312 Zhang and MacAlpine: A Phase-related Investigation of ac Corona in Air

A Phase-related Investigation of ac Corona in Air

C.H. Zhang and J.M.K. MacAlpine
Electrical Engineering Department
The Hang Kong Polytechnic University
Kowloon, Hong Kong, China

ABSTRACT
A computer-based partial discharge detector has heen used to measure ac cnrona
pulses in air in a point-cup electrode system. The pulse height distribution was
measured as a function of phase angle and pulse height for a range of voltages. The
nnset voltages for both positive and negative half-cycles are the same, a n d signifi-
cantly lower than for the dc case, due to space charge effects. In all other respects
the patterns for the two half-cycles are different: pulses only nccur a t the nnset
voltage (rising) for the positive half-cycle, but continuously for the negative half-
cycle, except for the higher applied voltages which exhibit a glow discharge transi-
tion near the (negative) peak. Differences between the distributions for avalanches
and streamers increase as the breakdown voltage is approached. Qualitative expla-
nations of the patterns and behavior are given, together with calculations showing
that space charge from one cycle can linger in the gap between half-cycles.
Index Terms - ac corona, plhase-related pulses, partial discharges.

1 INTRODUCTION heights and frequencies will vary with the voltage during
ORONA studies have been undertaken for many the half-cycle. Consequently the interpretations and de-
C years, not only because of the scientific interest in
corona mechanisms but also because of their practical en-
ductions are inherently limited.
Isa and Hayashi [4] were able to observe that positive
gineering importance. The majority of published studies is and negative onset voltages were virtually identical as well
for point-plane electrode systems and direct voltages as noting the large difference in positive and negative pulse
rather than ac. This is because, at least initially, dc sys- frequency. They surmised that space charge from preced-
tems provide a simpler system to understand although ing positive half-cycles would affect corona in the next
perhaps of less immediate practical application. ac corona positive half-cycle, but did not pursue the suggestion fur-
may also have the added complication of varying suffi- ther.
ciently slowly for ions to provide a “memory”effect such In his detailed paper on the stochastic aspects of partial
that the behavior at an instantaneous voltage differ:; from discharges (PD)[l], van Brunt was mostly concerned with
that of the same dc voltage [l].Studies of ac corona mostly partial discharges in voids within solid and liquid insula-
ignore the variation in voltage with time [Z]. Thus Peak‘s tion, giving phase-resolved data for certain such cases. For
well-known work [31 concentrated on determining corona ac corona discharges he points out that ‘when viewed over
onset voltages and power losses for overhead lineis with an entire half-cycle, the statistical distributions of pulse
various conductor diameters and various degrees of per- amplitudes or pulse shapes, for example, will represent at
turbing factors such as surface roughness, humidity, rain best averages of the equivalent dc behavior over a range
and pressure, but without distinguishing between half- of voltages from onset to the ac maximum’. At worst, they
cycles. will also he strongly affected by memory effects, as posited
More recently there have been several investigatitms of by Isa and Hayashi [4], due to the space charge from one
ac corona in air [e.g., 1,4-61, in which positive and nega- half-cycle affecting the phenomena of the next half-cycle.
tive onset voltages were measured as the peak ac voltages Examples of results which are affected by being averages
at which positive and negative pulses first appeared; and of the range of voltages in a half-cycle, and probably of
where average pulse heights or frequencies were given for memory effects, are the work by Al-Arainy and co-workers
the whole positive or negative half-cycle, for variouis volt- [51 and Florkowska and Wlodek [61.
ages. It follows from equivalent dc results that the pulse
van Brunt [l]gives equations which could he used to
calculate the movements of positive ions in the gap and
Monsxnpt receiwd on I July 2001, in finolform I I December 2002. hence indicate whether the phenomena in one half-cycle

1070-’9878/1/$17.00 0 2003 IEEE

IEEE Transactions on Dielectrics and Electrical Insulation Vol. 10, No. 2; April 2003 323

can affect the next. Although he did not carry out the
calculations in that paper it is suggested that these would
explain the differences he found between the ac peak on-
set voltages and the equivalent dc onset voltages in earlier
work with sulfur hexafluoride [7]. It is of course important Hemispherically-tipped
-
to distinguish between these relatively long-term ( 10 ms)
memory effects between half-cycles from the short-term
-
ones ( 100 y s ) whereby the time between two pulses af-
fects the height of the second pulse which in turn affects
the time delay before the next pulse [8,91.
In later work by Florkowska and Wlodek [lo], phase-re-
solved measurements are used t o show the variation in
pulse height with phase angle and how this pattern changes
as the voltage increases: increasingly around the peak
voltage there is a transition from pulsed to pulseless
corona.
In the present work, a computer-based commercial PD
detector was used to obtain phase-resolved corona data in
air, which could then be analyzed in terms of pulse height
and frequency with respect to phase angle. This approach
is new and necessary for increased understanding of ac RoW~OMIsyhmehy axis
Figure 2. The hemisphere-hemisphere point-cup electrode system.
corona. It allows a much more complex and enlightening The pointed electrode’k tip is concentric with the concave ground
analysis as the changing pattern may be observed through- electrode.
out the cycle and the changing effects of space charge
throughout the cycle to be inferred. It also enables the
onset voltages to he measured as instantaneous voltages
commonly used point-plane electrode system in respect of
during the half-cycle over a range of peak voltages, rather
understanding and modeling the complex mechanisms of
than as the peak voltage when corona is first observed.
corona [11,121.
Calculations are then introduced indicating how the ions
created in one half-cycle can persist in the gap until the A PD detector has the advantage of providing pulse
next half-cycle. height, number and phase angle analysis quickly and eas-
A point-cup electrode system [11,121was used as its high ily with built-in calibration. It also provides a variety of
symmetry gives clear advantages for analysis of data over methods of displaying the information. However there are
the more commonly used point-plane electrode system. some disadvantages stemming from the inability to access
the raw data.
2 EQUIPMENT The Haefely TE571 PD detector can he set to detect
A computer-based commercial PD detector, the Hae- pulses in ranges characterized by an upper limit Q,, that
fely ‘E571 (Figure I), was used for detecting and analyz- can he set to values from 1pC to 10 nC. The lower limit is
ing the corona pulses obtained when an ac voltage was automatically set at 16% of Q,=. Each 360” cycle is auto-
applied to a point-cup electrode system [7-91 in air. This matically divided into approximately 290 time “windows”
electrode assembly comprised a rod with a hemispherical by the software, which are therefore each of approxi-
tip, which is concentric with a n earthed hemispherical mately 69 ys (20 ms/290). Each pulse-height range is di-
concave “cup” (Figure 2). This arrangement has clear ad- vided into about 50 cells according to pulse height. Thus
vantages, because of its high symmetry, over the more each pulse within the set range (16% to 100% of Q,,,)
adds a “1”to the count in the appropriate cell of the 290
x50 database. These databases can he accessed and
opened in Excel. Data are accepted for 2 minutes (mini-
mum setting) so that 6000 cycles are sampled and minor
0 to fluctuations (and much of the noise) are removed.
100 kV rms
It appears that some near-neighbor weighted averaging
. .-
P\IF
Figure 1. Experimental set-up for detecting corona discharge pulses
(smoothing) then takes place; and that a number N is sub-
tracted from the number in each cell (and cells which be-
and analyzing them with respect to size and phase angle. CC, cou-
pling capacitor; CQ, coupling quadripole; PVE,prcssure vessel with come negative are set to zero). N appears to he about 30
electrode system; TE571. Haefely E 5 7 1 PD Detector. to 50% of the largest cell number 1131. This has the effect
314 Zhang and MacA[pine:A Phase-related Investigation ofac Corona in Air

of largely eliminating noise and, by analogy, is equivalent

to raising the “sea-level” in (for example) Figure 3, thus ,
( 12kV
“flooding” the minor valleys and hills. This signal condi-
tioning is built-in and cannot be altered. (These signal-
conditioning details have been deduced from usage of the
equipment: the information is not currently available from
the manufacturer).
The displayed range is generally less than the se(.range,
apparently being from the 16% threshold to the highest
charge window for which data is held, normally consider-
ably less than the set value of Q,,. Thus in Figure 3,
Q,, was set to 10 pC, but the displayed range is from
n
zero to about 5 pC with pulses shown only from 0.8 pC
upwards. A minor disadvantage for these “three-dimen-
siona1”graphs is that the scale i s automatically set and is
generally somewhat unusual: the pulse-height scale for the
20kV results in Figure 3 is 0.09, 1.2, 2.5, 3.7 and 1.9 pC,
for example.
It should be noted that only Q,, can be varied by the
operator as all other parameters are fixed or automati-
cally chosen and set by the software.
A1 16kV

PDs are most commonly detected in voids within solid

insulation by the pulses induced in an external circuit. Be-
cause of the capacitance ratios involved the pukes de-
tected are orders of magnitude smaller than the charge
movements occurring in the actual discharge. Relatively
few discharges occur per cycle (typically 1 to 5). 13y con-
trast, the discharges detected in the present case, the
avalanches and streamers between the electrodes, can oc-
cur in huge numbers. Consequently the system, which is
designed for the lower PD levels typically found cm elec-
trical plant, would, at times, have been “saturated’; more
pulses occurred than could he counted: This is because
only one pulse, the first, can be registered during each
69-ps “window”.
For the purpose of commercial discharge detection in
high-voltage equipment there is a need for a clear and
unambiguous display of the data and as a result consider-
able signal processing is evident in the PD detector, as
described above. Nevertheless v e q useful data can be ob-
n 20 kV

tained indicating when pulses occur during the cycls, when

pulses are more numerous and how large (pC) they are.

3 RESULTS
ac corona data were obtained by the above methods for
Figure 3. 3-dimensional displays of pul&~&bers vs size and ohase
air at a humidity of 50% or less for voltages ranging from angle far voltages from 12 to 20 kV,,
12 to 20 kV,,, and with a variety of pulse size ranges.
Figure 3 shows typical results obtained for a range: of ap-
plied (rms) voltages with the range set to a Q,, value of electron charges. The voltage range during which the
10 p c . avalanches occur is extremely short, less than 1 kV, pre-
The onset voltage for the avalanches in the positive cisely as for the dc case [121. As may be surmised from
half-cycle is consistently close to 11 kV (Figure 51, lower Figures 3 and 4, the onset is very sharply defined. The
by 2 kV than the case for positive dc corona [121. They are total number of pulses recorded during the 2 minute pe-
avalanches, as a 1 pC pulse is equivalent to about 0.6 x lo7 riod for each 69 ps time window could readily be obtained
ZEEE Transactions on Dielectrics and Electrical Insulation Vol. 10, No. 2; April 2003 315

12kV

A 14kV
Wbdvolaoes. W m r
Figure 5. Onset voltages for positive and negative half-cycle pulses
and extinction voltages for negative half-cycle pulses from the data
displayed in Figures 3 and 4.

was there an “intermediate” window with less than 10

A 16kV
.*
“.
pulses recorded. The criterion for onset was taken as the
first time window with 10 or more pulses recorded. Nor-
mally this was the first time window with any pulses
recorded. No doubt the signal conditioning referred to
earlier is a factor in this lack of ambiguity.
Figure 4 differs from Figure 3 in that Q, was set 100
times higher, at 1000 pC, so the pulses are typically at
values around 200 pC, corresponding to about lo9 elec-
tron charges, and may thus be considered as representing
18kV streamers. Again the positive pulses only occur at instan-
+
taneous voltages close to 11 kV, albeit only for applied
voltages of 18 kV,,, and up. No further discharges occur
until the negative half-cycle, when, at instantaneous volt-
ages of about -11 kV, streamers begin to occur.
Avalanches are observed from 16 kV,, and upward (Fig-
ure 5).
Discharge pulses occur over most of the rest of the neg-
ative half-cycle until the voltage has passed the maximum
20kV
c and dropped back to a magnitude of around 14 kV (Fig-
ure 5). The sudden increase in the streamer onset and
extinction voltage at 20 kV seen in Figure 5 is believed to
be affected by the processing software: the large positive
streamers have caused the upper range to he set at 867
pC with the consequent cutoff at just above 200 pC, and

/i
this loss of data has probably caused the apparently higher
onset and extinction voltages. The cutoff effect is demon-
22kV strated more clearly in Figure 6, a contoured version of
the 20 kV plot in Figure 4. By contrast, Figure 7, which is
a contoured version of the 18 kV plot in Figure 4, does
not show any sharp cutoff in the pulse distribution: they
grade down more gradually to zero for lower charge lev-
els. This gives added credence to the suggestion that the
apparent jump in the onset voltage at 20 kV,, may not be
Figure 4. As for Figure 3 but for e,,, = 1000 pc.
a true representation of the situation. This effect is also
apparent in Figure 4, for the 18 and 20 kV displays. This
from the spreadsheet data: typically there were no pulses cut-off is a noise threshold set by the software and not
in several windows and then the next windows had 50 or alterable by the operator, as mentioned earlier, in Section
more, rising quickly to more than 300. Only occasionally 2.
316 Zhang and MacAlpine: A Phase-rehzted Investigation of ac Corona in Air

600

* 100

d
5- 200
.... ... :. ..
:::: .....*:. ..
.....e
5%: -3,B;3?’<:
: A:.
E
..L .,.:,>E;Lx.:!:’ ...:.:!:<.,L.E
n
180 210 240 270 300 330 380
Phe.. angt.. d.(Ir.e.
Figure 6. Contour display for 20 kV,, with a,,, set to 1000 pC.

4 DISCUSSION
4.1 POSITIVE HALF-CYCLE PROCESlSES ilk"

As with the dc case [12], the positive ion density builds

up quickly close to the positive hemispherical point due to Figure 8. Diagrams illustrating the effect of space charge on ioniza-
the avalanches and streamers approaching toward:; it with
+
tion and corona. a, 13 kV increasing (first quarter-cycle, 41”, 2 ms);
b, + 11 kV decreasing (second quarter-cycle, 146”. 8 ms); c, zero-
exponential growth rates. In the absence of space charge, crossing, pasitivc 10 negative (180”, 10 ms); d. - 11 kV decreasing
it is straightforward to demonstrate (using, for example, a (third quarter-cycle, 214”, 12 ms); e, - 14 kV decreasing (third quar-
ter-cycle, 225”. 12.5 ms); f, -20 kV negative “ p e a k (third quarter-
subset of the corona simulation software of [ll]) that an cycle, 270”, 15 ms).
avalanche approaching a positive point of 1.5 mm radius
will achieve 90% of its growth, and “deposit” 90% of the
positive ions, in the last 50 p m before reaching the sur-
face. These positive ions drift away from the electrode at tively low (Figure pa). The effect is as if the radius of the
a rate which is balanced by the rate at which positive ions point electrode were increased. Consequently the ioniza-
are produced near the electrode by avalanche growth. This tion activity decreases to the point where only sufficient
is a quasi-stable situation as the drift of positive ions away avalanches to maintain the positive ion “space charge”wi1l
from the electrode is a continuous activity, whereas the occur, and streamers cease. As the voltage increases the
arrival of each avalanche produces a large and local incre- space charge will presumably increase, extending out fur-
ment in positive ion density. ther. So far, this mimics the dc case, except that the onset
The positive ions thus form a shell or shield around the voltage is significantly (15%) lower.
point electrode with the maximum charge density close to Once the voltage begins to decline from the maximum,
the surface. The field within the ion “cloud” will be rela- the ionization activity decreases and eventually ceases
(presumably close to the dc onset/extinction voltage of 13
kV). However the positive ion shell around the electrode
persists over the 3 ms until then so the streamers do not
reappear at around 13 kV as they would do in the dc case
when the voltage is usually decreased more slowly (manu-
ally). Below 13kV the positive-ion space charge will drift
away from the electrode (Figure 8b). However, given the
low fields of the last quarter of the positive half-cycle, and
the consequent low mobility of the ions, a high proportion
will remain in the gap, only reaching a maximum distance
of some 20 mm from the pointed electrode, when the
electrode voltage reaches zero (Figure 8c). The relevant
calculations are given in the Appendix. The electrode then
becomes increasingly negative, attracting the positive ions
180 210 240 270 300 330 360 back towards the electrode. As drift velocities are propor-
PhaHt angle. d o g ” tional to the field, it follows that Figures 8b and 8d should
Figure 7. Contour display for 18 kV,, with a,,, set to 1000 pC. be identical because no ionization takes place between
IEEE Transactions on Dielectrics and Electrical Insulation Vol. 10, No. 2; April 2003 31 7

these two events and the ions’ movements should there- dc breakdown voltage is 42 kV [15]. The ac breakdown
fore be symmetrical. voltage is at 25 kV,,,, or 35 kVpCzk,showing that by the
time the maximum positive voltage is achieved, there is no
4.2 NEGATIVE HALF-CYCLE longer any effect from space charge remaining from the
PROCESSES negative half-cycle: and the thin layer of positive space
charge is the same as for the positive dc case at the same
As the voltage becomes increasingly negative, the prox- voltage, As mentioned earlier, this space charge is quasi-
imity to the electrode of the returning cloud of positive stable in that the balance is between the steady drift of
ions increases the electric field between them (Figure Xd). positive ions away from the electrode and the discrete ap-
This increased near-electrode field explains the onset of pearance of large numbers of positive ions in the form of
streamers at - 11 kV, lower than the dc value of 13 kV.
~

a n avalanche. These fluctuations allow occasional stream-

Streamers continue to form up to and past the magnitude ers to occur and these, at 34 to 35 kV, can transform into
maximum (at - 20 kV) until the voltage has risen to about leaders and lead to breakdown.
- 14 kV (i.e., reduced in magnitude to about 14 kV). Dur-
ing the streamer-forming period, their positive ions drift By contrast the activity in the negative half-cycle iS rela-
inwards, but are replaced by those deposited by the tively unchanged with voltage at first, although appearing
avalanches and streamers developing outwards towards the to decline slightly at the higher voltages, and at the higher
positive ion cloud. Both activities increase in magnitude voltages within the half-cycle. This indicates that the space
as the voltage increases, so that the pattern remains rela- charge-free zone adjacent to the electrode is decreasing in
tively unchanged during this period (Figures 8d, Xe and extent faster than can be compensated for by the increase
Xf). in ionization with voltage.
The space charge remaining after the negative extinc- It is equivalent to the transition to a glow discharge seen
tion voltage will include both positive and negative ions. for negative dc corona, and was also noted by Isa and
This is because the electrons from these streamers will Hayashi [4]. By 22 kV,,, the glow discharge occupies much
drift into the attaching region (i.e., where the field is less of the negative half-cycle with streamers only seen around
than the critical field of about 2.3 kV/mm Ill]), and form 11-12 kV (rising) and 16-13 kV (falling).
negative ions. As the voltage reverses, going increasingly
positive, the negative ions will be attracted back towards 5 CONCLUSIONS
the pointed electrode, increasing the local field and re- HE discharge pulses produced by ac corona have been
leasing their “excess”e1ectrons as a result of colliding with
molecules [3,14]. The case is analogous to that of the posi-
T measured and analyzed. They have been shown to
have characteristics, which are fairly similar to those oc-
tive ions drifting out and back during the transition from curring at similar voltages during dc testing. The differ-
positive half-cycle to negative discussed earlier, and simi-
-
lar maximum distances ( 20 mm) are involved. The posi-
ences are, however, significant and it is shown that these
can be explained in terms of the expected space charge
tive ions created in the negative streamers will be quickly
-
( 100 ps) removed from the gap after ionization ceases.
distributions during the previous few milliseconds of the
cycle. The findings and proposed explanations may be
Both the increased field due to the negative ions and their summarized as follows.
release of electrons close to the pointed electrode encour-
During the positive half-cycle, pulses appear at a sig-
age the development of streamers and hence lower the
nificantly lower onset voltage than for the dc case because
positive half-cycle onset voltage below that for the dc case,
of the proximity to the pointed electrode of the negative
as was observed. space charge pulled back inwards after drifting away to-
The effect whereby negative oxygen ions encourage wards the end of the previous (negative) half-cycle and
streamers in the positive half-cycle has been observed in the subsequent release of electrons by the negative ions.
“air”of increased oxygen content where greatly enhanced Immediately above the onset voltage, the pulses dis-
streamer numbers are seen for oxygen contents of 25% appear, again as for the dc case.
and u p [9]. As the voltage decreases after the maximum, the
pulses do not re-appear, because the space charge does
4.3 CHANGES AS THE APPLIED not have time to disperse during the decrease from the
VOLTAGE IS INCREASED peak voltage.
In the positive half-cycle, an increasing changeover from Early in the negative half-cycle, pulses appear at a
avalanche (Figure 3) to streamer (Figure 4) at 20 kV and significantly lower onset voltage than for the dc case be-
upwards may be obsewed. This is consistent with the much cause of the of the proximity to the pointed electrode of
lower dc breakdown voltage found for positive point/plane the positive space charge pulled back inwards after drift-
gaps as compared with negative. For the point/cup system ing away towards the end of the previous (positive) half-
the positive dc breakdown voltage is 34 kV, the negative cycle.
318 Zi‘mng and MacAlpine: A Phase-related Investigation of ac Corona in Air

Pulses continue through the negative half-cycle until kV/mm [16]. 13 kV was chosen for the above example as,
the voltage drops to the extinction level, as the avalanches in the dc case, there would be no ionization taking place
are moving away from the point and so the positive space below that voltage. If I 1 kV is chosen instead of 13kV, the
charge is less dense and tends to increase the near-elec- ions will only trave1 out to r = 19 mm.
trode field. Alternatively, a simplified space-charge calculation may
Breakdown occurs during the positive half-cycle as the be made in which a positive space charge is assumed to
proportion of streamers to avalanches occurring at the’on- exist near the point electrode such that the field is con-
set voltage steadily increases with applied voltage, while stant, and equal to the critical field, E, = 2.3 kV/mm, from
the negative half-cycle streamers steadily decrease in the surface at r = a out to a radius rx (where the potential
number as the glow discharge region extends. is V,, say). Then, for r > rx

ACKNOWLEDGMENTS E(r) =v,/[r2(1/rX-l/b)],

This work was supported by HK Polytechnic University
grant G-V217. (because the electrodes are concentric hemispheres). The
potential difference across the space charge will be
APPENDIX
V - V ,= E,( rx - a ) .
It is suggested in the Introduction that a longer-term
“memory”effect may cause events in one half-cycle to af-
At r = rx the field is
fect those in the following half-cycle. This require!; signifi-
cantly longer “memory”times than cause the inverse cor-
relation between pulse amplitudes and the time intervals E(r,) = E, = v,/[r2(1/rx - 1/b)]
between pulses, which have been observed with Trichel
pulses under dc conditions. The gaps between Trichel These two equations allow V , and rx to be found for
pulses are typically of 5 to 500 ps, whereas times between any applied voltage V and hence the field at any radius.
positive and negative peak voltages are 20 ms apart (f = 50 The trajectory of any positive ion may now be found in
Hz). However the drift velocities decrease to zero at the the same way as for the space charge-free case above, ex-
end of the half cycles, as the field decreases. P. simple cept that the integration is numerical rather than analyti-
calculation of the movement of a positive oxygen ion in a cal.
gap may be done as follows. Using this method for an applied voltage of 14 kV,,,,
The electric field between the electrodes in Figure 1, is and considering posjtive ions left in the gap after the volt-
given by age has dropped from the maximum to 13 kV, it is found
that the positive ions move out to a radius of 24 mm be-
E(r)=V/[r2(l/a-l/b)]; fore beginning to return towards the pointed electrode.
Here the ion’s start point is assumed to be at the edge of
where a and b are the radii of the point electrode and the the space charge region ( r = rJ For the 11 kV case, the
cup electrode, respectively. furthest point of the trajectory will be at r = 22 mm.
The drift velocity is therefore Both the “space-charge free” and “space-charge” calcu-
lations necessarily involve simplifying assumptions, but al-
U = dr/df = KE = KV/[ r 2 (l / u - l/b)] though very different give very similar answers: a max-
mum “reach” of 20 to 25 mm. The reason is that, in the
=KV,sinwf/[r2(l/a-l/b)], mid-gap region, the fields are very low throughout the cy-
cle, reducing to zero during the time of interest, and drift
where K is the ions mobility, velocities are very low: around 10 mm/ms at r = 20 mm
Integrating, r is found to be proportional to (1 --coswt), and V = 11 kV.
and the maximum in r therefore occurs at f = 10 ms, the By contrast, velocities are very high in the region close
crossover from positive to negative polarity. Foi: an ap- to the point electrode (distances of the same order as the
plied voltage of 14 kV,,, and considering the period after tip radius), so that time-scales of the order of tens of mi-
the voltage has dropped from the maximum to 13 kV, it is cro-seconds prevail, and account for the short-term mem-
found that the positive ions move out to a radius (of 22mm ory effects already referred to (Section 1) whereby the time
before beginning to return towards the pointed el.ectrode. between two pulses affects the height of the next pulse
The same result is obtained, to 2-figure accuracy, whether and vice versa. For example, drift velocities are around
the ion’s starting point is at the electrode surface ( r = 1.5 100 times higher at the peak negative voltage (when
mm) or 4mm further out. This calculation neglects space Trichel pulses are occurring): around 1000 mm/ms at r =
charge effects and assumes a mobility of 136 mm/ms per 2mm and for V = - 20 kV.
IEEE Transactions on Dielectrics and Electrical Insulation Vol. IO, No. 2; April 2003 319

REFERENCES C. H.Zhang received the BS degree in Elec-

trical Engineering from the Harbin Institute
I11 R. I. Van Brunt, “Stochastic Properties of Partial Discharge
of Technology (HIT), China, in 1985 and the
Phenomena”; IEEE Trans. DEI, Vol. 26, pp. 902-948, 1991.
MS degree in Automation and Computer
121 F. W. Peek, Diekctic Phenomena in High Voltage Engineeriig, Engineering from the Navy Aeronautical
New York, USA, McGraw-Hill, 1920. Engineering Academy .(NAEA), China, in
131 M. Goldman and A. Goldman: “Corona Discharges” in M. N. 1988. the Ph.D. degree in Control Engineer-
Hirsh and H. 1. Oskam (Editors), Gaseous Electronics, Vol. 1, ing from South China University of Technal-
New York, U.S.A., Academic Press, 1978. ogy. China, in 1996 and the Ph.D. degree in
I41 H. Isa and M. Hayashi, “Behavior of Corona Discharges in a Electrical Engineering from the Hong Kong
Rod-to-plane Gap in Atmospheric Air under ac Voltage Appli- Polytechnic University in 2003. From 1988 to 1992, he worked as an
cation”,Elec. Eng. Japan, Vol. 107, NO. 6, pp. 9-16, 1987. engineer and lecturer at NAEA and in 1996 joined HIT as an
151 A. A. Al-Arainy, N.H. Malik and M.K. AI-Bahloul, “Statistical A s x i a t e Professor. He also works as a consultant engineer. His
Variation of ak Corona Pulse Amplitudes in Point-to-plane Air areas of interest include HV engineering, substation automation
Gaps”, IEEE Trans EI, Val. 24, pp. 681-687, 1989. and electric equipment testing and protection, computer modeling
L6l B. Florkowska and R. Wlodek, “Pulse Height Analysis of Partial and simulation, intelligent control, artificial neural networks, and
Discharges in Air”,lEEE Trans. El, Vol. 28, pp. 932-940, 1993. distributed computer control for power plants.
I71 R. 1. Van Brunt and M. Misakian, “Mechanisms for Inception
of dc and 60-Hz ac Corona in SF,”,IEEETrans. El, Vol. 17, pp.
106-120, 1982.
[SI R. I. Van Brunt and S. V. Kulkami, “Stochastic Properties of Mark M a d p i n e was bom in 1942 in the
Trichel-pulse Corona: A Non-Markovian Random Point UK. He obtained the BA and MA degrees
Process”, Phys. Rev. A, Vol. 42, pp. 4908-4932, 1990. at Cambridge University in 1963 and 1964.
respectively and the Ph.D. degree at London
191 C. H. Leung, “Pulse Height Analysis of ac Corana”,Final-year
University in 1968, the latter while working
BEng/MEng Project Report, Hong Kong Polytechnic Univer-
for the research laboratories of the UK Cen-
sity, 2002.
tral Electricity Generating h a r d . He then
[lo] B. Florkowska and R. Wlodek, “Analysis of Partial Discharges taught at the Haile Selassie I University in
in Air Using Phaseresolved Patterns”, IEEE CEIDP, USA, pp. Addis Ababa, worked with International
606-609, 1996. Computers Limited in the UK and, since
I111 E. N. Li and 1. M. K. MacAlpine, “Negative Corona in Air Us- le Hong Kong Polytechnic University His re-
ing a Point/cup Electrode systcm”,IEEE Trans. DEI, Vol. 7, pp. search interests are in electrical corona, statistical aspects of spark
752-757, 2000. and lightning paths, fiber-optic sensors and the condition monitor-
1121 J. M. K. MacAlpine and E. N. Li, “Positive Corona in Air Using ing of electrical plant. He has also published bn the evaluation of
a Point/cup Electrode System”,IEEE Trans. DEI, Vol. 10, 2003 teaching quality and on peer assessment by both students and staff.
(In press).
1131 C. H. Zhang and J. M. K. MacAlpine, the Analysis of ac
Corona in Air Using a Computer-based Partial Discharge De-
tector”. International Conference an Electrical Engineering
(ICEE), Hong Kong, pp. 168-171, 1999.
I141 J. L. Moiuzzi, and D. A. Price, “loniration, Attachment and
Detachment in Air and Air-CO, Mixtures”, I. Phys. D: App.
Phys., Val. 7, pp. 1434-1440, 1974.
I151 E. Li, “Investigation of Pre-breakdown Pulsed and Pulseless
Currents for Point Electrodes”, MPhil thesis, Hang Kong Poly-
technic University, 1998.
I161 E. Kuffel, W. S. Zaengl and J. Kuffel, Hfgh Voltage Engineering:
Fundamentuh, New York, USA. Newnes, Chapter 5, 2000.