PDI Vs NQN
PDI Vs NQN
PDI Vs NQN
+ + + +
=
where:
I
t
= average current, A,
t
0
= starting time, s,
t
1
= completion time, s, and
Q
1
, Q
2
, Q
n
= partial discharge quantity in a corona pulse 1 through n, C.
3.1.13 partial discharge (corona) power loss (P)the summation of the energies drawn from the
test voltage source by individual discharges occurring over a period of time, divided by that time
period.
When pulse height analysis is used, the summation over a period of time of pulses above a
preset level of corona usually determined by background noise multiplied by the instantaneous
test voltage at the time of the pulses in the specimen is approximately equal to:
=
=
i
j
j tj j
V Q n
T
P
1
) (
1
where:
P = pulse discharge power loss, W,
nj = recurrence rate of the jth discharge pulse in pulses/ second.
Qtj = the corresponding value of the partial discharge quantity in coulombs for the particular
pulse.
Vj = instantaneous value of the applied voltage in volts at which the jth discharge pulse takes
place.
3.1.14 partial discharge apparent power loss (Pa or PDI)the summation over a period of time of
all corona pulse amplitudes multiplied by the rms test voltage.
s t a
V I P =
where:
Pa = apparent power loss in time interval (t1 - t0), W,
It = average corona current, A, and
Vs = applied rms test voltage, V.
3.1.15 partial discharge (corona) pulse rate (n)the average number of discharge pulses that
occur per second or in some other specified time interval. The pulse count may be restricted to
pulses above a preset threshold magnitude, or to those between stated lower and upper
magnitude limits.
IEEE Guide of Measurement of Partial Discharges in Rotating Machinery 1434
defines an unique integral quantity called Normalized Quantity Number (NQN)
that is not presented in any other PD standard and is used specifically for rotating
machines.
Historically, this number was introduced by IRIS . By definition NQN originally
defined as the normalized area under a straight line fitted to the pulse counts in each magnitude
window of a pulse height analysis, where the pulse counts are expressed an a logarithm of the
pulses per second and the pulse magnitude window is a linear scale. Both the pulse magnitude
and repetition rate scales are normalized and the quantity obtained is divided by the gain of the
partial discharge detector. NQN can be represented mathematically as:
+ +
=
2
log
log
2
log
10
1
2
10
1 10 N
N
i
i
P
P
P
N G
FS
NQN
where P
I
= number of pulses per second in magnitude window i
N = number of magnitude windows
G = gain of the partial discharge detector
FS = maximum magnitude window in milli-volts at unity gain
Pulses with a high repetition rate have less of a contribution to the NQN than do pulses with low
repetition rate. It should be also noted that even if the normalizing process is followed, the value
of NQN for the same PD activity will vary as a function of the sensitivity of the PD detector and
thus the width and number of magnitude windows. Therefore a constant gain must be used if
various sets of pulse height analysis data are to be compared on the basis of NQN.
NQN, is to some extent, an attempt to represent the overall PD activity with a
single number. It should be noted that the more common Partial Discharge
Power Loss (P) or Apparent Power Loss (P
A
) representing actual energy
dissipation due to partial discharge serves this function.
As it is stated in the standard, NQN will vary significantly with a sensitivity of PD
detector and therefore NQN from different sensors or devices for the same PD
event can not be properly compared. For example, if you have a motor that
uses RFCTs on the surge capacitor grounding and wish to compare those results
to those on the same motor that uses 80 pF coupling capacitors, the data is not
comparable, even for the same PD event. If P
A
was used, this would not be the
case.
This is not the only problem with NQN. Even using the same PD detector and the
same instrument gain but varying number and size of magnitude windows and
maintaining the same magnitude range (FS), NQN will vary with magnitude
window.
The following example compares the relative variations in NQN and P
A
with a
change of magnitude and the number of magnitude windows for the same PD
event. Figure 1 shows the phase-resolved pulse height PD distribution and pulse
height distributions for positive negative and overall PD activity. Figure 2 shows
NQN and P
A
(positive, negative and overall) variation with magnitude window
size maintaining full scale (FS) at 1600mV. Magnitude window size was varied
from 10 to 200 mV and number of windows respectively from 160 to 8.
N(Q) Distribution
1
10
100
1000
10000
0 200 400 600 800 1000 1200 1400 1600
Q [mV]
N
[
p
p
s
]
N+ N- N
a. b. (Magnitude window at 100 mv)
Figure 1
NQN and PDI v.s. Magnitude Window Size
0
0.2
0.4
0.6
0.8
1
1.2
0 50 100 150 200
Magnitude Window [mV]
N
o
r
n
a
l
i
z
e
d
N
Q
N
a
n
d
P
D
I
NQN+ NQN- NQN PDI+ PDI- PDI
Figure 2
As can be seen from the Figure 2, an apparent power (PDI) maximum deviation
is for about 5% and PDI=(PDI+) + (PDI-). In the case of NQN, the variation is
about 3 times and NQN is not equal to (NQN+) + (NQN-).
The reason of all complications and problems with NQN is a use of a nonlinear
logarithmic operator applied to partial discharge pulse repetition rate. Therefore,
any linear operator can not be applied to NQN. A short list of issues are:
NQN data obtained with different PD sensors or instruments can not be
compared. Even a comparison with the same sensor and the same
instrument on different test objects may be incorrect due to possible different
sensor sensitivity on different objects. Sensitivity of a coupling capacitor
connected to a machine line terminals, for example, can very up to 2-3 times
depending upon a presence of a surge capacitor or a surge arrester, number
of cables, and the surge impedance of the cables and actual wiring inside a
terminal enclosure.
NQN does not allow flexibility in power frequency phase data analysis. This
means that NQN calculated over 180
0
phase range can not be compared in
any way to NQN calculated over 360
0
or 90
0
.
NQN does not allow for comparison of data taken at different full scales. In
other words, if pulse magnitude increased over time and one needs to use a
larger scale on PD analyzer, all previous data obtained at the lower scale is
not comparable to the current data. One should start the trend from the
beginning. (In our example above, the change of FS from 1600mV to 3200mV
maintaining 16 magnitude windows increases NQN by about 15% for the
same PD event.)
There is a difficulty in understanding the increase of PD with NQN. Lets
assume that PD pulse count increased twice in each magnitude window over
some time. NQN will reflect the increase as an addition to the existing NQN
that depends upon number of magnitude windows and their size. Lets
assume that maximum magnitude did not change and, therefore, FS factor,
window size, number of windows and gain are the same. In our example with
FS=1600mV and 16 magnitude windows the increase will be about 500.
Therefore, if initial NQN was 10 - the new one will be 510 (a 51 times
increase), but if initial NQN was 1,000 the new one will be 1,500 (50%
increase).
In conclusion, we recommend avoid NQN for all new PD technology users. Old
users should attempt switching to P and/or P
A
at all new measurements and also
to attempt recalculate older data to the same PD integral quantities for
comparison and trending purposes.