Voltage Sag in Three Phase System

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

M.H.J.

Bollen

Voltage Sags
in Three-Phase Systems
A voltage sag is a short-duration (up to a few
sseconds) reduction in voltage magnitude.
The voltage temporarily drops to a lower value,
e.g., from 230 V down to 170 V, and comes back
again after approximately 150 ms. Despite their
short duration, such events can cause serious
problems for a wide range of equipment. Pro-
cess-control equipment, computers, and adjust-
able-speed drives are most notorious for their
sensitivity.
Voltage sags are a three-phase phenomenon, in
which all three-phase voltages are involved and
sometimes even the neutral-to-ground voltage. In
this article, we will only consider the phase volt-
ages.
Note that most publications introduce a volt-
age sag as a drop in voltage in one phase. The
three-phase aspects are introduced later as a kind
of anomaly. Such a description fits single-phase
loads and single-phase laterals in rural distribu-
tion systems. But most of the power system is
three phase, and the loads that have the biggest
problems with voltage dips, i.e., adjustable-speed
drives, are three-phase loads. In such cases, a
more general description is needed. In this article,
voltage sags will be treated as a three-phase phe-
nomenon from the beginning. Such a three-phase
approach will also make it easier to understand
what happens with single-phase loads.

Origin of Voltage Sags


The principal cause of all voltage sags is a
short-duration increase in current. The main con-
tributions are motor starting, transformer energiz-
ing, and faults (earth faults and short-circuit
DIGITAL VISION LTD
faults). Examples of voltage sags with different
causes are shown in Figures 1 through 3. In these figures, the Figure 1 shows a voltage sag due to motor starting: a rather
rms voltages are plotted instead of the voltages as a function of small sudden drop in voltage, followed by a gradual recovery.
time. The rms voltage is calculated over a one-cycle window, Those who are familiar with motor starting current will recog-
and this window is then shifted across the time axis. Here the nize the shape. As electrical motors are three-phase balanced
rms calculation is updated each sample. In most power-quality loads, the voltage drops are the same in the three phases.
monitors this only occurs once or twice per cycle. The voltage sag shown in Figure 2 also shows a sudden drop
followed by a slow recovery. The drops are different in the three
This article is part of a series of articles on power quality appearing in this and phases, however. This event is due to the energizing of a large
future issues of IEEE Power Engineering Review. M.H.J. Bollen is with transformer. The inrush current is different in the three phases
Chalmers University of Technology, Gothenburg, Sweden. and associated with large second and fourth harmonic distor-

8 0272-1724/01/$10.002001 IEEE IEEE Power Engineering Review, September 2001


tion. These even harmonics are found back in the voltage re- ages. The effect of a Dy-transformer is the same as a change
cordings. When transformer saturation is involved, the use of a from star connection to delta connection. A star-connected mon-
half-cycle window to calculate the rms voltage will give a fast itor on the secondary side of a Dy-transformer will experience
fluctuating value. As one never knows beforehand whether the same sag as a delta-connected monitor on the primary side.
transformer saturation is involved, it is best to use a one-cycle The sags shown in Figures 4 and 5 give all possible combina-
window in any automatic recording. tions. The situation gets a bit more complicated when two
The voltage sag shown in Figure 3 is due to a single-phase phase-to-ground faults are also concerned, but it has been shown
fault: the voltage drop is sharp in two phases and recovers sharply that they fit in the same pattern. The reader is referred to the lit-
a few cycles later. Drop and recovery are associated with fault ini- erature for more details about this.
tiation and fault clearing, respectively. In this example, the volt-
age during the event is constant, and the voltage recovers
immediately. This is not always the case. Some events show mul-
tiple stages due to developing faults or due to the breakers on both 200
sides of a transmission line opening with some time delay. Due to
induction motor load, the voltage during the fault can show a slow 215
decay. After fault clearing, the voltage can show a slow recovery

RMS Voltage (V)


210
due to motor acceleration or due to transformer saturation.
205

Sags Due to Faults


200
The most severe voltage sags are due to short-circuit faults and
earth faults. Therefore, they have recently received the most at-
195
tention in the literature. Different types of faults lead to different
types of voltage sags. Methods for calculating the phase volt-
190
ages during a fault are treated in most books on power system 0 1 2 3 4 5 6 7 8 9 10 11
analysis. Time (Cycles)
The common way of describing voltages during a fault is
through complex phasors. (A phasor is nothing more than a Figure 1. Voltage sag due to motor starting
complex number represented through a vector.) The absolute
value (magnitude, length) of the phasor corresponds to the rms
voltage as obtained from measurements. The argument (phase 11
angle, direction) of the phasor corresponds to the so-called 10.9
phase-angle jump. The phasor diagram for a single-phase fault 10.8
is shown in Figure 4. The kind of sag experienced by the load (or
10.7
RMS Voltage (V)

by the monitor) depends on its connection. Star-connected load


experiences a drop in one phase but no change in the two other 10.6
phases (we assume that the nonfaulted phases are not affected). 10.5
Delta-connected load will experience a drop in two phases, 10.4
where the third phase is not affected. 10.3
In the case of a phase-to-phase fault, the voltage phasors are 10.2
as shown in Figure 5. A star-connected load experiences a volt- 10.1
age drop in two phases; a delta-connected load, a large drop in
10
one phase and a small drop in the two other phases. Note that a 0 5 10 15 20 25
phase-to-phase fault for a star-connected load gives a similar sag Time (Cycles)
as a single-phase fault for a delta-connected load. It can be
shown that it is not possible to distinguish between these from Figure 2. Voltage sag due to transformer energizing
the voltage recordings only.
A voltage sag is rarely recorded (or experienced by the load)
at the same voltage level as the one at which the fault occurs. 11.5
There are, thus, typically one or more transformers between the
fault and the monitor (or the load). These transformers change 11
the way in which the individual phases are affected.
RMS Voltage (V)

There are several different transformer winding connections 10.5


but a closer look shows that, as far as voltage sags are concerned,
there are just three types of transformers: 10
Those that do not affect the individual phases (Yy,

grounded on both sides) 9.5


Those that remove the zero-sequence voltage (e.g., Dd)
Those that change phase voltages into line voltages and the 9
other way around (e.g., Dy).
For most loads, the zero-sequence voltage is not of impor- 8.5
0 2 4 6 8 10 12 14 16
tance. It is either removed by a Dy-connected distribution trans- Time (Cycles)
former, or the load is delta connected, or both. Thus, only the
latter type of transformer changes something to the phase volt- Figure 3. Voltage sag due to a fault
IEEE Power Engineering Review, September 2001 9
Characteristics and Indices
Voltage sags are a multidimensional phenomenon. The voltage
drops are typically different in the three phases. The voltage
drop is not always constant. The drops in voltage are typically
associated with phase-angle jumps, and then we do not even talk
about point-on-wave, waveform distortion, or the transients at
the start and end of the event. To characterize all this with a
small number of values is rather difficult and still no interna-
tional standard has tackled all the issues.
In all characterization schemes, a characteristic voltage as
a function of time is calculated from the sampled phase volt-
ages. From this characteristic voltage, magnitude and duration
of the event are determined. The various documents mainly dif- Figure 4. Voltage sag due to single-phase fault for star-connected load (left)
fer on the way in which the characteristic voltage is calculated. and for delta-connected load (right)
The magnitude is typically defined as the lowest absolute value
of the characteristic voltage; the duration is the time during
which this absolute value is below a certain threshold.
Calculation of the characteristic voltage varies for the differ-
ent types of sags. Most monitors take the lowest of the three rms
voltages, but this will give different results for delta-connected
or star-connected monitors. A more acceptable method will con-
sider the three-phase nature of the power system. For balanced
sags (motor starting, three-phase faults), the three voltages are
the same so that the choice is easy. The resulting sag is classified
as Type A. For sags due to unbalanced faults, a distinction is
made between types C and D (left and right in Figure 5, respec-
tively). For Type D, the characteristic voltage is the voltage of
the affected phase. For Type C, it is the voltage between the
affected phases. In the latter case, a factor 3 should be used Figure 5. Voltage sag due to phase-to-phase fault for star-connected load (left)
to make the results of the two types compatible. Defining the and for delta-connected load (right); Type C (left) and Type D (right)
characteristic magnitude in this way makes it independent of the
monitor connection and independent on the transformer wind- Stochastic methods use a model of the system to predict the
ing connection. stochastic properties of the system, in this case voltage sag fre-
quency. The site-indices can be calculated for all locations
within a system, and for any period in the past or in the future.
The disadvantage is obviously that the results are as accurate (or
Statistics and Stochastics as inaccurate) as the model used. The current emphasis in the lit-
In the previous section the single-event indices magnitude and erature is very much in statistical methods; stochastic methods
duration were defined. They can be used to describe an individ- have not yet received the attention they require.
ual event. To describe the voltage quality at a certain location or
the performance of a certain network, a number of additional in- Mitigation
dices need to be introduced. Again, no definitive version has When a certain customer or installation suffers from voltage
been agreed upon, but, in almost all proposals, event magnitude sags, a number of mitigation methods are available. These can
and duration are used as a basis for the calculation of site indices be grouped as follows:
and system indices. A site index typically gives the number of Reducing the number of events

events per year more severe than a given magnitude and dura- Improvements in the power system

tion. A system index is typically a weighted average of the sites Mitigation at the interface between system and load

for which information is available. The ongoing activities in dif- Improving end-use equipment.

ferent working groups are a balancing act between the need to Reducing the number of faults tackles the problem at the
have indices that are suitable for a wide range of customers and source. It will not only reduce the number of voltage sags but,
loads, and on the other hand a small number of indices. Ideally, especially for distribution systems, also the number of interrup-
one would like to be able to capture the voltage sag performance tions. This is not always an easy task, and the costs could be very
of a system in one single index; something like the THD for high. Keep in mind that voltage sags at a certain location can be
waveform distortion. due to faults at distances up to hundreds of kilometers.
When collecting data to calculate site and system indices Improving the power system does not reduce the number of
(typically event magnitude and duration), two fundamentally voltage sags, but it makes them less severe. An apparently
different methods are being used. Statistical methods observe straightforward method is to reduce the fault clearing time.
the behavior of the system and calculate averages from the ob- Again, this may require large investments at the transmission
servations. This requires a certain number of monitors that ob- level. At the distribution level, the margin for speeding up is
serve the system for a certain time. To get accurate results, a much larger, although it may increase the risk of protection
large number of monitors and a long measurement period are maloperation. A method used regularly in distribution networks
needed. Monitoring will always give the real picture, but the re- is to remove long overhead feeders from busses feeding sensi-
sults have limited application for other locations and for future tive loads. Other system improvements, such as adding multiple
prediction. infeeds and strategically locating generators, are normally only
10 IEEE Power Engineering Review, September 2001
Normal Backup
Supply Supply

100 I II
Faults in Other Motors + Transformers

Transmission System Faults


Distribution
80 Networks
Magnitude (%)

A B

50 Faults in Other
Distribution
Networks C
Fuses

1 2 3
Interruptions
0
0.1 1
Duration (Seconds)

1 2 3
Figure 6. Voltage sags and interruptions, different origins Figure 8. Principle of the static transfer switch

feasible in industrial systems. A solution often neglected in the 100 ms, are mainly due to faults in the transmission system. Mit-
literature is system grounding. Using high-impedance ground- igating these at the system side is very difficult, whereas mitiga-
ing will prevent voltage sags due to single-phase faults from tion at load side is possible in most cases. The same holds for the
propagating through a Dy-transformer. This is common practice shallow events: switching of motors and transformers, remote
in European distribution and subtransmission systems as well as distribution system faults. Preventing an individual event is not
in many industrial systems. very difficult, but there are simply too many remote locations to
The most commonly used method is to install additional cover them all.
equipment at the system-load or system-customer interface. Ex- On the other side of the chart, it is very difficult to make
amples are the UPS (a must for any essential low-power equip- equipment immune to interruptions, whereas system-side solu-
ment, but batteries become a problem for larger load), the tions are easier (albeit not always cheap). The majority of inter-
motor-generator set (sometimes forgotten, but still an excellent ruptions originate in the distribution network. Reducing the
alternative), the ferroresonance transformer, the dynamic volt- number or duration of interruptions requires improvements in
age restorer (DVR), and the static-transfer switch. The latter two the distribution network.
solutions are discussed in more detail. From this reasoning, the following general guidelines can be
Improving end-use equipment has a larger potential than of- extracted: sags down to 75%, 100 ms should be solved by equip-
ten thought, but a combination of factors has kept this potential ment improvement; sags from 60%, 500 ms and more severe
untapped. The result is that the customer is often left with just should be solved by system improvement. Anything in between
one option: installing mitigation equipment. is worth discussing. In the future, these two limits should move
In Figure 6, different origins of voltage sags are indicated in a towards each other, e.g., through an equipment immunity stan-
magnitude-duration plot. For completeness, interruptions are dard.
also indicated. The events with the shortest duration, up to about A recently introduced solution for mitigating voltage sags for
large loads is the DVR. The principle is shown in
Figure 7. The voltage sag in the system is miti-
2
x 104
gated by injecting the difference between the de-
1.5
sired voltage and the actual voltage. This
1
0.5
Injected
injection requires a power-electronics converter,
0
Voltage a high-power source of energy, and a fast control
0.5
1
system. Sources of energy currently in use
1.5 include capacitors and superconducting coils.
2
0 200 400 600 800 1000 1200 1400 1600 To mitigate long and deep sags, a large energy
Voltage Sag Constant Load reservoir is needed.
in the System Voltage
2
x 104
x 104
Another rather new solution is the very fast
1.5
2
1.5
medium-voltage transfer switch (see Figure 8).
1
1 When a voltage sag (or interruption) occurs in
0.5
0
VSC 0.5
the supply, the load is switched to the alternate
0
0.5
0.5
supply. Using thyristors will give a transfer time
1
1.5
1 less than one cycle (manufacturers claim 4 ms),
1.5
2
0 200 400 600 800 1000 1200 1400 1600 2
whereas using vacuum switches will take one or
0 200 400 600 800 1000 1200 1400 1600

Energy
two cycles for the transfer. The transfer switch
Storage cannot mitigate sags that originate in the trans-

Figure 7. Principle of the DVR (continued on page 15)


IEEE Power Engineering Review, September 2001 11
The notation Psupplied refers to the total power taken by a load, [5] E.R. Collins, Jr. and R.L. Morgan, Three phase sag generator for
and the rms load voltage and current appear in the denominator. testing industrial equipment, IEEE Trans. Power Delivery, vol. 11,
The displacement factor is the cosine between the fundamental pp. 526-532, Jan. 1996.
voltage and current at a load. For the sinusoidal case, the power [6] G. Heydt and M. Negley, Alternatives for special power acceptabil-
factor and displacement factor are identical. ity services, in Proc. NSF Conf. Unbundled Power Quality Services,
Key West, FL, Nov. 17-19, 1996, pp. 176-181.
Summary [7] T. Key, Predicting behavior of induction motors during service
Power acceptability is important because it describes whether faults and interruptions, IEEE Ind. Applicat. Mag., vol. 1, no. 1, p. 7,
electric power delivered from the power system is compatible Jan./Feb. 1995.
with load characteristics. A popular measure of acceptability is [8] R. Waggoner, Understanding the CBEMA curve, Elect. Con-
the use of the power acceptability curves that measure the sensi- struction and Maintenance, vol. 90, no. 10, pp. 5557, Oct. 1991.
tivity of the load to voltage sags (and also overvoltages). The
best known power acceptability curve is the CBEMA curve. [9] G. Heydt and W. Jewell, Pitfalls of electric power quality indices,
IEEE Trans. Power Delivery, vol. 13, number 2, pp. 570-578
Rectifier loads and other electronic loads may be susceptible Apr.1998.
to bus voltage sags. The use of a switch-mode power supply can
alleviate the vulnerability. [10] G.T. Heydt, Computer Analysis of Power Systems, 2nd ed.
It is possible to quantify the number of low-voltage events Scottsdale, AZ: Stars in a Circle Publication, 1996.
per year seen by a set of customers. The SARFI index is exactly [12] IEEE Recommended Practice for Design of Industrial Power Sys-
this measure, and it may be calculated on the basis of different tems (Gold book), IEEE Standard 493-1980, IEEE Press, 1980.
voltage sag depths. [13] A.E. Turner and E.R. Collins, The performance of AC contactors
during voltage sags, in Proc. IEEE Int. Conf. Harmonics and Qual-
References ity of Power, Las Vegas, NV, Oct. 1996, pp. 589-595.
[1] Recommended Practice for Powering and Grounding Sensitive [14] N. Mohan, T. Undeland, and W. Robbins, Power Electronics. New
Electronic Equipment (Emerald book), IEEE Standard 1100, IEEE York: Wiley, 1995.
Press, 1992.
[15] R.S. Thallam and G.T. Heydt, Power acceptability and voltage
[2] Guideline on Electric Power for ADP Installations, United States
sag indices in the three phase sense, paper presented at the Panel
Federal Information Processing Standard (FIPS), Publication No.
Session on Power Quality: Voltage Sag Indices in the Three Phase
94, Washington, DC, 1978.
Sense IEEE PES Summer Meeting, Seattle, WA, July 2000.
[3] Guideline on Electrical Power for ADP Installations, United States
Federal Information Processing Standard (FIPS), Publication No. [16] R.S. Thallam, Power quality indices based on voltage sag energy
94, National Technical Information Service, United States Depart- values, paper accepted for presentation at the Power Quality 2001
ment of Commerce, Springfield, VA, 1978. Conference and Exposition, Chicago, IL, Sept 9-13, 2001.
[4] G.T. Heydt, Electric Power Quality. Scottsdale, AZ: Stars in a Cir- [17] American National Standard for Electric Power Systems and
cle Publications, 1995. EquipmentVoltage Ratings (60 Hz), ANSI Standard 84.1, 1989.

Voltage Sags in Three-Phase Systems


(continued from page 11)

mission system, as those will affect both feeders. The two-cycle [2] M.H.J. Bollen, Understanding Power Quality ProblemsVoltage
switch can be combined with 100-ms energy storage against Sags and Interruptions. Piscataway, NJ: IEEE Press, 2000.
transmission system sags. [3] L.E. Conrad and M.H.J. Bollen, Voltage sag co-ordination for reli-
able plant operation, IEEE Trans. Ind. Applicat., vol. 33, pp.
The Future 1459-1464, Nov. 1997.
Solving the voltage sag problem requires cooperation between [4] L.E. Conrad, K. Little, and C. Griff, Predicting and preventing
utilities and end-users, with support from equipment manufactur- problems associated with remote fault-clearing voltage dips, IEEE
ers and standards-setting organizations. It is possible to make Trans. Ind. Applicat., vol. 27, pp. 167-172, Jan. 1991.
better equipment when the demand for it is high enough. Stan- [5] M.F. McGranaghan, D.R. Mueller, and M.J. Samotej, Voltage sags
dards are needed that guide utilities, end-users, and equipment in industrial power systems, IEEE Trans. Ind. Applicat., vol. 29, pp.
manufacturers. It is also important that everybody becomes aware 397-403, Mar. 1993.
of voltage sags as a potential problem. Education of electrical and [6] L. Zhang and M.H.J. Bollen, Characteristic of voltage dips (sags)
mechanical engineers is very important to achieve this aim. in power systems,IEEE Trans. Power Delivery, vol. 15, pp.
827-832, Apr. 2000.
Further Reading
[1] M.H.J. Bollen, Characterisation of voltage sags experienced by
three-phase adjustable-speed drives, IEEE Trans. Power Delivery,
vol. 12, pp. 1666-1671, Oct. 1997.

IEEE Power Engineering Review, September 2001 15

You might also like