test system for analyzing the proposed UFLS
scheme is shown in Fig. 7. The DG test system consists
of two mini-hydro power plants units. Each DG unit has
2 MVA (maximum power dispatch is 1.83MW) capacity
and is modelled in PSCAD/EMTDC software. The
system consists of 27 buses and 20 lumped loads.
The DG units are operated at 3.3 kV voltage level and
are connected with transformer to step-up the voltage
level to 11 kV. Each node is connected with remote
circuit breaker (RCB) that can be remotely controlled for
load shedding purposes.

Fig. 5. Rate of change of frequency (df/dt) membership functions Load 5

Grid

1047 1079
Disconnected
from Grid

1046 1058

1056
Load 12 1075 1057
1039
1018 Load 6
Vital Load
Load 7
Load 9
1154
1019 1010 Load 13
1013
Load 1

Vital Load
1050
1020 Load 8
1012 Load 4
Load 3
1004 1141 1151 1064 1029
Load 11
1000

DG 1 Load 10 Load 2 Vital Load Load 14 Load 15

DG 2

Fig. 6. Lshed (p.u.) membership functions

The fuzzy logic load shedding controller input and
output membership functions are formed in C-coding by
using one dimensional array concept.
The triangular membership functions are divided into
two slope equations for fuzzification. In fuzzification, the
real input values are converted into fuzzy set values
which assign degree to which these inputs belong to each
of the appropriate fuzzy sets.
Fuzzy rule base is used in IF-THEN rule form to
assign the input and output control such as:
IF frequency is low and df/dt is HN THEN Lshed is
Sshed.
IF frequency is Vextlow and df/dt is HN THEN Lshed is
Vbshed.
The other rules of fuzzy logic controller are
summarized in Table. I.
The inference mechanism evaluates active signals for
taking control actions from the fuzzy rules. Finally,
Fig. 7. Mini hydro power plant with distribution network
disconnected from Grid
The distribution network having two DG units is
assumed to be disconnected from grid. The standard
model for exciter and governor and hydraulic turbine
provided in PSCAD/EMTDC library are used in this
study. The IEEE type AC1A excitation standard model is
chosen in this work and its model parameters are shown
in Table II.
The governor consists of PID controller including
pilot and servo dynamics models and its parameters are
shown in Table III.
The hydraulic turbine for this study is considered as
Non-elastic water column without surge tank and its
parameters are shown in Table IV. The distribution
network has load profile consisting of base load and peak
load capacity.

Copyright 2012 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 7, N. 4

4995
 H. Mokhlis, J. A. Laghari, A. H. A. Bakar, M. Karimi

TABLE II where P is active power at new voltage and frequency,

EXCITATION PARAMETERS
and P0 is active power at base voltage and frequency, Q
Parameter Value Parameter Value
TC 0 KF 0.03
is reactive power at new voltage and frequency, and Q0 is
TB 0 TF 1 reactive power at base voltage and frequency. Kpf and Kqf
KA 400 TE 0.8 are the co-efficient of active and reactive load
TA 0.02 KE 1 dependency on frequency. The parameter of this model
VAMAX 14.5 KC 0.2
VAMIN -14.5 KD 0.38
are the exponents a and b. With these exponents equal to
VRMAX 6.03 VRMIN -5.43 0, 1 or 2, the model represents constant power, constant
SE(VE1) 0.1 SE(VE2) 0.03 current, or constant impedance characteristics,
VE1 4.18 VE2 3.14 respectively [25]. f is frequency deviation (f - f0). In this
study, the exponent value for a and b is set to 1.0 and 2.0
TABLE III
GOVERNOR PARAMETERS respectively. Meanwhile, Kpf and Kqf is set to 1.0 and -1.0
Parameter Value Parameter Value respectively. Thus, by choosing these values, load is
KP 2 TC 0.2 voltage and frequency dependent type. The power
KI 0.35 TD 0.2 consumption of each load and its rank type is presented
KD 0.9 Max gate 0.16 in Table V. The loads are ranked based on their load
opening
TA 0.05 Max gate 0.16
priority and look-up table is created according to
closing prioritization. This load priority is created on active
RP 0.04 Dead band 0 power value of each load. Both event-based and
value response-based strategies of UFLS scheme are
Max gate 1.0 Min gate 0
position position
considered. Depending upon the system load capacity,
generating system can be operated at base load and peak
TABLE IV load capacity. respectively [25].
HYDRAULIC TURBINE PARAMETERS
Parameter Value Parameter Value
TW 2 Initial output power 0.7 IV. Case Studies
Initial operating
fP 0.02 1.0
head Under-frequency load shedding scheme is tested on
D 0.5 Rated output power 1.0 Event-based and Response-based case studies. All case
studies of load shedding scheme are tested on islanded
In a real system, the loads are always frequency and distribution network connected with DG. The
voltage dependent. This dependency of load descriptions of case studies are summarized in Table VI.
characteristics is usually modelled as:

V
a V. Result and Discussions
P P0 1 K pf f (3)
V0 V.1. Case I: Event-Based Case without Load Shedding
Scheme
b To simulate Event-based case without applying load
V
Q Q0 1 K pf f (4) shedding scheme, one of the DG unit is tripped-off from
V0 islanded distribution system.
TABLE V
LOAD RANKING TABLE
Load Bus Peak Load Base Load Load
Ranked Number P(MW) Q(MVAr) P(MW) Q(MVAr) Category
1 1013 0.0684 0.0423 0.0456 0.0282 Non-vital
2 1141 0.0795 0.0495 0.0531 0.033 Non-vital
3 1012 0.0795 0.0495 0.0531 0.033 Non-vital
4 1050 0.1095 0.0576 0.063 0.0384 Non-vital
5 1047-1079 0.1794 0.0792 0.11721 0.07281 Non-vital
6 1057 0.189 0.1152 0.126 0.0768 Non-vital
7 1058 0.198 0.123 0.132 0.0819 Non-vital
8 1010-1039 0.234 0.1101 0.15009 0.0933 Non-vital
9 1064 0.1488 0.0867 0.093201 0.057801 Semi-vital
10 1018 0.1743 0.108 0.11619 0.072 Semi-vital
11 1154 0.2097 0.1275 0.1401 0.0849 Semi-vital
12 1004 0.2121 0.1314 0.14151 0.0876 Semi-vital
13 1046 0.2535 0.1578 0.1701 0.1053 Semi-vital
14 1020 0.2745 0.1716 0.1845 0.11439 Semi-vital
15 1029 0.3468 0.2148 0.2313 0.1431 Semi-vital
16 1019 0.1902 0.099 0.10671 0.06609 Vital
17 1151 0.2208 0.0996 0.107199 0.06639 Vital
18 1056 0.345 0.3282 0.35259 0.2187 Vital

Copyright 2012 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 7, N. 4

4996
 H. Mokhlis, J. A. Laghari, A. H. A. Bakar, M. Karimi

TABLE VI one DG Trip

CASE STUDIES 50.5
Case Studies Description
50
Case I Event-Based case without load shedding scheme
Event-Based case with Load Shedding Scheme 49.5
Case II (1) Peak Load Scenario
(2) Base Load Scenario 49
Response-Based case with Load Shedding Scheme
Case III (1) Peak Load Scenario 48.5
(2) Base Load Scenario
48
0 10 20 30 40 50
Since, loads in islanded system are supplied by two Time (sec)
DGs, loss of one DG will give a great impact to islanded
system. Fig. 9. Frequency response for Event-based case at Peak Load
As a result, all load is shifted to the remaining
operating DG unit. The frequency response of DG Fig. 9 shows that by applying proposed load shedding
without applying load shedding scheme is shown in Fig. scheme, DG frequency has undershoot of 48.56 Hz and
8. From Fig. 8, it can be noticed that when event-based recovered to 50Hz after some time.
case occurred and load shedding scheme is not applied, The proposed scheme shed load up to 11th load ranked
frequency of operating DG unit goes below the safe for this case. It is noticed that DG frequency while
frequency limit of 47.5 Hz, stables at 28.67 Hz and never stabilizing has an overshoot up to 50.3 Hz. This
recovered to normal value. This will cause tripping of frequency overshoot explains that amount of load shed is
operating DG unit due to operation of protection relay not optimal.
which will leads to power blackouts. Hence, load To investigate this issue, load ranked 11th is not shed
shedding scheme is necessary in order to able DG unit to confirm whether amount of load shed is optimal or not
operating continuously, and prevent power system from and response is shown in Fig. 10.
power collapse.
one DG Trip
55

50

45

40

35

30
25
0 10 20 30 40 50
Time (s)
Fig. 10. Frequency response at peak load when load ranked 11th
is not disconnected
Fig. 8. Frequency response at Event-based case without load shedding
scheme
Fig. 10 shows that proposed fuzzy based UFLS
scheme sheds the optimal number of loads as the
V.2. Case II: Event-Based Case with Load Shedding frequency does not rise to normal value and stables at 46
Scheme Hz when one less load is shed.
DG unit enable to supply 1.83 MW load after one DG
V.2.1. Peak Load Scenario being tripped and its graph is shown in Fig. 11.
To simulate Event-based case with load shedding
scheme at Peak load (3.66 MW), one of the DG unit is
tripped-off from islanded distribution system at t=10 s.
The FLLSC checks first frequency limit of 49.5 Hz.
After checking this, FLLSC check about type of load
disturbance applied on islanded distribution system.
FLLSC by monitoring RCB status of DG units
determines that system encountered Event-based load
disturbance.
FLLSC estimates the amount of load to be shed and
sends signal to LSCM, which immediately trip
significant number of load feeders to stabilize frequency.
The frequency response of DG for this case is shown
Fig. 11. Power graph for Event-based case at Peak Load
in Fig. 9.

Copyright 2012 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 7, N. 4

4997
 H. Mokhlis, J. A. Laghari, A. H. A. Bakar, M. Karimi

V.2.2. Base Load Scenario TABLE VII

OPTIMAL LOAD SHED DURING EVENT-BASED CASE
To simulate Event-based case with load shedding Event-Based Power
Undershoot Load Shed
scheme at Base load (2.5 MW), one of the DG unit is Case Imbalance
tripped-off from islanded distribution system at t=10 s. Peak Load 48.56Hz 1.83MW 1.83MW
Base Load 47.62Hz 1.25MW 1.03MW
The frequency response of DG for this case is shown
in Fig. 12. It can be noticed from Fig. 12 that by applying One DG Trip
proposed load shedding scheme, DG frequency has 2.6
undershoot of 47.62 Hz and recovered to 50 Hz after
2.4
some time. The proposed scheme shed load up to 10th
load ranked. Fig. 13 shows DG frequency when one less 2.2

load is shed. It is observed that when ranked 10th is not 2

shed, DG frequency goes to 46.141 Hz. Since, DG 1.8
frequency goes below the safe limit of 47.5 Hz and will
1.6
cause tripping of generator due to operation of protection
relay. Hence, proposed fuzzy based UFLS scheme sheds 1.4
the optimal number of loads. 0 10 20 30 40 50
Time (s)
one DG Trip
50.5 Fig. 14. Power graph for Event-based case at Base capacity
50
By adapting proposed UFLS scheme, FLLSC checks
49.5 for frequency limit of 49.5 Hz. After checking this,
49 FLLSC check about type of load disturbance applied on
48.5 system. FLLSC by monitoring RCB status of DG units
48
determines that system encountered Response-based load
disturbance. The FLLSC by measuring frequency and
47.5
0 10 20 30 40 50
df/dt, estimates the amount of load to be shed for this
Time (s) case. If estimated amount is greater than Pmin, FLLSC
sends signal to LSCM, which immediately trip
Fig. 12. Frequency response during Event-based case at Base load significant number of load feeders to stabilize the
One DG Trip
frequency. However, if estimated amount is less than
50.5 Pmin, FLLSC does not sends signal to LSCM, DG units
remains operating without requiring any load to be shed.
49.5
The frequency response of DG for this case is shown
48.5 in Fig. 15.

47.5 Load Increment

50.2
46.5
50
45.5 49.8
0 10 20 30 40 50
49.6
Time (s)
49.4
Fig. 13. Frequency response without shedding load ranked 10th
49.2

The power supplied by operating DG unit is 1.47 MW 49

and its graph is shown in Fig. 14. The optimal amount of 0 10 20 30 40 50
Time (s)
load shed during peak load and based load are shown in
Table VII. Fig. 15. Frequency response for Response-based case at Peak Load

Fig. 15 shows that DG frequency has undershoot of

V.3. Case III: Response-Based Case with Load
Shedding Scheme
V.3.1. Peak Load Scenario
To simulate Response-based condition at peak load
capacity (3.66 MW); a new load feeder rated 0.54 MW is
suddenly connected to bus number 1056 in islanded
distribution network. Upon addition of this amount of
load, total load becomes 3.66 MW + 0.54 MW = 4.2
MW.
49.078 Hz and recovers to normal value after some time.
The proposed methodology shed load up to 5th load
ranked. It can be observed that DG frequency while
stabilizing has an overshoot up to 50.15 Hz. This
frequency overshoot explains that amount of load shed is
not optimal. To investigate this issue, load ranked 5th is
not shed to confirm whether amount of load shed is
optimal or not and response is shown in Fig. 16. It can be
noticed that the proposed fuzzy based UFLS scheme
sheds optimal number of loads as frequency does not rise

Copyright 2012 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 7, N. 4

4998
 H. Mokhlis, J. A. Laghari, A. H. A. Bakar, M. Karimi

to normal value and stables at 47.8 Hz when one less The power supplied by DG is 2.9413 MW as shown in
load is shed. The power supplied by DG is 3.68237 MW Fig. 19. The optimal load shedding values for response-
and is shown in Fig. 17. based case are shown in Table VIII.
Load Increment Load Increment
50.5 3.1

50
2.9
49.5

49 2.7

48.5
2.5
48

47.5 2.3
0 10 20 30 40 50 0 10 20 30 40 50
Time (s) Time (s)

Fig. 16. Frequency response at peak load when load ranked 5th Fig. 19. Power graph for Response-based case at Base load
is not disconnected
TABLE VIII
Load Increment OPTIMAL LOAD SHED DURING RESPONSE-BASED CASE
4.2 Event-Based Power
Undershoot Load Shed
Case Imbalance
4.1
Peak Load 49.078 Hz 0.54 MW 0.5163 MW
4 Base Load 47.92 Hz 0.54 MW 0.0987 MW
3.9

3.8 VI. Conclusion

3.7
This paper has presented a new fuzzy based adaptive
3.6 under-frequency load shedding scheme, suitable for
0 10 20 30 40 50
Time (s) islanded distribution network connected with DG. The
proposed scheme used frequency, rate of change of
Fig. 17. Power graph for Response-based case at Peak load frequency, load priority to develop the load shedding
strategy.
The proposed UFLS scheme is suitable for both
V.3.2. Base Load Scenario
Event-based and Response-based cases. The
To simulate Response-based load shedding case at effectiveness and robustness of this scheme has been
Base load capacity (2.5 MW); a new load feeder rated investigated on different cases from base load to peak
0.54 MW is suddenly connected to bus number 1056 in load. Form the simulation results it can be concluded that
islanded distribution network. Upon addition of this proposed load shedding scheme successfully stabilizes
amount of load, total load becomes 2.5 MW + 0.54 MW the frequency by shedding optimal number of loads.
= 3.04 MW. Response-based load shedding is applied at
t=10s. The frequency response of DG for this case is
shown in Fig. 18. Acknowledgements
Load Increment This work was supported by University of Malaya
50.5
Kuala Lumpur under UMRG research grant (grant code:
50 D000004-16001) and Quaid-e-Awam University of
49.5 Engineering Science & Technology (QUEST),
49 Nawabshah, Sindh, Pakistan.
48.5
48 References
47.5 [1] P. P. Barker and R. W. De Mello, "Determining the impact of
0 10 20 30 40 50 distributed generation on power systems. I. Radial distribution
Time (s) systems," in IEEE Power Engineering Society Summer Meeting, ,
2000, pp. 1645-1656 vol. 3.
Fig. 18. Frequency response for response-based case at Base load [2] Akash Davda, M. D. Desai and B. R. Parekh, "Assessment of
Reduction in Losses by Distributed Generation Penetration in a
In Fig. 18, The DG frequency has undershoot of 47.92 Radial Network: a Case Study," International Review on
Hz and frequency recovers to normal value after some Modelling and Simulations (IREMOS), Vol. 4 N. 1, pp. 120-124,
February 2011 (Part A) 2011.
time. The proposed methodology shed load up to 2nd load [3] H. Mohamad, H. Mokhlis, A. H. A. Bakar and H. W. Ping, "A
ranked. review on islanding operation and control for distribution network

Copyright 2012 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 7, N. 4

4999
 H. Mo
okhlis, J. A. La
aghari, A. H. A.
A Bakar, M. Karimi
K

connected with small hy ydro power plaant," Renewablee and National Proceeedings. Power anda Energy Con nference, PEConn
Sustainablle Energy Review ws, vol. 15, pp. 39952-3962, 2011. 2004, pp. 170-174.
[4] H. Laakso onen, K. Kauhan niemi and S. Vo oima, "DG Unit Fault [24] M. Lukic, I. Kuzle
K and S. Tessnjak, "An adapttive approach too
Behavior anda Protection off LV Microgrid,"" International Review
R setting underfrequency load sheddding relays for an
a isolated powerr
on Modellling and Simulatiions (IREMOS), Vol. 3. n. 3, pp.. 353- system with private generaation," in 9th Mediterraneann
367, June 2010.
2 Electrotechnical Conference, MMELECON 1998, pp. 1122-11255
[5] Ali Reza Fereidouni,
F Hameed Nafisi, Mehdi Garmrudi and HamedH vol.2.
Hashemi Dezaki, "The Effect E of Distriibuted Generatioon in [25] P. Kundur, "Pow wer System Stabbility and Control " Mcgraw-Hill,,
Distributioon Network on Coordination off Protective Devvices," 1994.
Internation nal Review on Modelling
M and Siimulations (IREM MOS),
Vol. 4 N. 4,
4 pp. 1773-1780, August 2011 (Part B).
[6] IEEE Stanndard, "IEEE Staandard for Intercconnecting Distriibuted A
Authors infformation
Resources With Electric PowerP Systems," IEEE Std 1547--2003,
1
pp. 0_1-16 6, 2003. Deppartment of Eleectrical Engineeering, Faculty of Engineering,,
[7] IEEE Reco ommended Practtice, "IEEE Reco ommended Practice for Univ
versity of Malaya, 50603 Kuala Luumpur, Malaysia.
Utility Intterface of Photov voltaic (PV) Systems," IEEE Stdd 929-
2000, p. i, 2000. 2
UMM Power Energy Dedicated
D Advannced Centre (UM
MPEDAC), Levell
[8] X. Ding anda A. A. Girgiss, "Optimal load d shedding strateegy in 4, Wisma
W R&D UMM, University off Malaya, 59990 Kuala Lumpur,,
power sysstems with distrributed generatio on," in IEEE Power
P Malaaysia.
Engineerin ng Society Winterr Meeting, , 2001, pp. 788-793 voll.2.
3
[9] V. V. Terrzija and H. J. Koglin,
K "Adaptivee underfrequencyy load Deppartment of Electtrical Engineerinng, Quaid-e-Awaam University off
shedding integrated with h a frequency estimation num merical Engineering Science & Technologyy (QUEST), 674450, Nawabshah,,
algorithm," Generation, Transmission an nd Distribution, IEE Sindhh, Pakistan.
Proceeding gs-, vol. 149, pp. 713-718, 2002.
[10] Noor Izzrri Abdul Wahaab and Azah Mohamed, M "Trannsient Hazlie Mokh hlis received his h B. Eng inn
Stability Emergency
E Contrrol Using Generaator Tripping Bassed on Electrical Engiineering in 1999 and M. Eng. Scc
Tracking Area-Based Ro otor Angle Com mbined with UFLS," in 2002 from University of Malaya,
M Malaysia..
Internation nal Review of Eleectrical Engineerring (IREE), Vol. 5. n. His obtained P
PhD degree from the University off
5, pp. 2317 7-2326, October 2010
2 (Part B). Manchester, U UK in 2009. Cuurrently he is a
[11] Ahmed Reeza Malekpour Ali A Reza Abbasi and a Ali Reza Seiifi, "A Senior Lectureer in the Departmment of Electricall
New Coo ordinated Approaach to Load Sh hedding in Integgrated Engineering, University
U of Malaya.
M His mainn
Transmission and Distributtion Systems " Innternational Reviiew of
research intereest is in distribuution automationn
Electrical Engineering (IRE EE), Vol. 5. n. 3,
3 pp. 1172-1182, June
area and power system
m protection.
2010 (Partt B).
[12] P. M. An nderson and M. Mirheydar, "An adaptive methood for J. A. Laghari received
r his B. Engg
E in Electricall
setting und derfrequency load shedding relay ys," IEEE Transactions engineering inn 2007 from BUET B Khuzdar,,
on Power Systems, vol. 7, pp.p 647-655, 1992 2. Pakistan and joined as lectu urer in Quaid-e--
[13] F. Shokoooh, J. J. Dai, S. S Shokooh, J. Tastet,
T H. Castrro, T. Awam Univeersity of Engg g Science andd
Khandelwal and G. Donnerr, "Intelligent Loaad Shedding," Inddustry Technology, QUEST, Naw wabshah, Sindh,,
Applicatioons Magazine, IEE EE, vol. 17, pp. 44-53,
4 2011.
Pakistan in 20008. He obtained his
h M. Eng. from m
[14] J. Jung, L. Chen-Ching, S. L. Tanimoto and V. Vittal, "Adapptation
University of Malaya, Malayssia in Electricall
in load shhedding under vu ulnerable operatiing conditions," IEEE
Engineering fieeld in 2012. He is currently PhD D
Transactio ons on Power Sysstems, vol. 17, pp. 1199-1205, 20002.
[15] P. Wang anda R. Billinton, "Optimum load--shedding techniqque to Student in Universityy of Malaya, Malaysia. His main research
r interestss
reduce the total customeer interruption cost c in a distribbution are intelligent poweer system controol, power systeem optimization,,
system," Generation, Trransmission and d Distribution, IEE islan
nding operation inn distributed geneeration and smart grid.
Proceeding gs-, vol. 147, pp. 51-56, 2000. Ab Halim Ab bu Bakar receivved his B.Sc. inn
[16] D. Hazarik ka and A. K. Sinha, "Method for optimal load sheedding Electrical E
Engineering in 1976 from m
in case of generation deficiiency in a power system," Internaational Southampton UUniversity UK and a M. Eng andd
Journal off Electrical Poweer & Energyy Systems, vol. 20, pp. PhD from University Technology Malaysia inn
411-420, 1998.
1 1996 and 20003. He has 30 years of utilityy
[17] W. P. Luaan, M. R. Irving anda J. S. Daniel, "Genetic algorithhm for experience iin Malaysia before joiningg
supply resstoration and optimal load shedd ding in power syystem academia. Currrently he is a Lecturer in thee
distribution networks," Generation, Transmission and Department of Electricall Engineering,,
Distributioon, IEE Proceedin ngs-, vol. 149, pp
p. 145-151, 2002.. Univ
versity of Malaya, Malaysia.
[18] P. Mahat, C. Zhe and B. B Bak-Jensen, "Underfrequency Load
Shedding for an Islanded Distribution Sysstem With Distriibuted M. Karimi reeceived his B. EngE from Islamicc
Generatorss," IEEE Transacctions on Power Delivery, vol. 25, pp. Azad Universsity, Iran and M. Eng. from m
911-918, 2010.
2 University of Malaya, Malaysia in Electricall
[19] E. E. Aponte and J. K. Neelson, "Time optiimal load sheddinng for Engineering field in 20002 and 2011,,
distributedd power system ms," IEEE Transactions on Power P respectively. He
H is currently PhD Student inn
Systems, vol.
v 21, pp. 269-27 77, 2006. University off Malaya, Malaaysia. His mainn
[20] D. Prasetijjo, W. R. Lachs and D. Sutanto, "A new load sheedding research intereest is in Distribbuted Generationn
scheme fo or limiting und derfrequency," IE EEE Transactionns on and Electric Poower System Stabbility.
Power Sysstems,, vol. 9, pp. 1371-1378, 1994 4.
[21] A. A. Giirgis and S. Maathure, "Application of active power p
sensitivity to frequency and d voltage variatio
ons on load sheddding,"
Electric Po ower Systems Ressearch, vol. 80, pp.
p 306-310, 20100.
[22] V. V. Terzzija, "Adaptive underfrequency
u lo
oad shedding bassed on
the magnittude of the disturrbance estimation n," IEEE Transactions
on Power Systems, vol. 21, pp. 1260-1266, 2006. 2
[23] A. A. Moh hd Zin, H. Mohd d Hafiz and M. S. Aziz, "A review of
under-freqquency load sheedding scheme on o TNB system m," in

Copyright 20012 Praise Worth

hy Prize S.r.l. - All rights reservedd Internationnal Review of Eleectrical Engineerring, Vol. 7, N. 4

5000