Improving Vector Shift Relaying for Islanding Detection

S. Martinez de Lafuente1, F. Uriondo1, J. Tavallo2 and J.M. Garca2

1

Department of Electrical Engineering

ESI Bilbao, University of the Basque Country
Alameda Urquijo s/n 48013 Bilbao (Spain)
Phone/Fax number:+0034 946 012 000, e-mail: sony.89.mtz@gmail.com, felipe.uriondo@ehu.es
2

INGETEAM POWER TECHNOLOGY

Parque Tecnolgico de Vizcaya
Edificio 108 48170 Zamudio (Spain)
Phone/Fax number:+0034 946 018 900, e-mail: josu.tavallo@ingeteam.com , juanmari.garcia@ingeteam.com

U(t)

or
ct
Ve hift
S

Abstract The two main dedicated passive methods for

detecting distributed generation islanding are the rate-ofchange-of-frequency (ROCOF) and vector-surge (VS)
relays. The paper presents an enhanced new algorithm for
VS relaying. The new algorithm takes into account that the
real voltage shift can happen in a wave zero-crossing or that
it can be split into two different half cycles. It also measures
the true vector shift magnitude in the three phases, allowing
for future changes of settings in order to adequate them to
the real relay working situation.
Keywords: Distributed generation, islanding detection,
vector-surge (VS) relays, vector-shift (VS) relays,
synchronous generators.

1. Introduction
The problem of islanding happens when a part of the
distribution network is disconnected from the rest of the
system but the distributed generators keep feeding the
islanded part of the network. This situation involves
several problems, for the equipment and for the personnel
safety also, so, it is common practice to disconnect all
distributed generators to minimize the risks associated to
the islanding situation.
The methods to detect the islanding situation can be
classified into active and passive [1]. The paper deals
with one of the most popular passive methods used to
determinate that the islanding has happened, the so called
vector shift, vector jump or vector surge relay.
The principle of operation of the vector shift relay is
quite simple [2]. During normal operation, the terminal
voltage of an embedded synchronous generator lags the
synchronous electromotive force by an angle . If the
grid supply is suddenly disconnected from the section of
the network partially supplied by the embedded
generator, the load on the generator increases or
decreases and this causes a shift in the rotor displacement
angle. The terminal voltage jumps to a new value and the
phase position changes as shown in figure 1.

T
T
T
Fig 1- Vector Shift and variation in the wave period

The vector shift in angle and in time are related by the

following expression:

T 'T
T

(1)

Where T is the wave period (typically 20ms) and T' is the

period during the vector shift. Usually relays measure the
period calculating the time difference between two
consecutive rising/falling zero edges of the voltage wave.
The calculation is done by the relays [3], [4] every half
cycle and the relay setting threshold must be exceeded
twice consecutively to issue a trip command.

2. Problems in the measure of the vector

shift.
The vector shift presented in figure 1 is purely
theoretical, if we simulate with certain degree of detail
the vector shift using the models (to be explained and
referenced in detail in the full paper) presented by Freitas

and other authors we see that the result is the one

presented in figure 2.

each zero crossing, we see that T1 does not include the

total vector shift and neither does T3, only T2 measures it.

T2

VS2

U(t)
R

or
ct 2
Ve ift
Sh

U(t)

VS1

or
ct 1
Ve ift
Sh

Fig 2- Simulated three-phase vector shift

The problems when trying to measure the vector shift are

the following:
The vector shift in two phases (R & T) is
divided in two instants of time (VS1 and VS2) in
different half cycles.
The vector shift in phase S (VS2) happens in a
zero crossing, so it is also divided in two half
cycles.
A manufacturer [4] solves the problem using a 5 out of 6
tripping method, this is, a measurement of each phase
voltage angle is performed every half-cycle and the
decision as to whether a trip condition exists is made
every full cycle. The angle difference between the
present and previous measurement is calculated. This
process yields six results at the end of each power system
cycle (2 half-cycles x 3 phase voltages). If 5 of those 6
results are above the setting threshold, a trip signal is
initiated.

3. Summary of the new algorithm.

3.1. Basis
The key of the new algorithm is that its decision criteria
is based on three measurements in each phase, this is, on
a total of nine measurements to decide if there has been a
breaker opening or not.
The algorithm measures the period every half cycle but
takes into account that only in one of the measures the
vector shift is exactly measured. This is due to the fact
that the vector shift can be divided in two different half
cycles so only the measure that includes both half cycles
measures the vector shift in its real value.
In order to see it more clearly, in figure 3, we have
isolated the vector shift in one phase for an opening of
the breaker at full load. As illustrated in the figure the
total (and real) vector shift is divided in two parts, the so
called Vector Shift 1 and Vector Shift 2, that happen in
two different half cycles. If we calculate the period for

T3

T1

Fig 3- Simulated Vector Shift

The algorithm proceeds as follows: every half cycle it is

checked whether the measured vector shift exceeds the
set threshold or not. If the setting is exceeded this vector
shift is compared to the ones in the previous and the
following half cycles and it searches for the biggest one.
The biggest value and the half cycle when it has taken
place are memorized. Then, the values of the vector shifts
in the previous and the following half cycles to the
recorded one are added (this must be a similar value to
the recorded one). If both values are higher than the set
threshold, it is accepted that a vector shift has happened
in that phase.
The same process is repeated in the other two phases and
if the setting is exceeded in the three phases a three pole
trip is provoked.
Another important aspect to point out, that the new
algorithm also takes advantage of, is that the addition of
the vector shifts of the periods T1 and T3 is equal to the
measured vector shift in the period T2.

3.2. Ancillary elements

The vector shift is calculated with the following formula
in each phase:

360

T 'Taverage
Taverage

Where the average period is calculated based on the

periods measured in the previous cycles.
Not to distort the measure, the following measures are
taken:
In the calculus of the average period, the periods
where the vector shift has a higher value than
the established one are not included. For
example, if we take a value of 2 degrees/cycle,
that fits with a ROCOF of 0.278 Hz/s, if the

ROCOF has a higher value than this, the queue

is not refreshed and the unit ends up operating
after a certain time. This avoids the distortion of
the average period due to faults and serves as a
backup of the ROCOFs unit.
Neither are included in the average period the
values of the period when voltages are below
75% of the rated value because it is assumed
that these values are polluted by the effect of a
fault or the connection of the source.
The value of the voltage is calculated using a
half cycle Fourier transform, in that way when
the decision of including or not a cycle in the
calculus of the average is taken, you already
know the effective value of the voltage with
precision, something that would not happen with
a full cycle Fourier transform.
In order to provoke the trip order in the three
phases, the voltages have to exceed the 75% of
the rated value, to avoid the three pole trip in
three phase faults.

analyse the stability and provoking islands with different

unbalances between the load and the generation.

5. The algorithms sensitivity.

As it has been mentioned before, one of the advantages
of the algorithm is the capacity of measuring the vector
shift precisely. This advantage has been used to analyse
the algorithms sensitivity. Starting with a situation of
full load in the source, we form islandings with different
levels of instability between generation and consumption
in the islanding and we save the data of the measured
vector shift getting the curve shown in the picture below.
In the picture, for example, it can be seen that for a
setting of 4 of the vector shift, it is needed a instability
of 20% between the generation and the consumption to
make the unit act. For the typical setting of 8 it is needed
an instability of 40% between generation and
consumption to make the unit act. In the same way, the
rest of the points of the curve can be read.

4. Simulated model.

22
20

VS
Relay

2MVA
0.69kV

Yg/
0.69/13,8kV

230kV
100MVA

Yg/
13,8/230kV

Line

GS

18

Vector Shift (degrees)

The system selected to make all the simulations is

the one used in most of the articles dedicated to the
topic of the vector shift[5][6].

16
14

Perfect Island
12
10
8

Island with excess of

generation

Island with lack of

generation

6
4
2
0
0

Fig 4- Model used in the simulations

<Stator v oltage v q (pu)>

<Stator v oltage v d (pu)>
<Rotor speed wm (pu)>
<Stator current>

wref

Pref

Vabc (pu)

Pm

we

Pm

Pe0
gate

dw
Vf _

HTG
1

A
1

Iabc (pu)

<Output activ e power Peo (pu)>

v ref

Vabc
a

Synchronous Machine
2 MVA 0.69 kV

vd

100MVA, 230 kV
source
A

Vf
vq
v stab

Excitation
System

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9 1 1.1 1.2

Consumptioni (p.u)

1.3

1.4

1.5

1.6

1.7

1.8

1.9

Fig 6- Sensitivity of the algorithm of the vector shift

The implemented MATLAB model is shown in the

next picture.

<Rotor speed dev iation dw (pu)>

0.1

Discrete,
Ts = 2.083e-006 s.

Fig5- MATLAB model used in the simulations

Due to the size of the picture, the details of the simulated

model cant be appreciated in the best way. It has been
included a voltage regulator and it also has been
simulated the turbine with its control. This last point
could seem unnecessary for the study of the vector shift
but is interesting to study the algorithms stability in
large simulations.
The model also gives the chance of connecting and
disconnecting different loads, provoking faults in order to

6. Security of the algorithm.

As important as the sensitivity is the security of the
algorithm. It has been proved that for a setting of 4, the
simulated system does not act against:
Single phase-faults.
Double phase faults.
Three phase faults: It depends on the duration of
the faults, because the longer it lasts, the
measured vector shift is higher. For example, for
a setting of 4, doesnt act for faults with a
duration shorter than 100ms.
Another important aspect is the capacity of detecting the
opening of the breaker just when the fault occurs (if we
want to be more precise, when it finishes due to the
opening of the breaker). As different mechanisms are
used to numb the unit against faults, it is important to
detect the opening of the breaker just before the fault, in
this sense we can affirm that in all the cases, when the
breaker opens during a fault, the unit acts properly with
almost every setting between 4 and 12 for one-phase,
double-phase and three-phase faults.
Finally, we have to highlight the influence of the
connection and disconnection of the loads and, in it the
relative strength of the system and the point of

connection with the load is fundamental, and the

algorithm does not run away from those influences. The
connections and disconnections of high power loads
make the unit act in a wrong way.
In the simulated system, we can calculate the vector shift
provoked by the connection and disconnection of
different power loads in the exit of the generator or in the
terminals of the generator in low voltage. If we simulate
the connection and disconnection of loads in the exit of
the source, in which changing P2 (power connected en the
AT side of the transformer) we get a table like the
following:
Table 1 Vector shift provoked by the connection and disconnection of
loads

Potencia
(MW/Scc)
-20/28
-10/28
-5/28
-2/28
+2/28
+5/28
+10/28
+20/28

Potencia(pu Scc
PCC)
-20%
-10%
-5%
-2%
+2%
+5%
+10%
+20%

Salto Vector
22.69
14.45
7.77
3.20
3.22
7.81
14.54
23.91

7. Conclusions.
The designed algorithm for vector shift, measures with
precision the pure vector shifts and the zero crossing
ones. Also measures the value of the vector shift to
enable changes in the settings if needed, with criteria of
the vector shift measured by the unit. In the other hand,
the unit is relaying as it does not act against faults and
trips properly in the openings of the breakers caused by
faults. The effect of the connection and disconnection of
the loads and the inaccurate tripping in the abrupt
changes in the loads in weak systems and for loads that
are next to the source is not sorted out, but this is a
problem inherent to the philosophy of operation of the
vector shift.

8. Acknowledgements
This study has been done as a part of the PROINVER
project which is financed by Spanish Ministry of Science
and Innovation (INNPACTO program).

9. References.
[1] "The Impact of Renewable Energy Sources and
Distributed Generation on Substation Protection
and Automation". CIGRE WG B5.34. August
2010.
[2] N. Jenkins, R. Allan, P. Crossley, D. Kirschen, and
G. Strbac, "Embedded Generation", 1st ed.
London, U.K.: Inst. Elect. Eng., 2000.
[3] Crompton Instruments, "Vector Shift and ROCOF
Relay Instruction Manual".
[4] AREVA, "MICOM P341 Relay Instruction Manual
P341"

[5] John Mulhausen, Joe Schaefer, Mangapathirao

Mynam, Armando Guzmn, and Marcos
Donolo,
"Anti-Islanding
Today,Successful
Islanding in the Future", 2010 Texas A&M
Conference for Protective Relay Engineer.
[6] Walmir Freitas, Wilsun Xu, Carolina M.
Affonso, and Zhenyu Huang,"Comparative
Analysis Between ROCOF and Vector
Surge Relays for Distributed Generation
Applications".
IEEE
Transactions
On
Power Delivery, VOL. 20, NO. 2, APRIL 2005

10. Other references.

[7] Jyh-Cherng Gu, and Cheng-Chun Wang and
Kuo-Chen Hsu, "Islanding Detection and
Protection of a Utility connected Wind Turbines
via a novel FPGA-based Relay", APAP2009,
October 18~21, 2009 Jeju, Korea.
[8] F. Katiraei, C. Abbey, I. Da Cunha and Y. Brisette,
"Performance Assessment of Commercial
Relays for Anti-Islanding Protection of Rotating
Machine Based Distributed Generation".
CIGRE CANADA, Conference On Power
Systems, Winnipeg 2008.
[9] Walmir Freitas, Wilsun Xu, and Zhenyu Huang," A
Practical
Method
for
Assessing
the
Effectiveness of
vector Surge Relays for
Distributed Generation Applications". IEEE
TRANSACTIONS ON POWER DELIVERY,
VOL. 20, NO. 1, JANUARY 2005