Electric Power Systems Research 77 (2007) 170–180

Characteristics of vector surge relays for distributed

synchronous generator protection
Walmir Freitas a,∗ , Wilsun Xu b , Zhenyu Huang c , Jose C.M. Vieira a
a Department of Electrical Energy Systems, State University of Campinas, 13083-852 Campinas, Sao Paulo, Brazil
b Department of Electrical and Computer Engineering, University of Alberta, Edmonton, Alta., Canada T6G 2V4
c Energy Science and Technology Division, Pacific Northwest National Laboratory, 99352 Richland, WA, USA

Received 12 September 2005; received in revised form 17 February 2006; accepted 17 February 2006
Available online 17 April 2006

Abstract
This work presents a detailed investigation on the performance characteristics of vector surge relays used to detect islanding of distributed
synchronous generators. A detection time versus active power imbalance curve is proposed to evaluate the relay performance. Computer simulations
are used to obtain the performance curves. The concept of critical active power imbalance is introduced based on these curves. Main factors affecting
the performance of the relays are analyzed. The factors investigated are voltage-dependent loads, load power factor, inertia constant of the generator,
generator excitation system control mode, feeder length and R/X ratio as well as multi-distributed generators. The results are a useful guideline to
evaluate the effectiveness of anti-island schemes based on vector surge relays for distributed generation applications.
© 2006 Elsevier B.V. All rights reserved.

Keywords: Distributed generation; Islanding detection; Synchronous generators; Vector surge relays; Vector shift relays

1. Introduction
Distributed generation has recently gained a lot of momentum
in the power industry due to market deregulation, technologi-
cal advances, governmental incentives and environment impact
concerns [1–3]. An important requirement for the connection of
synchronous generators to distribution networks is the protec-
tion system capability of islanding detection. Islanding occurs
when a portion of the distribution system becomes electrically
isolated from the remainder of the power system, yet continues
to be energized by distributed generators. This is also known as
loss of mains or loss of grid [3]. Failure to trip islanded gen-
erators can lead to a number of problems to the generator and
the connected loads. The current industry practice is to discon-
nect all distributed generators immediately after the occurrence
of islands [4,5]. Typically, a distributed generator should be dis-
connected within 200–400 ms after loss of main supply. In order
to achieve such a goal, each distributed synchronous genera-
tor must be equipped with an islanding detection device. The
∗ Corresponding author. Tel.: +55 19 37883740; fax: +55 19 32891395.
E-mail address: walmir@dsee.fee.unicamp.br (W. Freitas).
common devices used for this purpose are the modified ver-
sions of the under/over voltage and under/over frequency relays.
Among them, devices based on variations of frequency have
been recognized as the most reliable option by the industry so
far. Representative examples of such relays are the rate of change
of frequency relay (ROCOF) and the vector surge relay (VSR),
which is also known as vector shift or voltage jump relay [1–3].
However, it has been well recognized that such relays tend to
fail when the mismatch or imbalance between the generation
and the load in the islanded system is small. As more and more
distributed generators are added to utility systems, it has become
clear that a good understanding on the operating characteristics
of these relays is important. The results can be very useful for
utility engineers to evaluate the reliability and robustness of a
given distributed generation anti-islanding scheme.
The objective of this paper is to present our investigation
results on this subject. The concepts of detection time versus
active power imbalance curve and critical active power imbal-
ance are proposed to characterize the relay performance. Since
vector surge relays have been widely used by the industry, it is
selected as the study subject in this paper. However, the proposed
method can be applied to other types of anti-islanding relays as
well.

0378-7796/$ – see front matter © 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.epsr.2006.02.011
 W. Freitas et al. / Electric Power Systems Research 77 (2007) 170–180
Fig. 1. Equivalent circuit of a synchronous generator in parallel with utility.
This paper is organized as follows. The principle of vector
surges relays is explained in Section 2. The detection time versus
active power imbalance curves and the concept of critical active
power imbalance are discussed in Section 3. The main factors
affecting the performance of vector surge relays are investigated
in Section 4. The factors investigated are voltage-dependent
loads, load power factor, inertia constant of the generator, gener-
ator excitation system control mode, feeder length and R/X ratio
as well as multi-distributed generators. Finally, in Section 5, the
main conclusions are analyzed.
2. Principle of vector surge relays
When a synchronous generator is operating in parallel with
a distribution network, as depicted in Fig. 1, there is a voltage
drop
V between the terminal voltage VT and the generator
internal voltage EI due to the generator current ISG passing
through the generator reactance Xd . Consequently, there is a
displacement angle between the terminal voltage and the gen-
erator internal voltage, whose phasor diagram is presented in
Fig. 2(a). In Fig. 1, if the circuit breaker CB opens, due to a fault
for example, the system composed by the generator and the load
L becomes islanded. At this instant, the synchronous machine
begins to feed a larger load (or smaller) because the current ISYS
provided by (or injected into) the power grid is abruptly inter-
rupted. Consequently, the angular difference between VT and EI
is suddenly increased (or decreased) and the terminal voltage
phasor changes its direction, as shown in Fig. 2(b). Analyzing
such phenomenon in the time-domain, the instantaneous value
of the terminal voltage jumps to another value and the phase
position changes, as depicted in Fig. 3, where the point A indi-
cates the islanding instant. This behavior of the terminal voltage
is called vector surge or vector shift. It is possible to verify that
Fig. 2. Internal and terminal voltage phasors: (a) before the opening of circuit
breaker and (b) after the opening of circuit breaker.
171
Fig. 3. Voltage vector surge and vector surge relay cycle-by-cycle measure-
ments.
the cycle duration also changes. Vector surge relays are based
on such phenomena [3].
Vector surge relays available in the market measure the dura-
tion time of an electrical cycle and start a new measurement at
each zero rising crossings of the terminal voltage. The current
cycle duration (measured waveform) is compared with the last
one (reference cycle). In an islanding situation, the cycle dura-
tion is either shorter or longer, depending on if there is excess or
deficit of power in the islanded system, as shown in Fig. 3. This
variation of the cycle duration results in a proportional variation
of the terminal voltage angle
θ, which is the input parameter of
vector surge relays. If the variation of the terminal voltage angle
exceeds a pre-determined threshold α, a trip signal is imme-
diately sent to the circuit breaker. Usually, vector surge relays
allow this threshold to be adjusted in the range from 2◦ to 20◦
[3]. Another important characteristic available in these relays is
a block function by minimum terminal voltage. If the terminal
voltage drops below an adjustable level threshold Vmin , the trip
signal from the vector surge relay is blocked. This is to avoid, for
example, the actuation of the vector surge relay during generator
start-up or short-circuits.
3. Determination and analysis of the performance
curves
In this section, the determination and analysis of the detection
time versus active power imbalance curves, i.e. the performance
curves, are discussed. In this paper, the performance curves
are obtained by using repeated dynamic simulations. The pre-
islanding active power imbalance is gradually varied from 1 to
0 pu, referred to the MVA rating of the generator, by chang-
ing the generation–load profile in the islanded system. For each
active power imbalance level, dynamic simulation is conducted,
the detection time is determined once that the relay activation
criterion is met and then the curves are plotted.
In this work, the network components were represented by
three-phase models. Distribution feeders were modeled as series
RL impedances. Transformers were modeled using T circuit. The
synchronous generators were represented by a sixth-order three-
phase model in the dq rotor reference frame [6]. The generators
were considered equipped with an automatic voltage regulator,
which was represented by the IEEE-Type 1 model. The mechan-
ical power was considered constant, i.e. the primer mover and
172 W. Freitas et al. / Electric Power Systems Research 77 (2007) 170–180

Fig. 4. Algorithm implemented to represent vector surge relays.

governor effects were neglected, because the simulation interval

is short (0.5 s). Since load characteristics can have a significant
impact on the power imbalance after islanding, the exponential
voltage-dependent load model presented in ref. [7] was used.
The vector surge relays were simulated as follows. The gen-
erator terminal voltage angle θ is determined in each integration
step. Moreover, a comparative terminal voltage angle θ 0 is com-
puted and updated in the beginning of each cycle, i.e. it is updated
cycle-by-cycle. The absolute variation between these two angles,

θ = ||θ − θ 0 ||, is calculated in each integration step and com-
pared with the angle threshold α. Additionally, the rms value
of the terminal voltage is also determined in each integration
step. If the angle variation
θ is larger than the relay setting
α and the magnitude of the terminal voltage is larger than the
adjusted minimum voltage Vmin , the vector surge relay imme-
diately sends a trip signal to the generator circuit breaker. The
algorithm can be better understood through Fig. 4. In this fig-
ure, the calculation process of the angle variation considering
two complete electrical cycles and adopting an integration step
equal to 5.56 ms (third part of one electrical cycle in 60 Hz sys-
tems) is shown. In this figure, it is considered that the islanding
situation occurred at the beginning of the first cycle.
Fig. 5 shows the single-line diagram of the network used
in this paper. It comprises a 132 kV, 60 Hz, subtransmission
system with short-circuit level of 1500 MVA, represented by
a Thévenin equivalent (Sub), which feeds a 33 kV distribu-
tion system through a 132 kV/33 kV,
/Yg, transformer. In this
system, there is one 30 MW synchronous generator (SG) con-
nected at bus 5, which is connected to the network through one
33 kV/0.69 kV,
/Yg, transformer.
In all presented cases, the islanding situation was simulated
by opening the circuit breaker CB at bus 2 at t = 0.25 s, which
remained open during the rest of the simulation. Therefore, the
Fig. 6. Performance curves for different relay settings.
initial active power imbalance in the islanded system was equal
to the active power provided by the substation at the island-
ing moment. The total simulation time was 0.75 s. Therefore,
if the vector surge relay VSR installed at bus 5 did not detect
the islanding condition until 0.5 s after opening of the circuit
breaker CB, it was considered that the device was inoperative
for this case. Different power imbalance scenarios were created
by varying either the generator output or the total system load.
Several detection time versus active power imbalance curves
for the test system and different relay settings are shown in
Fig. 6. These curves were obtained by varying the active power
injected by the generator from 0 to 30 MW and keeping the
loads constant, as shown in Fig. 5. Consequently, after opening
of the circuit breaker CB, there is a deficit of active power in the
islanded system. For each power imbalance level, dynamic sim-
ulation was conducted to determine the detection time. There is
a family of curves, each one corresponding to a different relay
setting α. It can be observed that when the power imbalance
decreases the detection time increases. The increase is almost

Fig. 5. Single-line diagram of the test system 1. Fig. 7. Concept of critical active power imbalance level.
 W. Freitas et al. / Electric Power Systems Research 77 (2007) 170–180 173

exponential when the power mismatch is small. This is reason-

able because it takes longer for the relay to detect a small power
imbalance.
Such curves can be employed to assess the performance of
vector surge relays systematically. As an example, assume that
the required detection time is 200 ms and the relay setting is
5◦ . In this case, one can draw a horizontal line of 200 ms as
shown in Fig. 7. The active power imbalance level in the cross-
ing point between the horizontal line and the relay curve is the
minimum active power imbalance required for the vector surge
relay to trip the generator within 200 ms. In this case, the active
power imbalance level at this point is equal to 27%. Thus, if the
islanded system has a power imbalance level higher than 27%,
the relay would take less than 200 ms to detect the islanding con-
dition. So the relay can be used with confidence. On the other
hand, the relay would take longer than 200 ms to operate if the
power imbalanced level is lower than 27%. Consequently, the
relay is not suitable for such cases. In this paper, such an active
power threshold is called critical active power imbalance level
or simply critical power imbalance.

4. Performance characteristics of vector surge relays

In this section, the main factors affecting the performance

characteristics of vector surge relays are investigated using
the performance curves. The factors investigated are voltage-
dependent loads, load power factor, inertia constant of the gen-
erator, generator excitation system control mode, feeder length
and R/X ratio as well as multi-distributed generators. Excepting
the cases where the impact of voltage-dependent loads and multi-
generators are analyzed, the active components of the loads
were represented by constant current model and the reactive
components were represented by constant impedance model as
recommended in ref. [7].
Fig. 8. Relay performance curves as affected by voltage-dependent loads—
relays setting = 10◦ : (a) deficit of reactive power and (b) excess of reactive
4.1. Voltage-dependent loads power.

Since the nodal voltages change after islanding, voltage-

dependent loads affect the dynamic behavior of the active
power imbalance in the islanded system. Thus, typical constant
impedance, constant current and constant power loads are ana-
lyzed in this section. Moreover, the impact of voltage-dependent
loads on the performance of vector surge relays will depend on
if there is deficit or excess of active and reactive power in the
islanded system. Based on these facts, it is necessary to analyze
the following four different situations:
• Case (a): There is deficit of active and reactive power in the
islanded system.
• Case (b): There is excess of active power and deficit of reactive
power in the islanded system.
• Case (c): There is excess of active and reactive power in the
islanded system.
• Case (d): There is deficit of active power and excess of reactive
power in the islanded system.
The performance curves for different types of loads and
deficit of reactive power in the islanded system are shown in
Fig. 8(a) considering that the relay setting is 10◦ . The cases of
active power deficit were simulated maintaining constant the
loads and varying the generation from 0 to 30 MW. Whereas the
cases of active power excess were obtained keeping constant
the generation at 30 MW and varying the total active power load
from 0 to 30 MW without changing the load power factor. In
these cases, the generator exciter was controlled to keep termi-
nal voltage at 1 pu, as a consequence, there is deficit of reactive
power in the islanded system. As expected, it can be seen that
the type of load has much influence on the performance of the
relays. In this situation, deficit of reactive power, the most con-
servative case, i.e. the case in which the critical active power
imbalance assumes the highest value, occurs when the loads
are constant impedance type and there is deficit of active power
in the islanded system. On the other hand, the most optimistic
case, i.e. the case in which the critical active power imbalance
174 W. Freitas et al. / Electric Power Systems Research 77 (2007) 170–180

Fig. 9. Most optimistic cases vs. most conservative cases—case with deficit of
reactive power in the islanded system.

assumes the lowest value, is obtained when the loads are con-
stant impedance type and there is excess of active power in the
islanded system. In the case of constant power loads, there is
no difference between the cases with deficit and excess of active
power. The critical active power imbalances for the most conser-
vative and optimistic cases and different relay settings are shown
in Fig. 9, which were obtained considering that the required
detection time is 200 ms. It can be observed that if the loads
are impedance constant type and the relay setting is 2◦ , then
the critical active power imbalance will vary from 2.5 to 18.0%.
The cases with excess of reactive power in the islanded system
are presented in Fig. 8(b). In these cases, the generator exciter
was controlled to keep the generator working at 0.90 capacitive
power factor. The situations of excess and deficit of active power
were simulated as discussed in the previous cases. The results
are contrary to the previous cases, i.e. the most conservative case
can be related to the situation with excess of active power and
Fig. 10. Dynamic behavior of the active power load during islanding: (a) deficit
the most optimistic case is related to the situation with deficit of reactive power and (b) excess of reactive power.
of active power. The impact of voltage-dependent loads on the
performance of vector surge relays is explained in the sequence.
The dynamic behavior of the total active power load of the 24 MW and the generator is injecting 30 MW into the network.
system shown in Fig. 5 considering different types of loads In this case, the generator exciter is controlled to force the gener-
and sceneries of pre-islanding reactive power imbalance is pre- ator to operate at 0.90 capacitive power factor. Thus, before the
sented in Fig. 10. In this figure, the system becomes islanded islanding, the generator is injecting 14.5 MVAr into the network
at t = 0.25 s. In Fig. 10(a), the total active power load before the and the substation is consuming 3.4 MVAr. Therefore, there is
islanding is 30 MW and the generator is injecting 20 MW into the excess of reactive power in the islanded system. Consequently,
network. In this case, the generator exciter is controlled to keep the nodal voltages increase after the islanding; as a result, in
terminal voltage constant at 1 pu. Thus, the generator is injecting the case of constant current and constant impedance loads, the
7.9 MVAr into the network and the rest of the reactive power con- total active power load in the islanded system increases after the
sumed by the system is supplied by the substation (3.9 MVAr). In islanding. Again, the largest variation can be related to constant
this situation, after the islanding, the nodal voltages decrease due impedance loads. The impacts of the possible combinations of
to lack of reactive power, although the generator increases the active and reactive power in the islanded system (Cases a–d) on
reactive power injected after islanding. Consequently, in the case the relay performance are discussed below.
of constant current and constant impedance loads, the total active In the case of active and reactive power deficit (Case a),
power load in the islanded system decreases after islanding. The the total active power load decreases after islanding due to
largest variation can be related to constant impedance loads. In nodal voltage reduction and such fact decreases the active power
Fig. 10(b), the total active power load before the islanding is imbalance in the islanded system. Consequently, vector surge
 W. Freitas et al. / Electric Power Systems Research 77 (2007) 170–180 175

relays will take longer to detect the islanding situation. On the

other hand, in the case of active power excess and reactive power
deficit (Case b), the active power load reduction increases the
active power imbalance in the islanded system, so that vector
surge relays will detect the islanding situation faster. For the
same reasons, if there is excess of reactive and active power in
the islanded systems (Case c), the active power load increases
after islanding and the active power imbalance decreases, such
fact makes more difficult to detect the islanding. Finally, if there
is deficit of active power and excess of reactive power (Case d),
the total active power load increases after the islanding, conse-
quently, the active power imbalance also increases and the relay
can detect the islanding situation faster. In the case of constant
power loads, there is no difference if there is deficit or excess
of active and reactive power in the islanded system, since the
active power imbalance will be practically constant, neglecting
losses.
These important results show that utility engineers must ana- Fig. 11. Relay performance curves as affected by load power factor.
lyze the most conservative cases during studies of vector surge
relays, because it is not possible to know if there will be excess 4.3. Inertia constant of generator
or deficit of active and reactive power in the islanded system.
Typically, such cases are: The impact of the machine inertia constant (H) on the relay
performance is investigated in this section by using the test sys-
• There is deficit of active and reactive power in the islanded tem 1. The original machine has an inertia constant equal to
system and loads are constant impedance type. 1.5 s. In Fig. 12, the performance curves were obtained consid-
• There is excess of active and reactive power in the islanded ering the inertia constant equal to 0.5, 1.0 and 1.5 s, respectively,
system and loads are constant impedance type. and the relay setting equal to 10◦ . It can be seen that the inertia
constant has a huge influence on the performance curves. More-
over, the larger the inertia constant is, the larger the critical power
In the next cases, excepting the cases where multi-generators
imbalance level is for the same detection time requirement and
are analyzed, there is deficit of reactive and active power in the
relay setting. Such fact is due to the energy stored in the rota-
islanded system. These cases were obtained by varying the active
tional mass of the machine, which is proportional to the value
power injected by the generator from 0 to 30 MW and keeping
of the inertia constant. Consequently, a generator with a large
the loads constant, as shown in Fig. 5. Moreover, the active power
inertia constant will take more time to cause frequency devia-
loads were represented by constant current models and reactive
tion than a generator with a small inertia constant. The critical
power loads were represented by constant impedance models, as
power imbalances for a 200 ms detection time requirement and
recommended in ref. [7]. Although many other scenarios were
different relay settings and inertia constant values are compared
also simulated, such cases showed to be enough representative
to analyze the main technical factors affecting the performance
of vector surge relays.

4.2. Load power factor

The impact of the load power factor on the performance

curves was also investigated. The power factor of the loads was
varied while the active power was kept constant. The results are
shown in Fig. 11 for relay setting of 5◦ , 10◦ and 15◦ . The figure
reveals that the load power factor has little influence on the per-
formance of vector surge relays. The slight difference among the
curves can be explained by the following fact: as there is deficit
of reactive and active power in the islanded systems, the smaller
the power factor is, the larger the reduction of the voltage profile
is after islanding. As the loads have constant current character-
istics, the reduction of the active power consumed is larger in
this case. Consequently, the power imbalance after opening of
the circuit breaker decreases and the relay takes more time to Fig. 12. Relay performance curves as affected by the inertia constant of the
detect the islanding situation, however, the difference is small. generator (H)—relay setting equal to 10◦ .
176 W. Freitas et al. / Electric Power Systems Research 77 (2007) 170–180

Fig. 13. Critical power imbalance as affected by the inertia constant of the Fig. 15. Dynamic behavior of total active power load during islanding consider-
generator. ing different excitation system control modes (synchronous generator injecting
20 MW).
in Fig. 13. The results show that vector surge relays are more
effective when employed to protect small generators.
loads under distinct control modes. The dynamic behavior of
the active power load for different control modes is shown in
4.4. Excitation system control mode
Fig. 15 when the generator is injecting 20 MW. It can be seen
that the load reduction is larger when the generator is operated
A distributed synchronous generator typically has different
at unitary power factor. As a result, the active power imbalance
modes of controlling its excitation system. One is to maintain
reduction in this case is larger too, becoming more difficult to
constant terminal voltage (voltage control mode) and another
detect the islanding situation. For the same reason, the case in
one is to maintain constant power factor (power factor control
which there is excess of active power and deficit of reactive
mode) [3]. In the power factor control mode, in this section,
power after islanding showed that the critical power imbalance
the generator was controlled to keep unitary power factor oper-
is smaller in power factor control mode than that obtained in
ation. The performance curves for the two control modes are
the voltage control mode. It was also found that if the load is
compared in Fig. 14. In these cases, there is deficit of reac-
modeled as constant power load, there is no difference between
tive and active power in the islanded system. It is found that
the two control modes.
the critical power imbalance is larger if the excitation system is
controlled by power factor rather than by voltage. This is due
to the different dynamic behavior of the constant current active 4.5. Feeder length and X/R ratio

The influence of the feeder length on the relay performance is

verified in this section. In this case, the length of the feeder was
varied by multiplying the line impedance by a factor K. K was
selected to be 0.5, 1.0 and 2.0. The results are shown in Fig. 16
for relay setting of 5◦ , 10◦ and 15◦ . It can be seen that feeder
length has little influence on the relay characteristics. There is
a slight tendency for the critical power imbalance to increase
when the feeder length increases. This is caused by the higher
reduction of the nodal voltages when a feeder is longer because
there is deficit of reactive power in the islanded system. Such
voltage reduction leads to a diminution of the power imbalance
in the islanded system, becoming more difficult to detect the
islanding situation, however, the difference is very small.
The impact of the feeder X/R ratio was also investigated. The
original system has a X/R ratio of 4.30. The simulation results
with a X/R ratio of 2.15 are presented in Fig. 17 for relay setting
of 5◦ , 10◦ and 15◦ . In this study, the X/R ratio was modified
Fig. 14. Relay performance curves as affected by excitation system control considering two different cases. In the first case, the X/R ratio was
modes. reduced by halving the X value. This case is identified as “X/2”
 W. Freitas et al. / Electric Power Systems Research 77 (2007) 170–180
Fig. 16. Relay performance curves as affected by feeder length.
in the figure. In the second case, X/R was modified by doubling
the R value, which is labeled as “R*2” in the figure. It can be
seen that the X/R ratio has little influence on the performance
curves.
Fig. 17. Relay performance curves as affected by feeder X/R relation.
Fig. 18. Single-line diagram of the test system 2.
177
4.6. Multiple distributed generators
The distribution system presented in Fig. 18 was utilized to
analyze cases with multiple distributed generators. This system
is composed by a 138 kV, 60 Hz, subtransmission system with
short-circuit level of 1000 MVA, which feeds a 25 kV distribu-
tion system through a 138 kV/25 kV, Yg/
, transformer. In this
system, there are two synchronous generators with capacity of
4.5 and 6 MVA, respectively, connected at bus 7 and 9, which are
connected to network through 25 kV/2.4 kV, Yg/
, transform-
ers. In this case, there are two sets of vector surge relays and
circuit breakers: one at bus 7 (VSR1 + CB1 ) and another one at
bus 9 (VSR2 + CB2 ). It is important to mention that was consid-
ered a delay of 0.05 s between the instant of successful islanding
detection and the instant of circuit breaker opening [6]. The sys-
tem became islanded by opening the circuit breaker CB at bus
2. In multi-distributed generator case, results showed that it is
important to analyze the impacts considering situations in which
there is deficit and excess of active power and, simultaneously,
deficit of reactive power in the islanded system. Moreover, the
loads were represented by constant impedance models because
this type of load leads to the extreme behavior of vector surge
relays.
For the cases with deficit of active power after islanding,
the loads were maintained constant as shown in Fig. 18 and
the active power injected by generator 2 (SG2 ) was kept con-
stant at 5.5 MW. Then, the active power injected by generator
1 (SG1 ) was varied from 0.2 to 4.5 MW to achieve different
power imbalance levels. The results are shown in Fig. 19. In this
figure, only the performance curves associated with VSR1 are
shown. The performance curves with only VSR1 in operation
are used as a reference for comparison. These curves are labeled
as “no VSR2 ” in the figure. The curves for the cases in which
the VSR1 relay setting is equal to 5◦ , 10◦ and 15◦ are shown.
These reference cases are then compared with the cases with
two relays (VSR1 and VSR2 ) in operation, where the VSR2 set-
ting is fixed at 5◦ and VSR1 setting is again varied. Such curves
are labeled as “VSR2 = 5◦ ”. Ideally, one wishes to see that the
relay curve, which is VSR1 in this case, will not be affected by
the presence of another relay, which is VSR2 . It can be seen
from the figure, however, that when there is another relay in the
system, the performance curves of the VSR1 are much influ-
enced when this relay is set equal to 10◦ and 15◦ . On the other
hand, the curve obtained with VSR1 relay setting equal to 5◦
178 W. Freitas et al. / Electric Power Systems Research 77 (2007) 170–180

than 5◦ . If the VSR1 setting is equal to 10◦ , the relay has a

more sensitive response due to the presence of the VSR2 . This
is because it acts at the instant that the SG2 is disconnected. Such
event causes perturbations in the network, which can be large
enough to activate VSR1 . On the other hand, if the VSR1 setting
is equal to 15◦ , the presence of the second vector surge relay will
increase the critical power imbalance level from 54.1 to 90.1%,
making it difficult for VSR1 to detect the island condition. This
is because there is excess of active power after islanding. Thus,
when the generator SG2 is tripped before the generator SG1 due
to a tighter VSR2 relay setting, the power imbalance decreases
considerably. It makes the detection of islanding much more
difficult for the VSR1 . In summary, the multi-distributed gener-
ators case study results reveal that the relay characteristics can
interfere with each other, making the prediction of relay perfor-
mance very difficult. The critical power imbalances for the cases
analyzed above are shown in Fig. 21. It can be observed that the
Fig. 19. Relay performance curves as affected by multi-distributed generators
(deficit of reactive and active power).

is not affected. This is because both relays have the same set-
ting. Consequently, there is almost no interference between the
relays. The conclusion drawn from the results is the following:
if there is a deficit of reactive and active power in the islanded
system, a tight setting for one relay can reduce the critical power
imbalance level related to the other relay. This occurs because
the trip of the first generator will increase the power imbalance
level in the islanded system, leading to the other relay to operate
in a shorter time.
For the case in which there is excess of active power after
islanding, the simulation was conducted by maintaining con-
stant power for both generators at 6 and 4.5 MW. Then, the total
system active power load was varied from 10.5 to approximately
0 MW to create different power imbalance scenarios. The results
are shown in Fig. 20, where only the performance curves related
to VSR1 are presented. The results also suggest that there is inter-
ference between the two relays when VSR1 setting is greater

Fig. 21. Influence of multi-vector surge relays on the critical power imbalance:
Fig. 20. Relay performance curves as affected by multi-distributed generators (a) deficit of reactive and active power and (b) deficit of reactive power and
(deficit of reactive power and excess of active power). excess of active power.
 W. Freitas et al. / Electric Power Systems Research 77 (2007) 170–180 179

Table 1
Summary of the investigations concerning the vector surge relay performance
Case investigated Load model Description Results

The influence of
voltage-dependent loads
The influence of load power factor
The influence of the inertia
constant of the generator
The influence of the
excitation system control
modes
The influence of the feeder
length
The influence of X/R ratio
Constant power, constant current and
constant impedance
Active component: constant current
type
Reactive component: constant
impedance type
Active component: constant current
type
Reactive component: constant
impedance type
Active component: constant current
type
Reactive component: constant
impedance type
Active component: constant current
type
Reactive component: constant
impedance type
Active component: constant current
type
Reactive component: constant
impedance type
Four cases were investigate for each
load model:
• Case (a): deficit of active and
reactive power
• Case (b): excess of active power
and deficit of reactive power
• Case (c): excess of active and
reactive power
• Case (d): deficit of active power
and excess of reactive power
Deficit of active and reactive power
has been considered
Load power factors tested: 0.80, 0.90
and 0.95
Deficit of active and reactive power
has been considered
Values of the inertia constant of the
generator: 0.5, 1 and 1.5
Deficit of active and reactive power
has been considered and two control
modes of the excitation system were
investigated:
• Voltage control: set point equal
to 1 pu
• Power factor control: set point
equal to 1
Deficit of active and reactive power
has been considered. The feeder
length has been multiplied by 0.5, 1
and 2
Deficit of active and reactive power
has been considered. The feeder X/R
ration length has assumed the
following values: 4.3, 2.1 (by halving
X) and 2.1 (by doubling R)
Voltage-dependent loads have huge
influence on the relay performance
Most conservative cases:
• Case (a) and constant impedance
loads
• Case (c) and constant impedance
loads
Most optimistic cases:
• Case (b) and constant impedance
loads
• Case (d) and constant impedance
loads
Load power factor has little influence
on the relay performance
The inertia constant of the generator
has huge influence on the relay
performance. In fact, the larger the
inertia constant, the larger the critical
power imbalance
The control mode of the excitation
system has huge influence on the
relay performance. Under power
factor control mode, the critical
power imbalances are larger than
those obtained under voltage control
mode
Variations on the feeder length have
little influence on the relay
performance
The ration X/R has little influence on
the relay performance

presence of the second relay can decrease or increase the critical
power imbalance of the first relay.
5. Conclusions
This paper has presented an extensive investigation on the
performance characteristics of the vector surge relays employed
to detect islanding of distributed synchronous generators. The
proposed detection time versus active power imbalance curves
proved to be a useful tool to assist the evaluation of the relay
performance. The investigation on the relay performance has
been done through the analysis of various cases, which can be
summarized in Tables 1 and 2, as well as the main results.
Based on the results of the investigation, the main conclusions
are summarized below:
• The relay performance can be quite different depending on
if there is excess or deficit of active and reactive power
in the islanded system. The basic factor behind the differ-
ent responses is the voltage-dependent characteristics of the
loads. These important results show that utility engineers
must analyze the most conservative cases during studies of
vector surge relays. Typically, such cases are characterized
by:
- There is deficit of active and reactive power in the islanded
system and loads are constant impedance type.
- There is excess of active and reactive power in the
islanded system and loads are constant impedance
type.
• The main factors affecting the performance of vector surge
relays are inertia constant of the generator, load type and
the excitation system control mode. The excitation control
180 W. Freitas et al. / Electric Power Systems Research 77 (2007) 170–180

Table 2
Influence of multiple distributed generators on the vector surge relay performance (see Section 4.6)
Case investigated Load model VSR2 setting (◦ ) VSR1 setting (◦ ) Results

Deficit of active and reactive power
Excess of active and deficit of
reactive power
Constant impedance 5 5 No interferences between the relays
10 After VSR2 operation, the disconnection of SG2 causes
the active power imbalance rising and, consequently,
accelerates the VSR1 operation
15 After VSR2 operation, the disconnection of SG2 causes
the active power imbalance rising and, consequently,
accelerates the VSR1 operation
Constant impedance 5 5 No interferences between the relays
10 After VSR2 operation, the disturbances cause large
transients that provoke VSR1 actuation
15 After VSR2 operation, the disconnection of SG2 causes
the active power imbalance reduction and, consequently,
delays the VSR1 operation

mode affects the performance through the voltage-dependent References

characteristics of the loads.
• The feeder length and X/R ratio as well as the load power
factor have little influence on the performance of vector surge
relays.
• When there are multiple vector surge relays with dif-
ferent settings in an islanded network, there is a good
chance that the relays will interfere with each other, lead-
ing to very difficult-to-predict system responses to islanding
conditions.
Although the results presented in this paper are based
on two test systems, studies have also been conducted in
more complex systems and the results obtained were quite
similar.
[1] CIGRE Working Group 37.23, Impact of Increasing Contribution of Dis-
persed Generation on the Power System, CIGRE, Technical Report, 1999.
[2] CIRED Working Goup 4, Dispersed Generation, CIRED, Technical Report,
June (available: http://www.cired.be).
[3] N. Jenkins, R. Allan, P. Crossley, D. Kirschen, G. Strbac, Embedded Gen-
eration, first ed., Institute of Electrical Engineers, 2000.
[4] IEEE Standard P1547, IEEE Standard for Interconnecting Distributed
Resources with Electric Power Systems, IEEE, Standards Coordinating
Committee 21, 2003.
[5] Electricity Association, G59/1 Recommendations for the Connection of
Embedded Generating Plant to the Regional Electricity Companies Dis-
tribution Systems, Electricity Association Std., 1991.
[6] P. Kundur, Power System Stability and Control, McGraw-Hill Inc., New
York, 1994.
[7] IEEE Task Force, Load representation for dynamic performance analysis,
IEEE Trans. Power Deliv. 8 (1) (1993) 472–482.