OVERVOLTAGES IN HV DC TRANSMISSION SYSTEMS CAUSED BY TRANSFORMER MAGNETIZING INRUSH CURRENTS

John P. Bowles
Hydro-Quebec Institute of Research, Varennes, P.Q., Canada

ABSTRACT Excluding magnetizing inrushf effects of transform-

ers and failures within the HV DC convertors themsel-
HV DC transmission schemes can have system condi- ves, most transient events in HV DC systems (such as
tions in which the ac system is in resonance with the dc load changes, blocking or deblocking of bridges,
ac harmonic filters at particular harmonics of the and ac and dc faults) can give rise to fast changes of
supply frequency, and these resonances may be excited current at the' ac busbars which can shock the ac fil-
by transformer magnetizing inrush currents, generating ter - ac system combination into natural "ringing",
overvoltages at the ac busbars of the HV DC convertor but do not cause sustained oscillation. Such a shock
station. In the case of a rectifier station, supplied excitation may result in single peaks of voltage at
directly by generators with no local load, these over- the ac busbars approaching 1.7 times rated ac busbar
voltages may be as high as twice the rated voltage and voltage with ac systems having small losses, and if
may exist for many cycles of the supply frequency. the timing of the transient is particularly unfortuna-
This effect is examined in this paper and the results te. The harmonic components of such an overvoltage
of many thousands of model tests are included. generally decay naturally to negligible proportions
within a few cycles of the supply ftequencv, due to
While the paper concentrates on the effects within the damping within the ac system.
hv dc systems it should be realised that similar phe-
nomena are being observed in high voltage ac lines and A more important effect is the generation of trans-
that comments made in- the paper are relevant to the ac former magnetizing inrush currents during the tran-
case. sients mentioned above. These inrush currents can
repeatedly shock the ac system - ac filter combination
INTRODUCTION once per cycle, and as the decay time of inrush cur-
rents is very long for high power transformers (i.e.
Two of the main effects which contribute to over-~ of the order of seconds before the inrush currents
voltages at the ac busbars of a high voltage direct reach negligible proportions) it provides a mechanism
current (HV DC) transmission are: by which the natural resonances of the ac filter and
ac system can be excited for substantial times. This
(1) Fundamental frequency regulation due to rejection can result in overvoltages for many cycles of the sup-
of load. ply f requency.

(2) Excitation of the natural resonances between the It is important to note that these magnetizing
ac harmonic filters required in HV DC schemes, and inrush currents, and therefore the overvoltages asso-
the ac system. ciated with them, can be generated each time a trans-
former is energised; and that the transformer consi-
If the ac system impedances are known at fundamen- dered need not be a convertor transformer, but can be
tal frequency, regulation calculations, which are usu- any transformer of appropriate size electrically close
ally carried out by means of 'a digital computer, are to the ac busbars. Therefore such overvoltages can be
relatively straightforward. Dynamic voltages due to expected fairly frequently and it is important to know
regulation rarely exceed 1.25 times rated voltage at the severity and frequency of occurrence of overvol-
the ac busbars of HV DC schemes, but can exist for a tages in excess of 'critical magnitudes (e.g. surge
considerable time. If ac machines (e.g. synchronous diverter sparkover levels).
compensators) are close to the ac busbars, the busbar
voltage' can be controlled by the automatic voltage re- Although transformer magnetizing inrush phenomena
gulators of the machines, so that the duration of the have been examined in many papers and text books, some
dynamic overvoltages may be limited to a time of the of which are listed in references l1 6, the generation
order of 0.5s. Without such devices close to the ac of overvoltages by these inrush currents has not been
busbars the overvoltages may exist for times corres- so fully evaluated, particularly with respect to
ponding to tap changer operations. Dynamic regulation effects within hv dc transmission systems..' This paper
effects are not considered further in this paper. presents the results of a series of transformer ener-
gizations (20,000) within a simulated hv dc transmis-
HV DC transmission systems require ac harmonic fil- sion terminal. While these tests are directly rele-
ters at the ac busbars in order to divert the harmonic vant to the dc case similar circumstances can be found
currents generated by the convertor bridges away from within ac systems (i.e. the line capacitance or reac-
the ac systems. As the ac system impedance changes tive compensation of EHV lines is equivalent to the
(e.g. due to line switching or the addition or sub- effective capacitance of the hv dc filter), n these
traction of machines) the ac system may be brought tests are therefore relevant to the ac case.,
into resonance with the ac filter at particular harmo-
nics of the supply frequency. Due to the large number of variables involved it is
impracticable to conduct an exhaustive series of tests
on site. Most work of this type has therefore to be
carried out in conjunction with model networks, using
evidence available from site to verify the accuracy of
the model. This paper describes tests carried out on
a power system simulator, ref. 9. The modelling of
the convertor transformers and ac system for realistip
operation has been discussed in an earlier paper ref.
Paper T 73 433-0 recommended and approved by the IEEE Transmission
prv.
and Distribution Committee of the IEEE Power Engineering Society for presenta- In order to control and analyse the results of the
tion at the IEEE PES Summer Meeting and EHV/UHV Conference, Vancouver,
B.C., Canada, July 15-20, 1973. Manuscript submitted February 20, 1973; made many thousands of switchings required to give adequate
available for printing May 2, 1973. information concerning the magnitudes, frequencies,
487
and probability of occurrence of the anticipated over- B /B and Br /B should be the same in the model as in
voltages the simulator was connected to a hybrid com- the mreal transformer. Assuming typical values of
puter described in ref erence 9'. these quantities B = 20.5 kG, B = 13 kG and B =16 kG
B /B = 1.3, B /B = 0.8. r m
s m r m
MAGNETIZING INRUSH CURRENTS
Figure 2 shows the B-H characteristics of the model
On de-energising a transformer the core is general- transformer used in the following tests.
ly left with residual flux, 4, which in cold rolled
steel used in- modern high!r power transformers can I
approach 80% of the rated magnetizing flux swing 4)
If the transformer is re-energised at such a moment
that the- magnetizing flux adds to the residual flux I -T I -]
the total 'flux, 4, can exceed the saturation flux of
the transformer core 4 s', and the magnetizing current,
-I
i, then increases to a value limited by the air core
inductance X-, of the primary winding. This current I 1
can be many times the rated current of 'the transformer. - .
This effect is-considered more fully in references 1-6. I t -- I I
Figure 1 shows the timing sequence required for the (a) (b)
peak value of the magnetizing inrush current.
Fig. 2. B-H loops for the model transformer a) at
* ~~~~~~~~~I
-I rated voltage; B scale - 2 divisions corres-
pond to saturation of the core, magnetizing
current axis - 1 division corresponds to 1%
rated current (to the base of transformer MVA);
b) at 150% rated voltage; B scale - as a.,
_____- ;- _
2#m magnetizing, current axis 1 division corres-
ponds to 10% rated current.
Typical values of convertor transformer saturated
impedance (to the base of transformer MVA) are of the
order of 0.20 pu (calculated as in equation 3). The
maximum magnetizing inrush current is then anticipated
to be 7-8 times peak rated-current (to the base of
*1 transformer MVA).

The saturated inductance of the model transformer

was adjusted to be 0.20 pu and Figures 3(a) and, (b)
show the magnetizing inrush current for the model on
Fig. 1. Generation of maximum magnetizing inrush cur- closing a circuit breaker timed to be near to the
rent following transformer energisation. point for maximum inrush current in the phase shown.
Both primary and secondary windings of the transfor-
mer were connected in star (i.e. each phase complete-
For the case of a transformer connected to infin'tX
ly independent of the other two phases).
busbars, andvwith no delta winding, it can be shown
that the inrush current is described by
s/2E
i = k2
XT (coswt - cosO) (1)
-- J-,.1
where E is the rms value of the exciting voltage, k2
is a constant and O'is the saturation'angle of the
transformer.

O = k1cos
-l 4s4)m4)
( s ) = cos -l
B -B
r- (2)
m m
B , and B are the respective flux densities, and (b)
33. ..
rrJA- - f- - -' '-
B
ks is a constant.
1
2 2
0. 81T f N- A x 10 -8 ohms
XT - (3)

where Q is the length of the winding in cms, A 2is the

mean cross sectional area of the winding in cm , N is
the number of primary turns and f is the fundamental
frequency of. the supply.
H t iil
t! -! ->1-
From a simple analysis of inrush currents it would
--- ., - ~-_
appear that- the factors k and k sh uld be unity.
However it has been 'found in praciice that average (c) (d)
values of' k'1 = 0.90 andk = 1.15 compensate for the
errors involved in the simpie theory. Fig. 3. Magnetizing inrush
currents of the model
transformer', and b) star-star connected
a)
Transformer Model transformer and d) star-delta connected
c)
transformer. Each vertical division repre-
It can be seen from equation 2 that the ratios sents 1 pu peak rated current.

488
 It is common to have a tertiary delta winding added impedance angle of 850, i.e. a reasonable representa-
to a transformer if none of the main windings is tion of0 the system supplying an isolated rectifi
delta connected. It can'be seen from Figures 3(c) and station such as in the Nelson Ri4*er hv dc syste5c.
(d)' that the addition of a delta connected secondary
makes very little 'difference to the peak values of Each test in the series comprised 1000 transformer
magnetizing inrush, but the effective duration (20) of energizations with random switch closings
0
and-openings
the current pulse is increased due to-distortion of with a random (within a band of 60 ) spread of contact
the waveform, particularly at lower currents. closings and openings. An analysis of each test was
made and histograms produced to show the magnitude and
The rate of decay of the magnetizing inrush current frequency of the overvoltages produced.' A sample
corresponds to the resistance of the transformer win- histogram is shown in Figure 6.
ding which in a typical convertor transformer is of
the order of 0.OOlpu (on transformer MVA). The model Figure 7 indicates the preliminary tests carried
transformer has a winding resistance of 0.003pu and out and the maximum voltage observed during
therefore the decay rate of its inrush current is each run, the standard deviation, the voltage level
correspondingly faster than required, but is suffi- above which 2% and 50% of each run lie, and the mean
ciently realistic to -illustrate the phenomena inves- voltage.
tigated.
Three further series of tests were carried out in
the same manner to determine the effects of varying
the system impedance and the amount of damping present
i.e. the impedance angle. These results are tabulated
in Figures 8, 9, 10 and curves of the 2% levels are
AC system Convertors plotted in Figure 11.' Throughout the tests the hv dc
convertors were kept blocked.
Discussion of Test Results

Before attempting to explain the generation of the

as busbar overvoltages the significant general
features should be listed.
r AC filter
(1) The ac busbar overvoltages can build up slowly
L5 L7 L1 L13 CD over a number of cycles of the supply frequency (Figu-
re 12) and decay slowly as the magnetizing inrush cur-
TC5 C7 TCi TCI3TIL rent peaks decay. The overvoltages therefore exist
for a fairly large number of cycles of the supply fre-
R5 R7 fR1s R13 LD RD quency. It should be realised that as the model
transformer in -the tests described has a magnetizing
inrush current decay rate 2-3 times faster than that
of the real transformer, the number of overvoltage
Fig. 4. The system tested. peaks should be multiplied accordingly.
(2) The peak ac busbar voltages can occur at relative-
ly small inrush current peaks and can therefore be ge-
nerated fairly frequently (i.e. it does not require
the generation of the maximum inrush currents to ge-
nerate these overvoltages).

(3) For a system short-circuit ratio of 3 with a dam-

ping angle of 85 at the third harmonic (i.e. corres-
ponding to the damping given by the ac system genera-
tors) 2% of the generated overvoltages were in excess
of 2.0 pu.' This value is substantially independent of
the' number of transformers switched, their-winding
configuration, their saturated impedance, or the sys-
tem voltage level prior to-excitation (Figure 7).
(4) The peak overvoltages are very dependent on the
amount of system-damping present (Figure 11) and the
system impedance itself.
0 1 3 5 7 9 11 13 15 17 l6 21 23
Hormonic number n
(5) The histogram showing the frequency of the measu-
Fig. 5. The impedance characteristic of the harmonic red 'overvoltages, shown in Figure 6, has two peaks,
filter in Figure 4. one centred around the area of little overvoltage,
another centred around the 1.7 pu level.
Model Tests
Although 1000 transformer energisations were made
.The system of Figure 4, including a complete harmo- during each run, it is considered that higher overvol-
nic filter having the impedance characteristic shown tages may still be possible if each variable is maxi-
in Figure 5, was set up on the transmission simulator. mised to the fullest extent; i.e. there are so many
An initial series of tests was carried Qut to determi- variables (breaker closing instant, spread of contacts
ne how different transformer arrangements affected the on closing, breaker opening instant, spread of con-
ac busbar overvoltages subsequent to transformer ener- tacts on opening, filter ringing effects added to the
gization. The system impedance was'set up for this forced oscillation due to transformer energisation
series to giye a ratio of-short-circuit MVA to-rated etc.) that the worst overvoltages- may still not be
dc power (short-circuit ratio - S.C-.R.) of 3 with an indicated.

489
 * 4
* 4
* 4 Fig. 6. Results showing the overvoltages and frequency
of such overvoltages after 1000 energisations
* * * . . . of a single star-delta connected transformer
with an ac system S.C.R. - 2.7 impedance angle
*4 =85°.
850
C~ ~~~~~~~~~~~ *
* 4

a.
_.
*. 4
V * *

a. *
vI
v
.
u

.4
* 4

7
LI . 4 4

-. 4

a
4.
*

C,

Z 4~~~~~~~~~~~~~~~~~~~~~r"'lk
( . . . . . . . . . . . . . .

*~~~~~~
4 4 *

.* 4
* *
* 4*

*
*4
S,
v v v l-.le;t~~~~~~~~~~ rrs oDQAQTjv|qs * jr,
.44 4 9ucTFeIrTh !TNM1.iL 4 qQd
*
*. 4 ; F IeNzfF4"
TP~(. * ' 1
_

*
.** .
+~~~~~~~~~~~~
~~~~~~~~~~~~~~~~4+4
4 * * rv'axl,cj h. ,PPV
**. . -~~~~~~~~~~~ *
* 4. *
4~~~~~~ v f.1 1. r *~.P~1~
* .

- .44 U 4 44*4
4~~~~~~~~~~~~~I
V1 44444v¢
DN 04
7- -'If ,'%
¢¢¢ -I OX M~

ce .* *-,**-S+f*444 -

4- *4 4444444444444
** tt ¢¢

::
.44 ;
L . ¢4 ****- 44 4 44 4
***¢* 4444
.44* ** 44 4444444444*444v
c ..+ +* * * v v¢ v B B v - v
.44 + + ¢ * * *+ + ^ ^ * ¢ ¢ v *

- .44
" *++ * 444v+ vv @ @ 4444444444444 vv¢

lX4* +* 44 4 44444444444444444+++* ++* +- +#+ *+

r.444
.-4
-, 44 4-
4o1>s1w
4 4
4z>v_vssv-rXPors
4. .4
\Jutf b1 >
.444...r..4.4it
44449444444 .-9ns s_n
. -1 aS
wwes -.
r:4 44 - ...:C rrw^vXvr f trXs. s Jre(Co~
14 44 4 44 4 4 4444444444444444444X
**** eee @** @++ -9** ***

_______4 ___ ___ 44_ 4 __ 44 . 44

_ 44 4 44 44 44 4

490
 Maxi1m voltage Keen voltage S01 level 22 level Standard
pu pu pu pu deviation
Inergimimg one_
star-star transformer 2.03 1.40 1.41 1.73 0.178
at rated voltage
Emergislss one 2.0 2.0-
star-delta transformer 2.17 1.57 1.61 2.01 0.275 System impedance angle System short circuit
at rated voltage
constant at 850 ratio constant ot 3.0
ftergising both :j
star-star and 2.09 1.63 1.67 1.96 0.253
star-delta transformer
et reted voltage
Energlsing one
a:
°1.5 91.5
0.251
star-delte transformer
et 110X voltage
2.21 1.68 1.70 2.07
N
_~ 0
Energlsing one
ster-delta transformer 2.14 1.66 1.67 2.00 0.213
at 1202 voltage
Energising one
star-delta transformor 2.07 1.64 1.64 1.96 0.180 I I
at 1301 voltage 2.5 3.0 4.0 ,0 60 7.0 750 80 85°
Systen short circuit ratio Systen knpodance angle
Fig. 7. Results of preliminary series of tests to de- (a) (b)
termine the effects on the ac overvoltages of
Fig. 11. Plots of the 2% overvoltage levels versus
different transformer configurations and exci-
on transformer energisation system short-circuit ratio and system impe-
tation levels dance angle (defined at the 3rd harmonic of
S.C.R. 3 impedance angle 850 (at the third the fundamental).
harmonic of the fundamental). Each result was
obtained after 1000 energisations.
~f I I
41 1 71P. '
Impedance angle 85°
Short-circuit Maximum voltage Mean voltage 502 level 2X level Standard
'I'IJ2 Il I^ 0>,l [' i l
ratio pu pu pu pu deviation j_ l I r A l
2.5 2.02 1.52 1.59 1.88 0.258 C,.~~~~~~~~~~~1
3.0 2.17 1.57 1.61 2.01 0.275
-
3.5 1.79 1.37 1.40 1.64 0.174 9
I-~
4.0 1.65 1.29 1.30 1.51 0.136
it
l -
- w VW I;r ,.- .(
4.5 1.48 1.26 1.28 1.43 0.111 11.
5.0 1.49 1.22 1.23 1.39 0.103 (a) (b)
6.0 1.55 1.23 1.24 1.42 0.102
Fig. 12. Busbar voltage on energising a single star-
7.0 1.46 1.22 1.22 1.40 0.100 delta connected transformer a) with a system
impedance angle of 850 at the third harmonic
of varying the impedance
b) with a system impedance angle of 750 at
Fig. 8. The effect ac system
the third harmonic. Upper trace ac busbar -
keeping the impedance angle at the third har-
monic of fundamental constant at 85 Each voltage- 1 division corresponds to peak rated
inrush
.

result was obtained from 1000 energisations of voltage. Lower trace magnetizing -

a star-delta connected transformer.

current in the same phase- 1 division corres-
ponds to twice peak rated current (on the
transformer MVA).
Impedance angle 750
Short-circuit aoxims voltage Mean voltage 50% level 2% level Standard
deviation
Firstconsider the impedance characteristic of the
ratio pu pu pu pu
1.35 1.52 0.062
ac harmonic filter shown in Figure 5. Below
system
2.5 1.62 1.31
the 5th harmonic of the supply frequency the filter
1.51 1.25 1.28 1.43 0.132
3.0
appears like a capacitor at any single frequency. If
1.39
3.5 1.45 1.24 1.28 0.014
the ac system at these lower frequencies is considered
4.0 1.37 1.21 1.23 1.33 0.090 to be purely inductive (i.e. synchronous machines) the
4.5 1.29 1.16 1.18 1.27 0.080 system impedance characteristics can be added as shown
5.0 1.31 1.17 1.18 1.26 0.069 (the dotted lines). Two such characteristics are
6.0 1.25 1.14 1.15 1.22 0.062 indicated at system short-circuit ratios of 3 and 6.
7.0 1.25 1.15 1.15 1.23 0.057 Where the ac system impedance characteristic crosses
the capacitative part of the filter characteristic
Fig. 9. The effect of varying the ac system impedance there is a resonant condition between the ac system
keeping the impedance angle at the third har- and the harmonic filter.
monic of the fundamental constant at 750.
Each result was obtained from 1000 energisa- the characteristics shown it can be seen that
From
tions of a star-delta connected transformer. resonances can be anticipated at the third harmonic of
the supply frequency with a system having a short-
circuit ratio near 3, and at the 4th harmonic for a
System imdance Maximum voltage1 voltage 501 level 22 level Standard system impedance just in excess of 6. When adding the
1 &Vgle puU u pu | pu |devintlon effect of ac system damping the characteristics are
85° 2.17 1 .157 1 .1611 2.011 0.275 modified a little.
1.71 1.36 1.42 1.61 0.181
80 With ac system and ac filter combination reso-
the
750 1.51
j
1.25 1.28 1.43
j
0.132 nant say the 3rd harmonic of the supply frequency
at
70 1.41 0.103 any shock can cause the system to ring at its natural
frequency. The magnetizing inrush currents can provi-
Fig. 10 The varying the system impedance
effect of de the required shocks, and occurring once per cycle
angle (defined at the third harmonic of the tend to maintain the ringing. The diagram shown in
fundamental) with the system impedance kept Figure 13 is an equivalent circuit of the overall sys-
constant at a S.C.R. of 3. Each result is tem at one frequency (e.g. the third harmonic of the
obtained from 1000 transformer energisations. supply frequency).

491
 Significance of the Overvoltages
Equivalent A mechanism has been demonstrated for generation of
Rs
transformer many peaks of twice the busbar voltage for lightly
jxs --. r-
-- -..-
damped systems. If the fundamental busbar voltage is
temporarily 1.25 pu following a load rejection say,
peak voltages of 2.5 pu for many cycles can appear at
-jXF: the ac terminals. Referred to the dc voltage rating
of a typical convertor bridge the peak voltage appea-
ring on the dc side can be of the order of 1.3 x 2.5 =
3.25 pu. As surge diverters can not be used to pro-
tect against many repeated peak voltages this figure
may represent the migimum sparkover setting of the dc
side surge diverters . The ratio of minimum to maximum
sparkover settings are of the order of 1.25 for a ty-
Fig. 13. Equivalent circuit of the system of Fig. 4 on pical surge diverter. The protected level for the
energising the transformer-. plant is therefore of the order of 4.1 times its dc
voltage rating, and it is necessary that the plant be
able to stand this with a factor of safety.
The saturating transformer can be regarded as a METHODS OF REDUCING THE OVERVOLTAGES
short-circuiting switch which tends to discharge the
filter capacitance through its own saturated inductan- Selecting Filter Impedance
ce XT, once per cycle, and then allows the filter to
recover against the instantaneous supply voltage. As One apparently obvious way to reduce these overvol-
the inrush current peak (and saturation angle e- Figu- tages is to select the filter impedance to avoid re-
re 1) reduce, the current zero of the inrush current sonance.
tends to shift relative to the ac supply voltage until
the filter capacitance recovers against the crest Consider a rectifier station supplied by generators
value of the fundamental component (1 pu) and expe- and with no local load. The weakest systems envisa-
riences an overswing of about 1 pu, giving a total ged, as yet, have impedances of 0.4 pu at the funda-
voltage at the ac busbars of about 2 pu. The peak mental frequency of the supply, or approximately 1.2
overvoltages can exist for a number of cycles, but pu at the third harmonic, at rated dc power. If
then the magnitudes of the peaks reduce as the magne- however the dc power is reduced to its minimum value
tizing inrush current reduces. of the order of 0.1 pu, generators could be removed
from the ac system effectively increasing the impedan-
This explains the slow build up and decline of the ce of the ac system to a maximum value of 12 pu at the
overvoltages as shown in the oscillograms of Figure 3rd harmonic of the supply frequency. In order to be
12. sure of avoiding a third harmonic resonance with the
ac system the ac filter would have to have an impe-
The two peaks in the 2% overvoltage versus system dance in excess of 12 pu at the third harmonic of the
impedance curve (Fig. lla) can be explained by the ac supply frequency. This is a completely impractical
harmonic filter being in resonance with the system at filter.
the 3rd and 4th harmonics of the supply frequency, for
system short-circuit ratios of 3 and 6 respectively. It may be possible to switch sections of the filter,
if there is enough information available about the ac
system, to avoid resonances. It should be realised
It can be seen from Figure 12 that the maximum
that resonances at 2nd, 4th, or other harmonics of the
overvoltages can occur at relatively small magnetizing supply can occur in a similar way and also have to be
inrush currents as the width of the current pulse is avoided.
broadened by distortion at the ac bus, and by the pre-
sence of delta windings. Therefore any transformer Damped Filters
energisation that gives rise to magnetizing current
pulses in excess of these critical small values has The effect of ac system damping is demonstrated in
the potential to give rise to the large overvoltages. Figures 11 and 12. Similar effects can be produced
This explains the large concentration of high overvol- using damped ac filters. However the filtering effi-
tages in the histogram of Figure 6. The other concen- ciency of such filters is relatively low, and the cost
tration of events at the low overvoltage end is and losses are high.
explained by the energisation of the transformer at a
point in the cycle corresponding to the earlier AC Side Surge Diverters
de-energisation, which then gives rise to zero or
small overvoltages. As there is a tolerance in the Typical surge diverters may have a sparkover to
timing required for this (corresponding to the diffe- reseal ratio of about 1.6. If the surge diverter is
rence between the peak working flux B and the satura- to reseal against dynamic overvoltages of the order of
tion flux B ) a reasonable proportion of the events 1.25 pu it can be seen that the minimum guaranteed
fall into this area. protection level with no factor of safety to avoid
spurious sparkover is 2 pu. Such a surge diverter
From Figure 11(b) it is seen that the overvoltages would have to be capable of sustaining many cycle by
are very dependent on the amount of system damping cycle operations (i.e. the mechanisms exist for produ-
assumed. While a damping angle of 85 might be rea- cing many overvoltages in excess of the minimum break-
listic (if not a little pessimistic) for an isolated down level of the diverter).
rectifier supplied solely by synchronous machines with
a very light local load, such a damping angle would be
Switching Resistors
extremely pessimistic for the receiving end. A dam- It is possible to increase the decay rate of the
ping angle of 750 for the receiving end might be more
appropriate. magnetizing inrush currents of the transformer by in-

492
serting series resistors on energising the transfor- Following complete load rejections or ac system
mers, and therefore reduce the number of overvoltage faults these voltage levels can be increased by a fur-
peaks to within the capability of the surge diverter ther factor of 1.25.
(allowing lower protection settings). The energy
rating of such resistors would have to be very large The presence of local load introduces damping which
to be useful. can considerably reduce these overvoltages. Curves
are included to illustrate this effect.
Timed Transformer Energisation
Techniques are suggested which may lead to ways of
If the transformer energisation could be timed to reducing these overvoltages or the frequency of these
coincide with the point on wave of the original overvoltages.
de-energisation little or no overvoltages are genera-
ted (Figure 6). REFERENCES

An alternative might be to bias the transformer by (1) L. F. Blume, "Transformer Engineering", New York,
injection of dc into the primary winding of the trans- Wiley, 1938, pp. 23-36.
former prior to energisation in order to ensure the
correct polarity of the remanence flux. The energi- (2) R. L. Bean, N. Chackan, H. R. Moore, E. C. Wentz,
sing breaker could then be timed for generation of "Transformers for the electric power industry",
minimum overvoltage as above. McGraw-Hill, 1959, pp. 103-109.

Such techniques would have to be applied to all (3) T. R. Specht, "Transformer Magnetizing inrush
transformers electrically close to the ac busbars con- current", AIEE Trans. Vol. 70, pp. 323-328, 1951.
sidered. In an isolated rectifier station this might
be possible. (4) T. R. Specht, "Transformer inrush and rectifier
transient currents", IEEE Trans. Vol. PAS-88, No.
ADEQUATE RATINGS 4, pp. 269-274, 1969.

In an isolated rectifier station with little local (5) L. F. Blume, G. Camilli, S. B. Farnham, and H. A.
load the generation of high overvoltages on energising Peterson, "Transformer magnetizing inrush currents
transformers is quite likely. It has been shown that and influence on system operation", AIEE Trans.
2% of energisations can result in overvoltages in Vol. 63, pp. 366-374, 1944.
excess of 2 pu rated voltage for a system having a
short-circuit ratio of 3 and an impedance angle of 850 (6) W. K. Sonnemann, C. L. Wagner, and G. D.
at the third harmonic of the supply frequency. Rockefeller, "Magnetizing inrush phenomena in
transformer banks" AIEE Trans. (Power Apparatus
It is essential that the plant has an adequate and Systems) Vol. 77, pp. 884-892, October 1958.
rating to withstand these overvoltages. The most sen-
sitive parts of the dc convertor station are the con- (7) H. C. Barnes, "Preliminary analysis of extensive
vertor switching elements. Many voltage peaks of 3.25 switching surge testing of American Electric
times the rated dc voltage may occur across the con- Power's First 765 kV Line and Stations", IEEE
vertor valves in the worst case. Mercury pool valves Trans., March/April 1971, PAS-90, No. 2, p. 785.
must be capable of holding off these levels of voltage
but it is probably acceptable that they temporarily (8) R. W. Flugum, "Operation of lightning arresters on
break down for the higher voltages. However, a abnomal power frequency voltages", IEEE Trans.
thyristor valve must be capable of withstanding the 1970, PAS-89, No. 7, p. 1444.
highest voltages without failing.
(9) J. P. Bowles, L. A. Snider, R. L. Vaughan, "An
The maximum guaranteed surge diverter protection integrated computing facility for power system
level could be of the order of 4.1 times rated bridge studies, comprising a transmission simulator, a
voltage, to which a safety margin should be added digital computer, and an analogue computer",
Canadian Communications & EHV Conference, Procee-
dings, November 1972, p. 126.

CONCLUSIONS (10) J. P. Bowles, "AC system and transformer repre-

sentation for hv dc transmission studies"', Trans.
In an HV DC system resonances may occur between the IEEE PAS-89, No. 7, 1970, 71 1603.
ac system and the ac harmonic filters at particular
harmonics of the ac supply frequency. These resonan- (11) L. A. Bateman, R. W. Haywood, R. F. Brooks,
ces can be excited by magnetizing inrush currents "Nelson River DC Transmission Project", IEEE
which occur on transformer energisation, and can give Trans. PAS-88, No. 5, May, 1969.
rise to overvoltages at the ac busbars.

In the case of a rectifier station supplied direc- ACKNOWLEDGEMENTS

tly by generators' with very little local load 2% of
the overvoltages generated by transformer energisation I would like to express appreciation to my collea-
can be in excess of twice rated voltage in a typical gues at the Hydro-Quebec Institute of Research for
system having a short-circuit ratio of 3 (i.e. short- their help, particularly to Dr. L. A. Snider, and Mr.
circuit MVA/dc rated power = 3). J. Lemay who were responsible for developing the
digital programs used for the simulator-hybrid compu-
ter interactions.

493
 Discussion filter re-energisation before bridge de-blocking (c) as (b) without the
re-energisation of the filter arms, when de-blocking to produce 1 2-pulse
Casjen F. Harders (Erlangen, Germany): The author is to be com- operation with existing 6-pulse conversion. Comments by the author
mended on this in-depth study on effects of transformer inrush would be appreciated.
phenomena. Should the above procedure be judged inappropriate, the use of
On energization, the magnetizing flux added to the residual flux switching resistors appears to be the most practicable alternative.
(which is assumed to be 80% of rated flux) may considerably exceed Finally, the author may wish to clarify the use of a simple series
the saturation flux; this in turn is assumed to be around 130% or more R-L network to represent the ac system and to specify the Q-values of
of rated flux. the filters used in the tests.
For the worst case consideration this would mean, that an inrush
current approaching the order of magnitude of a fault current would be REFERENCE
generated, were it not for the effects of hysteresis, source resistance and [1] J. Reeve and T. Subba Rao, "Dynamic analysis of harmonic inter-
other damping as well as the fact that the residual flux never attains action between ac and dc power systems", IEEE Trans. paper
100% rated flux. no. T73 411-5 presented at Summer Power Meeting, Vancouver,
The ferroresonant phenomena initiated by this magnetic dis- July 1973.
turbance will be largely determined as to their form of occurrence by
the loss-frequency relation of the system components and, as to their
duration by the aggregate damping losses of the system.
Would the author care to comment on the possibility of sustained J. D. Ainsworth (G.E.C. Ltd, Stafford, England): The most important
overvoltage phenomena following inrush and occurring above the fifth results given in the paper are the exceptionally high overvoltages pos-
harmonic when frequency dependent losses in a system usually increase? sible at a rectifier station when switching, on transformers in the pres-
This would primarily concern a DC station with a negligible local load. ence of ac filters. This was discovered some years ago in designing the
Nelson River and Kingsnorth Schemes, and is an important design cri-
Manuscript received July 23, 1973. terion, particularly for its effect on valves and their protection. Mer-
cury-arc valves may temporarily break down, but this is not very serious.
However with thyristor valves the overvoltage figures are alarming, and
A. Ekstrom and G. Flisberg (ASEA, Ludvika, Sweden): We would first the author does not appear to offer much consolation to thyristor de-
congratulate the author on his interesting paper. This paper is of special signers. In fact from the paper the only important alternatives appear to
interest to us as we ourselves have initiated fairly extensive studies, both be either to use an uneconomically large number of thyristors in series,
on analog simulators and digital computers, with special emphasis to or to use surge diverters as self-sacrificing short-circuiting devices. An-
compare the results from these studies with oscillograms recorded at other possibility which is technically almost ideal (except for its finite
HVDC substations in operation. cost) is to add a constant-voltage shunt saturated reactor, e.g. of the
The only case where overvoltages due to transformer magnetizing treble-tripler type.
inrush current might be a problem seems to be for substations con- I would like to mention an allied phenomenon which may be
nected only to generator stations. When an AC network also is con- equally serious. As the author mentions, fundamental frequency voltage
nected, the damping seems to be so efficient that almost no overvoltages due to 100% load rejection can reach 1.25 (or 1.3) per unit at a rectifier
of this special type have been recorded. However, it is highly desirable supplied from isolated generators. This is enough to cause partial satura-
that the power utilities can continue to measure the AC bus voltage tion of transformers, which then have a moderate magnetising current,
during energization of transformers, AC filters and at various faults that containing all odd harmonics, which does not decay naturally. With an
can cause overvoltages, as extensive simulator investigations are of only ac filter containing arms for harmonics of orders 5, 7, 11, 13, etc. there
limited value if it cannot be shown that the simulation is correct. is no problem since the other odd harmonics (3rd, 9th, etc.) are ab-
Regarding the simulation of transformer energization performed sorbed by delta windings. But if filter arms for 5th and 7th harmonics
by the author, the results seem in principle to agree with the results are omitted, which is a strong temptation for thyristor schemes oper-
achieved by us although we have not performed such a large number of ating in 12-pulse groups, the combination of system impedance plus
tests. The histogram presented by the author is really very interesting. remaining filters may well be in resonance near to 5th or 7th harmonic,
However, we think that one important item of information is missing in and large steady state overvoltages can then occur at one of these fre-
the paper. It is for instance not only the maximum amplitude of the quencies, giving a total peak voltage of 2 p.u. or more, until either
overvoltages that is essential, but also the number of cycles during machine AVR's or tapchangers reduce voltage. As in the author's paper,
which the overvoltages can be higher than the sparkover levels of the this time may be sufficiently long to cause damage to surge diverters.
arresters that are used for protection of the converters.
Regarding the method of reducing the overvoltages we have partly Manuscript received August 3, 1973.
arrived at different conclusions. We have for instance found that the
overvoltages at transformer energizations can be completely avoided by
the use of preinserted resistors in the transformer breakers. The power W. A. Derry, D. Woodford, (Manitoba Hydro, Winnipeg, Manitoba) and
rating of the switching resistors does not seem to constitute a problem. R. W. Haywood (Atomic Energy of Canada Limited, Winnipeg, Mani-
That does not mean that the overvoltage problem can be forgotten toba): The author is to be commended on this paper and his approach
as similar overvoltage can occur at other events than operation of the to the problem of switching surges at dc converter stations. However,
transformer breaker, e.g. at certain blocking or deblocking events of the the question arises whether the impedance characteristic of the har-
convertors, disconnection of earth faults in the AC network etc. How- monic filter as shown in Figure 5 can be used to effectively predict
ever, the studies so far performed indicate lower overvoltages for these resonant conditions between the ac system impedance and the filter.
events. The number of overvoltage peaks having amplitudes higher than With the dc system in operation, the impedance looking into the dc
the sparkover levels of the arresters will also be considerably reduced. station from the filter busbar must be considered in addition to the ac
Ordinary arresters, or arresters with a slightly increased energy capa- system impedance and converter transformer magnetizing reactance.
bility, can then be used for protection without having to increase the DC station switching surge studies conducted by Manitoba Hydro on a
protection levels of the arresters. dc simulator have shown that maximum switching surges from trans-
former energizations with the dc valve groups on full load do not cor-
Manuscript received August 2, 1973. respond to conditions of harmonic resonance between the ac system
and filter impedances. Were the studies reported in this paper conducted
without the dc system in operation? Would the author comment on the
J. Reeve (University of Waterloo, Waterloo, Ontario, Canada): The effect of commutation in operating valve groups on switching surge
author has provided much insight into a phenomenon of recurring magnitudes with converter transformer energizations, and in particular,
speculation over several years. In common with harmonic overvolt- whether these worst conditions occur with the dc system in operation
ages [1], it is largely a result of system resonance behaviour, the ac at peak or light loads.
harmonic filters playing a major role. The maximum switching surge results with a probability of occur-
The possible methods described in the paper for reducing the over- rence number associated with each are much more useful than merely
voltages appear to be sound and naturally assume that the filters are in stating a maximum value. We do note, however, that the paper does not
themselves not amenable to modification. In applications where the indicate how many peaks above a certain value occur within the same
resonance conditions as described could apply, removal of the 5th and switching operation. This is of importance when choosing arresters.
7th harmonic filter arms would probably aleviate the problem. This Does the author plan to incorporate this feature as well as including a
could be achieved by (a) operating with exclusively 12-pulse converter probability value to its occurrence.
units. (b) switching out the 5th and 7th (or all) filter arms prior to The paper suggests that the protected level for the plant might be
energising the transformer for a single 6-pulse bridge start followed by of the order of 4.1 times its dc voltage rating. This appears high when
Manuscript received August 2, 1973. Manuscript received August 3, 1973.
494
compared with existing systems. Would the author comment on the excess of 60. The relatively simple 'T' representation of the ac system
differences? has been used in the tests because it has been found to give useful re-
IThe introduction of the paper infers the possibility of controlling sults, and can give a means whereby results from different simulators
dynamic overvoltages by use of transformer tap changers. One would can be compared. We have also developed elaborate electronic micro
suggest that control of these overvoltages would be accomplished by machines ,1 vfor more in depth representation of the ac systems.
means of rotating machines or static voltage control devices. Mr. Ainsworth points out that the overvoltages discussed in the
A specific situation where transformer energizing would take place paper have been used in the design criteria for'the Nelson River and
during the period immediately following a load rejection (i.e. during a Kingsnorth schemes. The application of the constant-voltage shunt
temporary 1.25 pu fundamental busbar voltage) is difficult to visualize. saturated reactor appears very promising for reducing the overvoltage
Does the author feel that the cost of providing insulation for this situa- problem, and should be studied in depth.
tion is justified? If this situation is to be catered for, then the effects of The problem of transformer saturation on load rejection is ex-
large frequency excursions must be considered. tremely important and is directly allied to the magnetizing inrush prob-
The paper suggests methods of reducing the overvoltages. One is lem. If the 5th and 7th harmonic filter arms are omitted resonances be-
to change the system impedance of an isolated generation - rectifier tween the' ac filter and system can occur and be excited by the mag-
complex by temporarily reducing the number of machines on the sys- netizing inrush currents, and the steady state harmonic currents due to
tem while transformer energizing is accomplished. The ac filters would overexciting the trans-former (caused by regulation effects) at har-
presumably still be on the bus, and therefore the possibility of a self- monics up to that of the lowest tuned arms (i.e. 1 1th harmonic).
excitation situation arising must be examined. Mr.'Derry, Mr. Woodford and Mr. Haywood made a number of
comments and asked several questions which will be replied to in-turn.
The resonances between the ac system and filter are defined (for
harmonic purposes) by the sub-transient impedance of the generators
J. P. Bowles: I would like to thank the discussers for their comments at the sending end. As this sub-transient impedance varies on a function
and interest. In reply to the question by Mr Harders concerning the of load it is not surprising to find that the system resonances vary with
possibility of sustained voltages above the 5th harmonic I would like to the d.c. line loading. In order to define these resonances the sub-
refer to the curves of fig. 1. It can be seen that as the system im- transient impedance of the system as a function of loading is required.
pedance reduces (i.e. the system short circuit ratio increases) the over- To avoid this problem the model tests were carried 'out- with the sys-
voltages reduce. The overvoltages are seen to be a maximum for a sys- tem blocked. With the electronic machine 1,2 we will be able to repeat
tem resonant at the 3rd harmonic (S.C.R. = 3.0 for the system con- these tests and to show the effects of machine saturation etc.
sidered) and are much lower for a system resonant at the 4th harmonic The discussers ask about the possibility of monitoring multiple
(S.C.R. = 6.0). It is possible therefore that in systems not having 5th peaks. As explained earlier this is already possible but much more in-
and 7th harmonic filters that the overvoltages for systems with reso- formation is necessary about the surge diverter performance before it
nances above the 4th may be smaller, but as the filter impedance char- can be adequately applied.
acteristic is different this is not obvious. The discussers question the figure of 4.1 times the d.c. voltage
It should be appreciated that the system representations used in rating that might under certain circumstances be used for the pro-
the tests take into account the frequency dependent losses. The reason tected level of converter plant. The purpose of this paper has been to
for the T representation for the ac system, is to provide a system in attempt to give a probability figure for the 2% overvoltage due to trans-
which the impedance angle of the system remains approximately con- former-energisation. Th'e probability of load rejection plus transformer
stant with frequency (i.e. the losses are proportional to frequency). It energisation can be assessed and then the probability of such a condi-
should be appreciated that these overvoltages are very dependent on tion producing a significant overvoltage can be applied. The risk in set-
the losses assumed (fig. 11 b). ting the protective levels can than be assessed for the anticipated condi-
Mr. Ekstrom and Mr. Flisberg agree substantially with the results tions. This might giye a required protection level of 4.1 times the d.c.
of the paper but point out that an indication of the number of peaks voltage, but this would depend on the case considered and the risk
above a certain level would be useful. I agree and in fact we are carrying accepted. It should be realised that the B.I.L. of the mercury arc valves
out work in this direction as such information is extremely important used in the Nelson River Scheme' is 650 kV (i.e. 4.33 times the d.c.
for surge diverter rating. However, such information is difficult to pre- bridge rating of 150 kV).
sent and the application to surge diverter rating is not straightforward Considering the control of overvoltages by means of tap change
as one might anticipate that the surge diverter characteristics will controls; the point was made in the paper that in the absence of syn-
change with the number of consecutive strikes applied. I would agree chronous compensator or machine controls (or other such voltage con-
that the application of pre-insertion resistors might alleviate the prob- trol devices) that regulation overvoltages would exist for times corre-
lem (as mentioned in the paper), but the problem still occurs with sponding to tap changer operations.
fault clearing (although less frequently), where it has been our experi- I would agree with the discussers that self excitation of the ma-
ence that the overvoltages can be severe due to regulation effects being chines would have to be considered if filter switching were employed
added to' the magnetizing inrush effects. to avoid system resonances.
I would agree with Dr. Reeve that it is possible that removal of the
5th and 7th harmonic filters may tend to reduce the magnetizing in- REFERENCES
rush generated overvoltages, as discussed earlier. Consideration is being
given to complete removal of the filters of an isolated generating sta- 1. J. P. Bowles, G. Jasmin, "An Electronic Machine in Power System
tion thereby removing this particular problem completely. Switching Studies", IEEE, Montreal, October 1972.
the 5th and 7th filters might give rise to operational problems and
would not avoid the problem caused by the clearing of ac faults. 2. J. P. Bowles, L. A. Snider,' R. L. Vaughan, "An Integrated Com-
The Q values of the filter components used in the tests were in puting Facility for Power System Studies Comprising a Transmission
Simulator, a Digital Computer and an'Analogue Computer, IEEE,
Manuscript received October 29, 1973. Montreal, October 1972.

495