Sei sulla pagina 1di 8

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 19, NO.

2, APRIL 2004

745

An Islanding Detection Method for Distributed


Generations Using Voltage Unbalance and
Total Harmonic Distortion of Current
Sung-Il Jang, Student Member, IEEE, and Kwang-Ho Kim, Member, IEEE

AbstractDistributed generation (DG) units are rapidly


increasing and most of them are interconnected with distribution
network to supply power into the network as well as local loads.
Islanding operations of DG usually occur when power supply
from the main utility is interrupted due to several reasons but the
DG keeps supplying power into the distribution networks. These
kinds of islanding conditions cause negative impacts on protection,
operation, and management of distribution systems; therefore, it is
necessary to effectively detect the islanding conditions and swiftly
disconnect DG from distribution network. Generally, if there are
large changes in loading for DG after loss of the main power
supply, then islanding conditions are easily detected by monitoring
several parameters: voltage magnitude, phase displacement, and
frequency change. However, in case of small changes in loading for
DG, the conventional methods have some difficulty in detecting
such a particular islanding condition. This paper presents a new
islanding detection algorithm for DG effectively working in the
most of DG loading conditions. This paper proposes two new
parameters for detecting islanding operation of DG: voltage
unbalance and total harmonic distortion of current. The proposed
method effectively combines the conventional parameters with the
newly proposed parameters for detecting the islanding conditions.
The proposed methods were verified using the radial distribution
network of IEEE 34 bus.
Index TermsDistributed generations (DGs), islanding detection.

I. INTRODUCTION

HESE DAYS, electric power utilities are concerned with


distributed generators including photovoltaic, wind farm,
fuel cells, micro-sized turbine, and internal combustion enginegenerators as many good alternatives to solve environmental
problems and to cope with rising energy prices and power plant
construction costs. Distributed generation (DG) may make a
contribution to improve quality of power, minimize peak loads
and eliminate the need for reserve margin [1], [2]. Most DGs
may be connected in parallel and supply power into power grids
as well as local loads. Therefore, DG must be operated in such
an inherently safe manner that DG should supply the generated

Manuscript received November 6, 2002. This work was supported by the


Korea Ministry of Science and Technology and by the Korea Science and Engineering Foundation through the ERC program.
S.-I. Jang is with the Department of Electrical and Computer Engineering,
Kangwon National University, Chuncheon 200-701, Korea (e-mail: sijang@
kwnu.kangwon.ac.kr).
K.-H. Kim is with the Department of Electrical and Computer Engineering,
Kangwon National University, Chuncheon 200-701, Korea and Next-Generation Power Technology Center, Yongin 449-728, Korea (e-mail: khokim
@kangwon.ac.kr).
Digital Object Identifier 10.1109/TPWRD.2003.822964

power to the network loads only if the utility power supply is


present. If DG is feeding the power to the networks without
the utility supply, then it produces several negative impacts on
utility power system and the DG itself, such as the safety hazards to utility personnel and the public, the quality problems of
electric service to the utility customers, and serious damages to
the DG if utility power is wrongly restored [2], [3]. Therefore,
during the interruptions of utility power, the connected DG must
detect the loss of utility power and disconnect itself from the
power grid as soon as possible.
Until now, lots of detection methods for power islanding
operation of DG have been proposed. The proposed detection
method can be classified into two groups. The active techniques,
such as the system fault-level monitoring method [4] and the
reactive export error-detection method [5] (directly regulate the
DG output, measure the variation of the DG output, and decide
whether the DG is operating in islanding condition or in normal
condition). Despite of their effectiveness in detecting island
operation of DG, these active techniques need to continuously
vary the DG output and may negatively affect the operations of
the DG and the utility system. Others detection methods can be
called passive techniques since they detect islanding operation
of DG by monitoring the system parameters: voltage magnitude
[6], phase displacement [7], the rate of change in frequency [4],
and impedance monitoring [8]. Even though they are unlikely
to influence the operating strategies and management of utility
power system, if there are little changes in the DG loadings
after islanding, these methods have difficulties in finding out
the islanding operation since the monitoring parameters do not
change enough to detect these islanding conditions.
In general, after loss of the main source, the DG has to take
charge of the remaining network and the connected loads;
therefore, the loading condition of the DG is suddenly changed
after islanding. Since the distribution networks generally
include single-phase loads, it will be highly possible that
the islanding changes the load balance of DG. Additionally,
different loading condition might result in different harmonic
current in the network since the amount and configuration
of load are changed. Therefore, this paper proposes two new
monitoring parameters for detecting the islanding operations
of DG: voltage unbalance and total harmonic distortion (THD)
of current. The proposed method utilizes not only two new
monitoring parameters but also voltage magnitude used in
the conventional islanding detection techniques. The method
monitors changes in three parameters and diagnoses the operating conditions of DG. The proposed method enhances the

0885-8977/04$20.00 2004 IEEE

746

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 19, NO. 2, APRIL 2004

TABLE I
MODIFIED IEEE 34BUS NETWORK LOADS DATA

Fig. 1.

Structure of distributed generator system model.

performance of the conventional methods and also overcome


their demerits. We tested the proposed method using the radial
distribution network of IEEE 34-bus model. The test results
showed that the method correctly detects the islanding operation, even in case of little change in DG loadings after islanding
and does not mal-operate for normal variation of network load,
where conventional detection methods may become confused
during island operation.
II. MODELING OF DISTRIBUTION NETWORK WITH DG
This paper, by using PSCAD/EMTDC, models the IEEE
34-bus distribution network interconnected with DGs, simulates various islanding operation conditions and normal load
variations, and verifies the efficiency of the proposed method.
A. Modeling of DG System
In this study, we assume that the three-phase DG system consists of a dc source, an inverter, a filter, a transformer, and a controller. This kind of dc-ac inverting DG structure is easily found
and commonly used in photovoltaic systems, fuel cell systems,
micro-turbines, and modern wind power systems. Fig. 1 shows
the DG system modeled in this paper. The controller plays a key
role in operating, managing, and protecting the DG system. In
this model, the controller measures the secondary current and
voltage of transformer and monitors the condition of the DG
system. And, it also equalizes the voltage magnitude, phase,
and frequency of the DG output power with the distribution network using the pulsewidth modulation (PWM) method and the
phase-locked loop (PLL) method.
B. Modeling of Distribution Network
We used an IEEE 34-bus distribution networks in [10], which
consist of single-phase and three-phase laterals connected with
R-L bus loads and distributed loads. For the purpose of the realistic islanding operations of DG, we modified R-L network load
on the IEEE 34-bus (see Table I). There are also two balanced
three-phase capacitor banks at bus 844 and 848, respectively.
Fig. 2 shows the circuit diagram of the distribution network
connected with several DGs. In this paper, DGs are installed at
different locations in order to simulate various islanding conditions and network load varying conditions. In simulations, we
assumed that the islanding operations of DG occur after loss of
main source power by operation of the upstream circuit breakers
or switches.

III. CHARACTERISTICS OF ISLANDING OPERATION


This section presents the characteristics of two typical cases
of islanding operation conditions: abrupt changes in the loading
for DG (case I) after the circuit breaker between Bus 834 and
Bus 860 opens and little changes in the loading for DG (case II)
after circuit breaker 1 opens, as shown in Fig. 2. The islanding
condition of case I is easily detected since the conventional monitoring parameters, such as voltage magnitude, phase displacement, and frequency, show quite different value after islanding.
However, even though it rarely happened, in the latter conditions, these monitoring parameters do not vary enough to detect
the loss of the main power.
A. The Large Variation in the Loading for DG After Islanding
In the distribution network model of Fig. 2, we locate the DG
at Bus 840 and open the circuit breaker between bus 834 and
bus 860. Then, the loading for the DG abruptly increases since
the DG takes charge of all the loads within the power island.
Immediately after islanding, due to the large variation in the
loading for the DG, there are sudden changes in the voltage magnitude, frequency, and phase displacement of the DG as shown
in Fig. 3(a)(c). Therefore, there are no problems in detecting
this kind of power islanding condition using the conventional
monitoring parameters.
B. The Small Variation in the Loading for DG After Islanding
The power islanding condition having little changes in the
loading for DG is simulated by locating the DG at bus 832 and
then opening the circuit breaker 1. As we expect, after islanding,
voltage magnitude, phase, and frequency do not change enough
to detect the islanding operation as shown in Fig. 4(a)(c). In

JANG AND KIM: AN ISLANDING DETECTION METHOD FOR DGs

747

Fig. 3. Conventional monitoring parameters for large changes in the loading


for DG after islanding. (a) Voltage magnitude. (b) Phase. (c) Frequency.

Fig. 2.

Distribution network model with DG.

Fig. 5, we depict the one-cycle average of the THD of the current and voltage unbalance in this case. Fig. 5 shows the large
variation in the THD of the current and voltage unbalance after
islanding; therefore, these parameters can be effectively utilized
to detect the islanding operations.

Fig. 4. Conventional monitoring parameters for little changes in the loading


for DG after islanding. (a) Voltage magnitude. (b) Phase. (c) Frequency.

IV. MONITORING PARAMETERS TO DETECT ISLANDING


OPERATIONS
In this study, we propose two new monitoring parameters for
detecting the islanding operations of DG: voltage unbalance and
the THD of the current. Additionally, the voltage magnitude is
also utilized as one of monitoring parameters, which is conventionally used for power islanding protection [6].
A. Voltage Unbalance Variation
Generally, even though the loading for DG has little changed
after the loss of main source, due to the topology changes of the
networks and the load, the voltage unbalance varies. So, if we
keep monitoring the unbalance of three-phase output voltage of
the DG, then it is possible to effectively detect an islanding oper-

Fig. 5. Newly proposed monitoring parameters for little changes in the loading
for DG after islanding. (a) THD of phase A current. (b) Voltage imbalance.

ation of DG. In order to do this, we define the voltage unbalance


at the monitoring time by (1)
(1)

748

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 19, NO. 2, APRIL 2004

where
and
mean the magnitude of negative and positive
sequence of voltage at , respectively.
This study defines the one-cycle average of voltage unbalance
as (2), and also defines the voltage unbalance variation as (3),
which measuring how much the monitored voltage unbalance
deviates from the steady state and normal loading conditions.

defined in (8), this method discards the value and goes to the
next time step
(8)
where

is set to be 1/4 cycle (4.17 ms).

C. Three-Phase Voltage Magnitude Variation


(2)
(3)
where
is the sampling number of one-cycle, is the monis the VU reference value initially
itoring time, and
set for the steady-state and normal loading conditions. After
is initially set, if
remains within
through
for one-cycle, the
is updated by the
in order to adapt to the normal load variation. And
also, to avoid inaccurate decisions during too short transient
above 0.05%
state, if there are abrupt changes in
during the 1/4 cycle, as defined in (4), this method discards the
value and goes to the next time step
(4)
where

is set to be 1/4 cycle (4.17 ms).

B. THD Variation of the Current


The changes in the loading for DG due to loss of main power
source obviously result in variations on the harmonics of the current. So, we propose the THD of the current as one of the monitoring parameters. The total harmonic distortion of the current
at the monitoring time is defined as (5)

(5)
where is the rms of the harmonic components and is the
rms value of fundamental component. In this paper, we set to
the 31th component.
over one-cycle is defined as follows:
The average of

In this paper, the proposed method also monitors the variation of the voltage magnitude defined in (9) as the three-phase
average rms value of line to line voltage [11]

(9)
where is a sampling number of one cycle, and , , and
are instantaneous voltages of phase A, B, and C, respectively,
and is the monitoring time.
V. PROPOSED ISLANDING DETECTION METHOD
Each of three monitoring parameters described in Section IV
can be used alone for detecting islanding operation. However,
any of them does not guarantee that it, in any kind of network
conditions, would be able to detect islanding and not to mal-operate for normal load variations. So, the hybrid decision approach, which monitors three parameters and makes final decision, would be suitable for power islanding detection.
of
At every sampling time, this method calculates
of a phase A current, and
three phase voltages,
of a line to line voltage. In the first, it checks whether the
is lower than the preset value (0.5 pu). If the value is below
the preset value, then this method immediately makes a trip
signal for islanding protection. In this manner, we can easily
and rapidly detect the power islanding operation in the case
of the large variation in the loading for DG. Otherwise, the
and
method checks the other monitoring parameters:
. If
and
have been satisfying the following logic equation in (10) during more than one cycle, the
method decides it as an islanding condition in the case of little
variation in the loading for DG and makes a trip signal +

(6)
where is the sampling number of one cycle.
We also define the THD variation as (7), which measures how
much the monitored THD at deviates from the steady-state and
normal loading conditions
(7)
where
is the THD reference value for the steady state
is initially set,
and normal loading conditions. After
remains within
through
for one cycle,
if
the
is updated by the
. And if there are
above 0.1% during 1/4 cycle, as
abrupt changes in

(10)
Fig. 6 represents the detection flow diagram of the proposed
islanding protection method. The preset criteria for the monitoring parameters were selected through the various simulations
for different islanding operation conditions and also normal load
variation conditions on the distribution network model shown
in Fig. 2 using PSCAD/EMTDC. We conducted 40 case simulations in total by moving a 0.14 MVA DG with a 0.125-MVA
local load of 0.88 power factor lagging, to different locations.
VI. TEST RESULTS
The test of the proposed method was conducted with several network conditions, which are not included in the simula-

JANG AND KIM: AN ISLANDING DETECTION METHOD FOR DGs

749

Fig. 8. Conventional parameter values for islanding operation with large


variation in DR loading. (a) Phase. (b) Frequency.

Fig. 6.

Flow diagram of the proposed islanding detection method.

Fig. 7. Result for islanding operation with large variation in DG loading. (a)
Voltage magnitude. (b) Detection trip signal.

tion cases used for deriving the detection criteria. We tested the
method not only with islanding operation conditions of DG but
also with normal network load variation in order to see its selectivity.
A. Islanding Operation Conditions
We made a typical islanding condition of the DG connected
at Bus 840 in Fig. 2 by opening the circuit breaker between Bus
858 and 834 at 53 ms. Due to a large variation in DG loadings, as shown in Fig. 7, there is a sudden change in the line to
line voltage. So, the islanding operation condition is easily and
rapidly detected by voltage parameter. In the figure, the thin line
represents parameter value and the bold line shows trip signal
for islanding detection. The trip signal for detecting islanding
operation is made at 75 ms, which is 53 ms after islanding.
Fig. 8 also shows the other parameters, which have been used in
the conventional islanding protections. The results show that the
conventional approaches are also useful for detecting islanding
operation in case of large variations in loading for DG.

Fig. 9. Result for islanding operation with small variation in the loading of
DG located in Bus 858. (a) Voltage magnitude. (b) THD of phase A current. (c)
Detection trip signal.

We also tested two kinds of particular cases in Fig. 2. One


case (A) is for islanding operation of the DG connected at Bus
858 by opening the circuit breaker 2. The other case (B) is that
the DG connected at Bus 840 is in islanding condition by disconnecting the circuit breaker between Bus 840 and 836. Case
B was simulated without the capacitor banks of phase A located
in Bus 844 and Bus 848. Due to the small variation in DG loadings, if we apply the previously used criteria of the conventional
parameters, the islanding operations are not able to be detected
since the voltage magnitude, phase displacement, and frequency
do not change enough, as shown in Figs. 912. However, unlike other parameters, the changes in the THD of the current
and voltage unbalance are large enough to detect the islanding
operation. The proposed method provides the trip signal for islanding detection at 129 ms in the case A, 126 ms in the case
B, respectively, as shown in Figs. 9 and 11. This result shows
that the propose method can outperform the conventional approaches, which have some difficulties to detect the islanding
operation conditions with small changes in DG loadings.

750

Fig. 10. Conventional parameter values for islanding operation with small
variation in the loading of DG located in Bus 858. (a) Phase. (b) Frequency.

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 19, NO. 2, APRIL 2004

Fig. 12. Conventional parameter values for islanding operation with small
variation in the loading of DG located in Bus 840. (a) Phase. (b) Frequency.

Fig. 13. Conventional parameter values for staring the induction motor. (a)
Phase. (b) Frequency.

Fig. 11. Result for islanding operation with small variation in the loading of
DG located in Bus 840. (a) Voltage magnitude. (b) THD of phase A current.
(c) Voltage imbalance. (d) Detection trip signal.

B. Normal Load Variation Conditions


We also tested the proposed method with the normal load
variation conditions, which are sometimes confused with islanding conditions, to show that the method does not maloperate in nonislanding conditions. The test result shows that the
proposed method has better selectivity than the conventional approaches.
Induction Motor Start: For this case, we located and operated the DG at Bus 840. After that, we started an induction motor
rated 500 HP connected at Bus 834. After starting the induction motor, the phase and voltage magnitude vary remarkably;
therefore, the conventional approaches may confuse this case
with the islanding conditions. However, the proposed method
does not mal-operate in this nonislanding case since the voltage
unbalance does not much change except for the short transient
period. This simulation result is shown in Figs. 13 and 14.
Switching-In a Capacitor Bank: For this case, we located
and operated the DG at Bus 840. After then, we switched-in a

Fig. 14. Test result for starting the induction motor. (a) Voltage magnitude.
(b) THD of phase A current. (c) Voltage imbalance. (d) Detection trip signal.

capacitor bank of 300 kvar at Bus 844. In this case, the conventional method monitoring the changes in frequency confused it
with islanding conditions due to the natural frequency oscillations caused by the switching-in a capacitor bank as shown in

JANG AND KIM: AN ISLANDING DETECTION METHOD FOR DGs

751

Fig. 15. Conventional parameter values for switching-in a capacitor bank.


(a) Phase. (b) Frequency.

Fig. 17. Test result for removing two single-phase loads. (a) Voltage
magnitude. (b) THD of phase A current. (c) Voltage imbalance. (d) Detection
trip signal.

Fig. 16. Test result for switching-in a capacitor bank. (a) Voltage magnitude.
(b) THD of phase A current. (c) Voltage imbalance. (d) Detection trip signal.

Fig. 15. However, the proposed method does not mal-operate


for this normal switching, as shown in Fig. 16.
Unbalanced Network Load Variations: We simulated two
cases for unbalanced load variations. A case for load reduction
was simulated by disconnecting two single-phase loads rated 62
and 4.2 kVA on phase A at Bus 820 and Bus 842, respectively.
For load addition, two single-phase loads rated 8 and 10 kVA
were connected at phase A and phase B in Bus 836, respectively.
In this case, the DG was connected at Bus 840. Due to the unbalanced load variation and the switching operation, these test
cases for normal load changes show similar behaviors with islanding. However, the simulation results show that the proposed
method does not mal-operate on these normal load variations.
The test results would be understood that, due to the large balanced network load, comparatively small load variations barely
affect the overall voltage unbalance value. Also, the THD of the
current does not change enough to trip except for the short transient period and the variation during the short transient period

Fig. 18.

Test result for connecting two single-phase loads.

are discarded for reliable decision in this paper. These simulation results are shown in Figs. 17 and 18.
VII. CONCLUSION
This paper introduced and proposed the THD of the current
and voltage unbalance of the DG terminal output as new monitoring parameters for power islanding detection. This paper also
presented a simple and novel detection algorithm, which effectively combines the detection results of the conventional param-

752

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 19, NO. 2, APRIL 2004

eter, voltage magnitude, and two newly proposed parameters.


We tested the proposed method using several distribution network conditions including not only islanding operation conditions, but also nonislanding conditions of normal network load
variations. The test results showed that the proposed parameters and algorithm are capable of correctly detecting the islanding operation not affected by variation of DG loading and
also have a good selectivity for islanding conditions and nonislanding conditions. Therefore, the results verify the proposed
method can be efficiently implemented in the actual distributed
generators and able to improve the performance of them. So, the
proposed method is expected to be an effective islanding detection approach for implementation in industrial fields.
REFERENCES
[1] P. P. Barker and R. W. de Mello, Determining the Impact of Distributed
Generation on Power Systems: Part1-Radial Distribution Systems,
IEEE, 0780364201/00, 2000.
[2] T. Ackermann, G. Andersson, and L. Soder, Electricity Market
Regulations and their Impact on Distributed Generatiron, IEEE,
078035902-X/00, 2000.
[3] O. Usta and M. A. Refern, Protection of Dispersed Storage and Generation Units Against Islanding, IEEE, 0780317726/94, 1994.
[4] C. B. Cooper, Standby generation Problems and prospective gains
from parallel running, in Proc. Power System Protection, Singapore,
1989.
[5] J. W. Marin, Loss of mains protection, in Proc. ERA Conf. Circuit
Protection Industrial Commercial Inst., London, U.K., 1990.
[6] K. Takigawa and H. Kobayashi, Development of Compact and Reliable
Protective Control Unit for Grid Connected Small Residential PV Systems, CH33654/94/00001081, 1994.

[7] Schaltanlagen Electronik Gerate GMBH & Co., Generator/Mains Monitor GW2, DEG GMBH Publication GW2/E/810.
[8] P. OKane and B. Fox, Loss of mains detection for embedded generation by system impedance monitoring, in Inst. Elect. Eng. Conf. Developments in Power System Protection, Aberdeen, U.K., 1990.
[9] Recommendation for the Connection of Private Generating Plant to the
Electricity Boards Distribution System, Engineering Recommendation
G.59, 1985.
[10] W. H. Kersting, Radial distribution test feeders, IEEE Trans. Power
Syst., vol. 6, pp. 975985, Aug. 1991.
[11] M. H. Rashid, Power Electronics Circuits, Devices, and Applications. Englewood Cliffs, NJ: Prentice-Hall, 1993.

Sung-Il Jang (S01) received the B.S. and M.S. degrees in electrical engineering
in 1996 and 1998, respectively, from Kangwon National University, Chuncheon,
Korea, where he is currently working toward the Ph.D. degree in the Department
of Electrical Engineering.
His research interests include distributed generation interfaces with power
systems and power quality analysis.

Kwang-Ho Kim (M95) received the B.S., M.S., and Ph.D. degrees in electrical
engineering from Seoul National University, Seoul, Korea, in 1988, 1990, and
1994, respectively.
Since 1995, he has been with the Department of Electrical Engineering,
Kangwon National University, Chuncheon, Korea, where he is currently an
Associate Professor. He has also been with Next-Generation Power Technology
Center, Yongin, Korea, since 2001. His research interests include distributed
generation interfaces with power systems and power quality analysis.

Potrebbero piacerti anche