Professional Documents
Culture Documents
Tesis Factor K PDF
Tesis Factor K PDF
By
Jian Zheng
A Thesis
2000
ABSTRACT
ongoing concern of electric utilities and consumers. UL Standards 1561 and 1562 suggest
This work focuses at investigating the concept of K-factor and the relationship be-
tween K-factor, transformer derating, and the transformer winding eddy-current loss.
test procedures for measuring the AC winding resistance of two type of distribution trans-
formers are developed and explained. Test procedures for checking the linearity and su-
From the test results, it is found that linearity and superposition holds very well
for the test transformers while the K-factor overestimates the losses in transformer wind-
ings. The difference between K-factor results and lab test results is explained. Another
approach for estimating the total stray loss in transformer winding, the Harmonic Loss
i
ACKNOWLEDGEMENTS
nard Bohmann, for his insights and direction through this research. He has
been an excellent advisor, patient, and helpful all through the time.
I take this opportunity to thank Dr. Bruce Mork, for his help and sug-
Konrad Heuvers for their time spent on reviewing this work. Their insights
Besides the professors I have listed, I would also like to thank all of
Scott Ackerman, John Miller, and Chuck Sannes, for being so helpful.
Finally, I wish to thank my family and friends for all the support they
have provided. Your support made my stay at Michigan Tech one that I will
ii
TABLE OF CONTENTS
ABSTRACTi
ACKNOWLEDGEMENTSii
TABLE OF CONTENTS...iii
CHAPTER 1 INTRODUCTION................................................................................................................. 1
iii
APPENDIX D MATLAB PROGRAM FOR ANALYSIS OF 2 KVA TRANSFORMER SHORT
CIRCUIT TEST RESULTS ...................................................................................................................... 55
iv
LIST OF FIGURES AND TABLES
FIGURE 2-1 CORE AND SHELL FORMS WITH WINDINGS ................................................................................... 4
FIGURE 2-2 SIMPLIFIED SINGLE-PHASE TRANSFORMER MODEL ..................................................................... 5
FIGURE 2-3 FOUR WINDING CORE-SECTION WITH MAIN LEAKAGE PATHS SHOWN ........................................ 5
FIGURE 2-4 10KVA, AMORPHOUS STEEL CORE SINGLE-PHASE DISTRIBUTION TRANSFORMER ........................ 6
FIGURE 2-5 WINDING EDDY-CURRENT INDUCED BY MAGNETIC FLUX IN THE WINDING CONDUCTORS ............. 8
FIGURE 3-1. LABORATORY SETUP FOR SHORT-CIRCUIT TESTS ON 2 KVA DISTRIBUTION XFMR................ 17
FIGURE 3-2. LAB SETUP FOR SHORT-CIRCUIT TESTS OF THE 10 KVA DISTRIBUTION XFMR....................... 18
FIGURE 3-3. LINE SPECTRUM......................................................................................................................... 21
TABLE 3-1TIME STEP VALUES AND CORRESPONDING DFT FREQUENCY SPACINGS FOR DIFFERENT NUMBERS
OF POINTS TRANSFORMED. ................................................................................................................... 20
TABLE C-1 2 KVA DISTRIBUTION TRANSFORMER HARMONIC GROUP TEST RESULTS 1.............................. 54
TABLE C-2 KVA DISTRIBUTION TRANSFORMER HARMONIC GROUP TEST RESULTS 2................................. 54
TABLE C-3 DFT ACCURACY CHECK (10 KVA TRANSFORMER)................................................................... 54
v
________________________________________________
CHAPTER 1
INTRODUCTION
________________________________________________
With the ever-increasing use of solid state electronics in electrical load devices,
such as switching power supplies, variable-speed drives and many types of office
equipment [6], the power system network is being subjected to higher levels of harmonic
currents. One result of this trend is excessive internal heating in power distribution
heating problems. Design changes include enlarging the primary winding to withstand the
inherent triplen harmonic circulating currents, doubling the secondary neutral conductor
to carry the triplen1 harmonic currents, designing the magnetic core with a lower normal
flux density by using higher grades of iron, and using smaller, insulated secondary
conductors wired in parallel and transposed to reduce the heating from the skin effect and
associated AC resistance.
Several methods of estimating the harmonic load content are available. Crest-
Factor and Percent Total Harmonic Distortion (%THD) are the two common methods.
1
Triplen harmonics are created by non-linear loads. They flow in the neutral conductor and windings of the
power transformer. They are odd harmonics devisable by three, including the 3rd, 9th, 15th, and 21st.
1
The third method K-Factor can be used to estimate the additional heat created by non-
sinusoidal loads
The crest factor is a measure of the peak value of the waveform compared to the
(I
h=2
h )2
%THD =
I1 (1.2)
current.
Both of the above methods are limited because frequency characteristics of the
The third method, K-factor, is defined as the sum of the squares of the per unit
K = ( I h ( pu ) ) 2 h 2
h =1 (1.3)
where Ih(pu) is the harmonic current expressed in per unit based upon the
2
K-factor was introduced in UL standards 1561 [14] and 1562 [15] for rating
transformers based on their capability to handle load currents with significant harmonic
content.
harmonic load current magnitudes expected. Several manufacturers have utilized this
standard to market transformers that are specifically designed to carry the additional
harmonic currents.
This thesis is aimed at investigating the concept of K-factor and the relationship
between K-factor, derating, and the winding eddy-current loss of harmonic currents.
Chapter 2 presents the structure of the transformer under study, the K-factor theory and
existing work. Chapter 3 documents the test procedure and data processing methods
developed for determining the winding eddy-current loss, AC winding resistance and K-
factor. Chapter 4 compares the results of the measurements with the ideal results of the
K-factor theory and explains the difference. Chapter 5 provides conclusions and
3
________________________________________________
CHAPTER 2
There are two basic core designs for single-phase transformer: core form and shell
form.
/2
a) Core b) Shell
Figure 2-1 Core and Shell forms with Windings
Due to insulation requirements, the low voltage (LV) winding normally appears
closest to the core, while the high voltage (HV) winding appears outside. The windings
are usually referred to as primary and secondary winding(s) as denoted by the P and S. In
the shell form, the flux generated in the core by the windings splits equally in both "legs"
of the core. Winding configurations may vary with core design and include concentric
4
windings, pancake windings and assemblies on separate legs. A commonly used
P2 Lp Rp Ls Rs
S2
Lc
Rc
P1 S1
transformer. It includes the winding resistance and leakage as well as the core losses so it
is widely used for all core and winding configurations of the single-phase two-winding
variety.
/2 /2
depending on a series or parallel connection of the windings. Amperage rating for 120V
5
winding connection is 8.33A while it is 4.17A for the 240 V winding connection. In this
particular design, the high voltage windings are nearest to the core while the low voltage
core. This transformer consists of two low-voltage and two high-voltage windings
Tank
Oil Amorphous
Low Voltage winding
High Voltage
High Voltage
High Voltage
Core type
winding is placed closest to the core, with the high-voltage (primary) winding is outside.
The two high-voltage windings of the transformer are permanently connected in series. A
center tap in secondary winding can be used to provide different output voltages.
The core of the test transformer is made of wound amorphous steel ribbons and
has a core-type structure. Amorphous steel is made by rapidly cooling the metal at a rate
of 106 K/s. Thinner gauge steel, lower electrical conductivity, and a disorderly crystalline
6
structure are characteristics that separate amorphous from silicon steel. Compared to a
typical silicon steel core, an amorphous core offers an impressive reduction in average
core losses of up to 60-70% [19]. The reduction in average power losses due to hysteresis
can be attributed to the disorderly crystalline structure. The reduction in eddy current
loss (excitation loss); load loss (impedance loss); and total loss (the sum of no-load loss
and load loss). Load loss is subdivided into I2R loss and stray loss. [1].
where Ptotal is the total loss, Pno-load is the no-load loss, Pload is the load loss and the
Stray Loss is the loss caused by stray electromagnetic flux in the windings,
core, core clamps, magnetic shields, enclosure or tank walls, etc. Thus, the stray loss can
be subdivided into winding stray loss and stray loss in components other than the
windings (POSL).
The winding conductor strand eddy-current loss is caused by the time variation of
the leakage flux through the winding conductors [21], as shown in Figure 2-5. The other
7
stray loss is caused by the same mechanism within the tank wall, core clamps, etc.
Magnetic flux in
Magnetic flux
the winding
in the core
Core
Figure 2-5 Winding eddy-current induced by magnetic flux in the winding conductors
where PEC is the winding eddy-current loss and POSL is the other stray loss.
According to [1], all of the stray loss is assumed to be winding eddy current loss
and winding eddy-current loss for sinusoidal currents is approximately proportional to the
square of the frequency. The total load loss (copper loss) can be stated as
By measuring the copper loss and the rms current, RAC can be measured.
8
2.4 K-FACTOR
UL standards 1561 [14] and 1562 [15] introduced a term called the K-factor for
rating transformers based on their capability to handle load currents with significant
The K-factor is an estimate of the ratio of: (a) the heating in a transformer due to
winding eddy currents when it is loaded with a given nonsinusoidal current to (b) the
winding eddy-current heating caused by a sinusoidal current at the rated line frequency
which has the same RMS value as the nonsinusoidal current. For example, if the current
in a transformer winding is 100 A, and this current has a K-factor of 10, then the eddy
current losses in that winding will be approximately 10 times what they would be for a
Although the K-factor formula was defined for transformer currents, K-factors of
individual load currents are sometimes computed. This practice can be misleading
because, in general, K-factors measured at transformers are significantly lower than the
devices. The reduction is primarily due to other sinusoidal load currents, power system
impedance and the essentially random phase angles of the harmonic currents produced by
various loads.
The AC loss in a transformer winding is mainly due to the sum of the I2R losses
produced by the fundamental and harmonic components of the current, recognizing that
for each component, R depends on the frequency of that component. For lower-order
9
harmonics, the frequency dependence of the winding resistance is primarily due to the
proximity effect, a phenomenon that occurs in coils because the magnetic field
surrounding each conductor in a coil depends on the fields produced by other conductors.
The proximity effect produces greater losses than those predicted by the skin effect,
The K-factor formula does not account for the core eddy current losses and other
losses that occur in transformer cores. Core losses due to harmonics depend primarily on
the voltage distortion across the transformer windings. The voltage distortion appearing
impedance of the transformer, the impedance of the system feeding the transformer, and
the voltage distortion of that system. Although K-rated transformers are usually
constructed to withstand more voltage distortion than other transformers, this capability
The K-factor formula is based on the assumption that the winding eddy current
the square of the harmonic order as well as being proportional to the square of the
K = ( I h ( pu ) ) 2 h 2
h =1 (2.6)
where Ih(pu) is the rms current at harmonic h (per unit of rated rms load
10
(3) K-factor rated transformers have not been evaluated for use with harmonic
loads where the rms current of any singular harmonic greater than the tenth
(a) Winding eddy-current loss (PEC) is proportional to the square of the load
(b) Superposition of eddy current losses will apply, which will permit the direct
According to [1], suppose the eddy current loss under rated conditions is
where PEC-R is the eddy current loss under rated conditions and
From the first assumption, the eddy-current loss due to harmonic component is
fh 2 2 I
PEC ( h ) = REC R ( ) I h = REC R h 2 I h2 = PEC R ( h ) 2 h 2 (2.9)
f1 IR
where PEC(h) is the eddy current loss due to harmonic current of order h, IR is the
rated load current, fh is the harmonic frequency at order h and f1 is the fundamental
frequency.
According to the second assumption, the eddy-current loss due to the total
h = hmax h = hmax
Ih
PEC = PEC ( h) = PEC R
h =1
(I
h =1
) 2 h 2 = PEC R K (2.10)
R
11
In subsequent chapters, it is found that the K-factor assumption is too restrictive.
So I suggest that the assumption that the winding eddy-current loss (PEC) is proportional
Ih 2
PEC ( h ) = REC R h I h2 = REC ( h ) I h2 = PEC R ( ) h (2.11)
IR
where
than 2
K = ( I h ( pu ) ) 2 h (2.13)
h =1
The Harmonic Loss factor, as defined by IEEE Std C57.110-1998 [17], is given below
h = hmax h = hmax
Ih
2
Ih h 2
[I
h =1
]2 h 2
FHL = h =1
h = hmax
= h = hmax
1
(2.14)
I
h =1
Ih
2
h =1
[ h ]2
I1
From (2.7) (2.9), the K-factor was derived based on the assumption that the
measured application currents are taken at rated currents of the transformer. This is
seldom encountered in the field. This is where the FHL comes in handy because it can be
12
calculated in terms of the actual rms values of the harmonic currents and the quantity Ih/I1
h = hmax 2
Ih
K factor = h =12 FHL (2.15)
IR
According to [17], a Harmonic Loss Factor for other stray losses is defined as
h = hmax h = hmax
Ih
2
I h h 0.8 [I
h =1
]2 h 0.8
FHL STR = h =1
h = hmax
= h = hmax
1
(2.16)
I
h =1
Ih
2
h =1
[ h ]2
I1
based on the assumption that the other stray losses are proportional to the square of the
Because the other stray losses can not be ignored, in [17], an assumption is made
a) 67% of the total stray loss at rated frequency is assumed to be winding eddy
losses for dry-type transformers and 33% of the total stray loss at rated
b) 33% of the total stay loss at rated frequency is assumed to be winding eddy
losses for oil-filled transformers and 67% of the total stray loss at rated
13
________________________________________________
CHAPTER 3
LABORATORY TESTS
________________________________________________
This chapter presents the test procedure and data processing methods developed
for determining the winding eddy-current loss, AC winding resistance and K-factor.
typically been based on meter measurements. Voltages and currents are measured with
RMS meters, and power is measured with an average reading wattmeter. Significant
measurement errors are possible for harmonic study. Only a true RMS meter can take
measurements which correctly include the effect of all harmonics within the meters
oscilloscope was used to record the waveforms of the voltage and current. A voltage
probe of ratio 1:100 was used. Hall effect current probe with a 1:1 ratio was used to
obtain current waveforms. The digital scope could save the sampled data on floppy
diskette. This allowed waveform data to be transported to a PC for analysis using the
14
3.2 TEST DEVICES
3.2.1 Power source
The power source used, AMX-3120, is a product of Pacific Power Source
test purposes, 3-phase voltage with programmable harmonic contents can be generated
from this device. It is configured with an interchangeable digital controller called the
allows control of voltage and frequency, but also allows the user to simulate virtually any
transient (including sub-cycle waveform disturbance). Main features of the power source
are: [11]
External Sense input This is required for precise control of the output
voltage of the power source. The line drops are taken care of by using external
sense inputs.
15
3.2.2 UPC-32
UPC 32 is a programmable controller designed to directly plug into Pacific
single, two or three phase signal generator and can be remotely controlled from a PC
either through a GPIB interface or through a serial interface. Main features of the UPC-
32 are: [12]
Operations in 4 modes:
Calculation time: 45 sec + 10 sec for each non-zero magnitude of the harmonic
99 user programs that contain steady state and transient parameters can be stored
the metering point based on the metered voltage at that point. Therefore, accurate
Control Local/Remote. In remote control mode, the source can be either controlled
16
3.2.3 Oscilloscope
The oscilloscope used is a Nicolet Pro20, a digital oscilloscope from Nicolet
The Nicolet Pro20 can be configured with a wide variety of input channels and
The laboratory setup used to perform the short-circuit test of the 2 kVA dry-type
AMX 3120 AC
Power Source
Master
transformer
H4 X1
Slave
UPC-32 H3 X2
H2 X3
Current Amplifier
H1 X4
Nicolet Pro 20
Oscilloscope
Figure 3-1. Laboratory Setup for Short-circuit Tests on 2 kVA Distribution XFMR
17
This transformer is a 2 kVA single phase, dry type, 4winding 120/240 Volt
general purpose transformer. It is excited at the high-voltage winding (H4-H3) with low-
The laboratory setup used to perform the short-circuit test of the 10 kVA
AMX 3120 AC
Power Source
transformer
H1 X3
X2
H2
X1
Current Amplifier
Nicolet Pro 20
Oscilloscope
Figure 3-2. Lab Setup for Short-circuit Tests of the 10 kVA Distribution XFMR
This allowed waveform data to be transported to a PC for analysis using the VuPoint
18
The average power is calculated from v(t) and i(t):
T
1
P = v(t )i (t )dt
T 0
which can be acquired using the statistic function Mean in the Vupoint program.
S = VRMS*IRMS
Q = S 2 P2
P Q
Rsc = 2
X sc = 2
I RMS I RMS
The laboratory setup used to perform the harmonic test is the same as short-circuit
tests above. The only difference is that the voltage applied to the transformer in this test
The harmonic test requires FFTs (Fast Fourier Transform) of the current
19
obtain a discrete spectrum or line spectrum for periodic waveforms, the waveform data
The waveform data must cover the range of an integral number of cycles.
Table 3-1Time step values and corresponding DFT frequency spacings for different numbers of
points transformed. [13]
The relationship between the time step, numbers of points and DFT frequency
spacing is:
1
f =
t NPTS
where
f is the DFT frequency spacing
20
T = 1, 2, 5 S;
10, 20, 50 S;
For the test purpose, a sweep length of 8192 points and time setting of 200 S
were chosen. The total length of the waveform is 1.6384 second. Then the data was cut
cosine-tapered rectangular, Bartlett, Hanning and Parzen. If the data being transformed
21
was an integer number of waveform cycles, a rectangular window with no tapering was
sufficient.
For the processing of harmonic test data, the last row in Table 3.1 was used. In
VuPoint, the data was first interpolated to a sampling time of 244.14 S, then cut off to
only 1 second long. The FFT result is a perfect discrete spectrum (line spectrum). (Figure
3.3)
measurements. It removes any residual magnetism from the attached current probe and it
initiates an operation to remove any undesired DC offsets from probe circuitry. This
operation is recommend each time a new measurement is started or any setting on the
probe is changed.
The short-circuit tests for the 2 kVA transformer were done both manually and
automatically. The manual test was done continuously. At each frequency, the test was
repeated three times and the average value was recorded. The automatic test procedure is
The short-circuit tests for the 10 kVA transformer was done only manually
because of the limitation of the voltage output of the Power source. At each frequency
point, only one measurement was made. But the whole test sequence was repeated three
times in different order. The first and the third test were done from high frequency to low
frequency while the second was done from low frequency to high frequency. The average
22
________________________________________________
CHAPTER 4
chapter 3. Detailed analysis of the test results is provided. The test results can be found in
4.1 LINEARITY
Because the voltage limitation of the power source, when doing the short-circuit
test, the rated current of the test transformer may not be reached at high frequencies.
Because only resistance is our concern, an alternative way is to do the short-circuit test at
lower voltage level if the linearity of the resistance holds. A check on the linearity of
Table 4.1
Because the error of R is between 2.6%, the linearity of the resistance holds
very well.
23
4.2 SUPERPOSITION
assumptions. One of them is that superposition of eddy current losses will apply, which
will permit the direct addition of eddy losses due to the various harmonic. This
First, a group of harmonics is applied to the transformer together. The voltage and
current waveforms are recorded. The load loss is measured as Pgroup. An FFT is then used
on the voltage waveform to get the amplitude and the frequency of the individual
harmonics in the group. Then individual harmonic in the group is applied to the
transformer one by one at the same amplitude and frequency, the load losses are recorded
as Pindividual. If the sum of the Pindividual is equal to Pgroup, the superposition assumption is
correct.
24
The error is (45.568 46.02)/46.02*100% = 0.98% which is small enough to
400
350
300
c
a 250
R
200
150
100
Because when short-circuit tests are made continuously, heat may accumulate in the
transformer and the temperature in the transformer winding conductor may rise which
will cause the increase of the resistance. To test how much the effect will be, two set of
short-circuit tests were performed on the same 10 kVA transformer. The first set did the
test from high frequency (2940 Hz) to low frequency (60 Hz) while the second test set
was done from low frequency (60 Hz) to high frequency (2940 Hz). The results are
25
plotted in Figure 4.1. There is a small difference between the two sets of test results. The
The test was repeated for three times. Please see Appendix A for original data. In
Table 4.3, the quantities are the average values of these measurements.
26
The resistance and the reactance are plotted in Figure 4-2.
The first and the third test were done from high frequency to low frequency while
the second test was from low frequency to high frequency. Although temperature effect
has been considered to be small enough, this kind of test scheme can reduce possible
3000
R
X
2500
2000
s
m 1500
h
O
1000
500
0
0 500 1000 1500 2000 2500 3000
Frequency(Hz)
Figure 4-2 Short-circuit Test Results: R.X vs. Frequency (10kVA Transformer)
4.4.1.1 2nd order fit for the AC Winding Resistance RAC (10 kVA, all points)
According to (2.7), a least square fit of a second order polynomial is used for all
the data points in Table 4.2. The original RAC curve and the fit curve are shown in Figure
4.3.
27
From Figure 4.3, it is very clear that the second order polynomial is not a good
choice for fitting the test data. Using this fit, the RAC would be:
400
350
300
c 250
a
R
200
150
100
50
0
0 500 1000 1500 2000 2500 3000
f (Hz)
Figure 4-3 2nd fit for Rac (fh/f1)2 from 60 Hz to 2940 Hz (25 points)
4.4.1.2 One Section Optimal fit for the AC Winding Resistance RAC ( 10 kVA)
From Figure 4.3, it is clear that the second order polynomial for the whole data set
is not a good match. Next step would be to use an optimal fit of a constant plus a 2nd term
It turned out that the optimal exponent found is 1.03 and the total error is 20.32.
The original RAC curve and the fit curve are shown in Figure 4-4
28
RAC = 52.69 + 7.464(fh/f1)1.034 (4.3)
Optimal fit for 10 kVA Rac (fh/f1) expo from 60 Hz to 2940 Hz ( 25 points)
350
300
250
c
a
R 200
150
100
50
0
0 500 1000 1500 2000 2500 3000
f (Hz)
Figure 4-4 Optimal fit for 10 kVA Rac from 60 - 2940 Hz ( all the 25 points)
transformer winding because of the assumption that the eddy current losses are
proportional to the square of the frequency for all frequencies. In fact, for high enough
to the square root of the frequency instead of the square of the frequency.
One improvement suggested from this explanation is to use a two-section fit for
the RAC test data. It is necessary to find where the transition between the 2nd order
29
It would have the form:
So the transition point is moved from the 3rd point to the 22nd point in the data to
find a best position (minimum fit error). The first part is the 2nd order polynomial fit
Total Error (transition point moves from the 3rd - 22nd(300 Hz- 2580 Hz) 10 kVA
120
100
80
r
or
r
E 60
l
at
o
T
40
20
0
0 500 1000 1500 2000 2500
f (Hz)
Figure 4-5 Total fit error while transition point moves. (Square/non-square)
From Figure 4.5, it can be seen there is no optimal transition point found when the
Naturally, further improvement is to use optimal fit for both sections. It would
30
RAC = RDC + Rco1(fh/f1)x1 (fh transition frequency ) (4.5)
In Figure 4.6, an optimal transition point is found on 1260 Hz (the 11th point in
Total Error (transition point moves from the 3rd - 22nd(300 Hz- 2580 Hz) 10 kVA
20
18
16
14
12
r
or
r
E 10
l
at
o
T 8
0
0 500 1000 1500 2000 2500
f (Hz)
Figure 4-6 Total fit error while transition point moves (both sections are optimal fit)
4.4.1.4 Summary of the fitting tests for the RAC (10 kVA)
The results from these different fitting methods for the 10 kVA transformer AC
31
Fitting Method Exponent Error
Section 1 Section 2
One section (total 25 points) 2 150.5
One Section (total 25points) Optimal found = 1.034 20.32
Two sections(first fixed at 2) Best transition points not found N/A
Two sections (both optimal) 1.304 1.118 13.0
Table 4-4 Fitting methods comparison for 10 kVA Transformer data
4.4.2 2 kVA distribution transformer
The short-circuit tests for the 2 kVA distribution transformer were done both
automatically and manually. The AC winding resistance derived from the automated
Because the frequency step used in the automated tests is 30 Hz, only part of the
data set will be used for the following fitting experiment. (Table 4.5)
32
2 KVA XFMR AC Winding Resistance (automatic test results)
1.8
1.6
1.4
1.2
1
s
m
h 0.8
O
0.6
0.4
0.2 R
0
0 500 1000 1500 2000 2500 3000
Frequency(Hz)
4.4.2.1 2nd Order Polynomial Fit for the AC Winding Resistance RAC (2 kVA
distribution XFMR)
First, a 2nd order polynomial fit is used for all the data points in Table 4.5.
The original RAC curve and the fit curve are shown in Figure 4.8.
33
One section 2nd order fit for 2 kVA XFMR (60 - 1680 Hz) ( 27 points)
1.2
Test data
Fit data
0.8
c 0.6
a
R
0.4
0.2
0
0 200 400 600 800 1000 1200 1400 1600
f (Hz)
Figure 4-8 One Section fit for 2kVA XFMR AC Winding Resistance data
4.4.2.2 Optimal fit for the AC Winding Resistance RAC (2 kVA distribution XFMR)
From Figure 4.8, it can be seen that the 2nd order polynomial fit for the whole data
set is not a good match. Next step would be to use optimal fit for the whole length of
data.
It turned out that the optimal exponent found is 1.7087 and the total error is
0.0278. The original RAC curve and the fit curve are shown in Figure 4.9.
34
One section optimal fit for 2 kVA XFMR Rac (60 - 1680) Hz ( 27 points)
1.2
Test data
Fit data
0.8
c 0.6
a
R
0.4
0.2
0
0 200 400 600 800 1000 1200 1400 1600
f (Hz)
Figure 4-9 One Section Optimal fit for 2 kVA XFMR RAC (60-1680 Hz)
According to [2], a two-section fit is used for the whole data set of RAC. The first
section will use a 2nd order polynomial fit while the second part uses a non-2nd order
optimal fit. The transition point is moved from the 3rd point to the 27th point to find the
best fit.
35
Total Error when the transition point moves from 3 - 27
0.06
0.05
0.04
r
or
r 0.03
E
l
at
o
T
0.02
0.01
0
5 10 15 20 25
Harmonic order
Figure 4-10 The total fitting error while the transition points between 2nd order fit and optimal fit
moves
The total error while the transition points moves is plotted in Figure 4.10. It can
be observed that the minimum error is found when the transition point is at 1080 Hz and
The same process was repeated for a two-section fit which both sections use an
optimal fit.
36
The total error while the transition points moves is plotted in Figure 4.11. It can
be observed that the minimum error is found when the transition point is at 1560 Hz and
0.05
0.04
r
or
r 0.03
E
l
at
o
T
0.02
0.01
0
5 10 15 20 25
Harmonic order
Figure 4-11 The total fitting error while the transition points between two optimal fit regimes moves
The results from these different fitting methods for the 2 kVA transformer AC
37
Fitting Method Exponent Error
Section 1 Section 2
One section (total 27 points) 2 0.0536
One Section (total 27points) Optimal exponent found = 1.709 0.0278
Two sections(first fixed at 2) 2 1.189 0.0342
Two sections (both optimal) 1.706 1.531 0.0271
Table 4-6 Fitting methods comparison for 2 kVA Transformer data
winding eddy currents when it is loaded with a given nonsinusoidal current to the
winding eddy-current heating caused by a sinusoidal current at the rated line frequency
which has the same RMS value as the nonsinusoidal current. [2]
Transformers with K-factor ratings are constructed so that their winding eddy
current losses are very low for sinusoidal currents at the rated line frequency. This allows
them to have acceptable losses when they are fully loaded with non-sinusoidal currents
that have a K-factor less than or equal to the K-rating of the transformer.
As stated in chapter 2, the K-factor formula is based on the assumption that the
current is proportional to the square of the harmonic order as well as being proportional
to the square of the magnitude of the harmonic component. However, this assumption is
not always true which can be seen from the test results presented in this chapter.
For example, the best fit curve for the 10 kVA distribution transformer tested is
38
So at lower frequencies the exponent of (fh/f) is about = 1.3037 and at higher
the K-factor. The power of the harmonic order should not be limited to 2. The K
K = ( I h ( pu ) ) 2 h (2.13)
h =1
Part of the reason that the exponent is less than 2 is that in (2.2)
The other stray loss (POSL) was ignored when defining K-factor. So the actual
Because POSL are proportional to the square of the load current while not
proportional to the square of the harmonic frequency, the total PEC-A is not proportional to
Another weak point of the K-factor formula is that it overestimates the high-
frequency losses in transformer windings. According to formulas in [16], for high enough
frequencies, winding eddy current losses in transformers are not proportional to the
square of the frequency. The geometry of the windings in a given transformer determines
when the transition between the square and the non-square regimes occurs.
made is separating other stray loss (POSL) from winding stray loss (PEC)
39
Because the other stray losses can not be ignored, in [17], an assumption is made
a) 67% of the total stray loss is assumed to be winding eddy losses for dry-type
transformers and 33% of the total stray loss is assumed to be the other stray
loss.
b) 33% of the total stay loss is assumed to be winding eddy losses for oil-filled
transformers and 67% of the total stray loss is assumed to be the other stray
loss.
This assumption can be checked using an optimal search. Using the assumption
that the winding eddy-current loss vary with the square of the frequency and the other
stray loss vary with the frequency raised to the 0.8 power, the fit formula is:
It is found these assumptions are not accurate for the tested transformer but it can
help explain the difference of the exponent of (fh/f) between 2 kVA dry-type transformer
proportional to the square of the frequency, takes a larger part in the total stray loss than
40
________________________________________________
CHAPTER 5
_______________________________________________________
This chapter presents the conclusions drawn from this work. Topics including
closing comments regarding the lab tests, the K-factor concept and the Harmonic Loss
5.1 CONCLUSIONS
The K-factor does not apply to the two tested transformer and overestimates the
transformers tested are not proportional to the square of the frequency, instead, they
For the two transformers tested, the eddy-current loss is a function of frequency with
The Harmonic Loss Factor is a better approach for estimating transformer load loss.
Compared with the K-factor, the Harmonic Loss Factor is a function of the harmonic
current distribution and is independent of the relative magnitude while the K-factor is
dependent on both the magnitude and distribution of the harmonics. Harmonic Loss
Factor also has a separate definition for the other stray losses assuming that they are
41
proportional to the square of the load current magnitude and the harmonic frequency
transformer structure such as the geometry of the windings is necessary for further study.
A laboratory test method for separating winding eddy current losses from stray
losses in components other than windings are important in the future work.
42
Reference:
[1] "An American National Standard: IEEE Recommended Practice for Establishing
ANSI/IEEE C57.110-1986
[3] Tom Shaughnessy, "Use Derating and K-Factor Calculation Carefully", Power
for publication.
[5] D.Yildirim and E.F.Fuchs, Measured Transformer Derating and Comparison with
43
[9] Keith H. Sueker, Comments on Harmonics: The Effects on Power Quality and
[10] Gregory W. Massey, Estimation Methods for Power System Harmonic Effects on
[11] AMX Series AC Power Source Operation Manual, Pacific Power Source, Oct,
1996.
[13] Bruce Andrew Mork, Ferroresonance and Chaos: Observation and Simulation of
[14] Standard UL1561, Dry-Type General Purpose and Power Transformers, April 22,
1994.
Institute of Electrical and Electronics Engineers, Inc., New York, NY, 1998.
[18] Manjunatha Rao, Development of a Laboratory Test Setup Using LabView for a
44
[19] Michael J. Gaffney, Amorphous Core Transformer Model for Transient
[21] Richard L. Bean, Transformers for the Electric Power Industry, McGraw-Hill
45
APPENDIX A: 10 KVA DISTRIBUTION XFMR SHORT CIRCUIT TEST RESULTS
46
Table A-2 10 KVA distribution Transformer Test No.2
Order Freq Iraw Irms(A) Vraw V rms (v) Mean P(W) S(VA) Q (VAR) R(Ohm) X(Ohm)
(Hz) rms(mv) rms (v) (Vpp*Ipp)
1 60 0.9695 0.048475 0.04574 4.574 31.9948 0.159974 0.22172465 0.153526 68.07909 65.33493
3 180 0.5445 0.027225 0.06074 6.074 11.89376 0.059469 0.16536465 0.154301 80.23307 208.1777
5 300 0.901 0.04505 0.1608 16.08 35.8912 0.179456 0.724404 0.701824 88.42364 345.8108
7 420 1.0255 0.051275 0.25075 25.075 52.6592 0.263296 1.28572063 1.258472 100.1458 478.6657
9 540 1.0335 0.051675 0.3213 32.13 64.0256 0.320128 1.66031775 1.629163 119.8844 610.1037
11 660 0.939 0.04695 0.35195 35.195 60.2624 0.301312 1.65240525 1.624701 136.6927 737.0591
13 780 0.9215 0.046075 0.402 40.2 65.0752 0.325376 1.852215 1.823412 153.2692 858.9227
15 900 0.9075 0.045375 0.4529 45.29 72.832 0.36416 2.05503375 2.022511 176.872 982.3305
17 1020 0.9005 0.045025 0.50315 50.315 78.7456 0.393728 2.26543288 2.230956 194.2177 1100.484
19 1140 0.89 0.0445 0.55415 55.415 85.4016 0.427008 2.4659675 2.428716 215.6334 1226.469
21 1260 0.933 0.04665 0.6351 63.51 100.5568 0.502784 2.9627415 2.919768 231.0352 1341.668
23 1380 0.955 0.04775 0.7058 70.58 113.9712 0.569856 3.370195 3.321668 249.93 1456.832
25 1500 0.9845 0.049225 0.7869 78.69 128.6656 0.643328 3.87351525 3.819719 265.4978 1576.376
27 1620 0.9985 0.049925 0.8575 85.75 141.2096 0.706048 4.28106875 4.222445 283.2684 1694.057
29 1740 1.0095 0.050475 0.9287 92.87 151.9104 0.759552 4.68761325 4.625667 298.1294 1815.607
31 1860 1.0315 0.051575 1.0092 100.92 167.1168 0.835584 5.204949 5.13744 314.1316 1931.383
33 1980 1.067 0.05335 1.1101 111.01 188.6208 0.943104 5.9223835 5.846809 331.3529 2054.235
35 2100 1.0795 0.053975 1.1908 119.08 203.6736 1.018368 6.427343 6.346154 349.5582 2178.338
37 2220 1.0915 0.054575 1.2706 127.06 217.4976 1.087488 6.9342995 6.848495 365.1211 2299.364
39 2340 1.0695 0.053475 1.3118 131.18 217.3952 1.086976 7.0148505 6.930123 380.1179 2423.48
41 2460 1.054 0.0527 1.3616 136.16 222.6176 1.113088 7.175632 7.088775 400.7821 2552.407
43 2580 1.0415 0.052075 1.4112 141.12 225.792 1.12896 7.348824 7.261588 416.313 2677.769
45 2700 1.016 0.0508 1.441 144.1 223.9488 1.119744 7.32028 7.234132 433.9017 2803.232
47 2820 0.955 0.04775 1.4096 140.96 206.4384 1.032192 6.73084 6.651224 452.7034 2917.124
49 2940 0.9665 0.048325 1.5 150 221.4912 1.107456 7.24875 7.163653 474.2232 3067.544
47
Table A-3 10 KVA distribution Transformer Test No.3
Order Freq Iraw Irms(A) Vraw V rms (v) Mean P(W) S(VA) Q (VAR) R(Ohm) X(Ohm)
(Hz) rms(mv) rms (v) (Vpp*Ipp)
1 60 3.1305 0.156525 0.1505 15.05 346.5216 1.732608 2.355701 1.596057 70.71848 65.14498
3 180 2.264 0.1132 0.25075 25.075 195.2512 0.976256 2.83849 2.665324 76.18524 207.997
5 300 1.6955 0.084775 0.3002 30.02 125.8752 0.629376 2.544946 2.465894 87.57388 343.1143
7 420 1.4425 0.072125 0.3514 35.14 107.0848 0.535424 2.534473 2.477271 102.9263 476.2137
9 540 1.297 0.06485 0.402 40.2 98.1248 0.490624 2.60697 2.560387 116.6618 608.8154
11 660 1.2125 0.060625 0.4526 45.26 100.864 0.50432 2.743888 2.697143 137.2153 733.8384
13 780 1.151 0.05755 0.5034 50.34 102.144 0.51072 2.897067 2.851695 154.2028 861.0183
15 900 1.114 0.0557 0.5536 55.36 109.696 0.54848 3.083552 3.03438 176.787 978.0467
17 1020 1.083 0.05415 0.6058 60.58 111.8208 0.559104 3.280407 3.23241 190.6758 1102.375
19 1140 1.062 0.0531 0.6554 65.54 117.0944 0.585472 3.480174 3.430573 207.6429 1216.684
21 1260 1.0415 0.052075 0.7072 70.72 124.2624 0.621312 3.682744 3.629955 229.1138 1338.575
23 1380 1.0285 0.051425 0.7546 75.46 134.9632 0.674816 3.880531 3.821406 255.1742 1445.022
25 1500 1.015 0.05075 0.8074 80.74 134.7584 0.673792 4.097555 4.041777 261.6096 1569.279
27 1620 1 0.05 0.8576 85.76 147.0464 0.735232 4.288 4.224497 294.0928 1689.799
29 1740 0.9945 0.049725 0.9094 90.94 142.4896 0.712448 4.521992 4.465515 288.14 1806.018
31 1860 0.9805 0.049025 0.9594 95.94 150.784 0.75392 4.703459 4.642642 313.6823 1931.657
33 1980 1.069 0.05345 1.1108 111.08 189.3376 0.946688 5.937226 5.861266 331.3687 2051.616
35 2100 1.058 0.0529 1.1614 116.14 191.5904 0.957952 6.143806 6.068664 342.3201 2168.611
37 2220 1.0455 0.052275 1.2118 121.18 197.0688 0.985344 6.334685 6.257581 360.5785 2289.91
39 2340 1.031 0.05155 1.2614 126.14 200.704 1.00352 6.502517 6.424615 377.6319 2417.63
41 2460 1.0225 0.051125 1.3112 131.12 206.1312 1.030656 6.70351 6.623805 394.3184 2534.2
43 2580 1.015 0.05075 1.3684 136.84 213.4016 1.067008 6.94463 6.86217 414.2815 2664.338
45 2700 0.9985 0.049925 1.4106 141.06 216.7808 1.083904 7.042421 6.958508 434.8652 2791.772
47 2820 0.9875 0.049375 1.4606 146.06 220.0576 1.100288 7.211713 7.127283 451.3279 2923.545
49 2940 0.972 0.0486 1.4998 149.98 222.8224 1.114112 7.289028 7.20338 471.6896 3049.747
48
Table A-4 10 KVA distribution Transformer Rdc Test Results
49
APPENDIX B
50
44 2640 1217.6 100 468.16 0.5 1 22.8884
45 2700 1244 100 468.8 0.5 1 23.4004
46 2760 1257.6 100 464 0.5 1 23.38
47 2820 1289.6 100 466.88 0.5 1 24.0353
48 2880 1321.6 100 469.12 0.5 1 24.7685
49 2940 1348 100 469.12 0.5 1 25.3379
50 3000 1377.6 100 470.72 0.5 1 25.8908
51
1260 69.4 8.28 22.45 1.01
1290 70.5 8.28 22.94 1.03
1320 71.3 8.28 23.44 1.04
1350 72 8.29 23.92 1.05
1380 73.1 8.3 24.44 1.06
1410 73.9 8.3 24.94 1.07
1440 74.7 8.3 25.43 1.08
1470 75.2 8.25 25.82 1.11
1500 76.1 8.28 26.32 1.11
1530 76.8 8.28 26.81 1.12
1560 78.5 8.32 27.38 1.13
1590 79.6 8.32 27.88 1.15
1620 80.6 8.3 28.37 1.17
1650 80.8 8.28 28.76 1.18
1680 81.6 8.28 29.26 1.19
1710 83 8.29 29.73 1.21
1730 83.5 8.3 30.14 1.21
1760 84.4 8.3 30.61 1.22
1790 85.1 8.28 31.02 1.24
1820 86.7 8.32 31.61 1.25
1850 86.9 8.29 32 1.26
1880 88.2 8.29 32.47 1.28
1910 88.6 8.3 32.99 1.28
1940 89.5 8.3 33.44 1.3
1970 90.9 8.32 33.94 1.32
2000 91.2 8.28 34.35 1.33
2030 92.7 8.3 34.94 1.34
2060 94 8.3 35.41 1.36
2090 94.3 8.28 35.8 1.37
2120 95.5 8.29 36.29 1.39
2150 96.5 8.3 36.79 1.4
2180 97.3 8.3 37.27 1.41
2210 98.4 8.3 37.72 1.43
2240 99.6 8.3 38.22 1.44
2270 100.4 8.3 38.74 1.46
2310 101.6 8.29 39.26 1.48
2340 102.9 8.29 39.78 1.5
2370 104.3 8.32 40.39 1.51
2400 105.3 8.32 40.87 1.52
2430 106.3 8.29 41.23 1.55
2460 107.5 8.29 41.72 1.56
2490 108.2 8.32 42.11 1.56
2520 108.5 8.3 42.58 1.57
2550 109.9 8.29 43.06 1.6
2580 111.1 8.3 43.54 1.61
2610 112.5 8.3 44.06 1.63
2640 113.2 8.29 44.53 1.65
2670 113.7 8.29 44.44 1.65
2700 115 8.28 45.03 1.68
2730 116.3 8.3 45.53 1.69
2760 117.3 8.29 46.02 1.71
2790 118.5 8.28 46.39 1.73
52
2820 119.4 8.29 46.39 1.74
2850 120 8.29 46.39 1.75
2880 120.7 8.29 46.39 1.76
2910 121.8 8.29 46.39 1.77
2940 122.2 8.26 46.39 1.79
2970 122.8 8.28 46.39 1.79
3000 123.7 8.3 46.39 1.79
53
APPENDIX C
54
APPENDIX D
List D.1 Program for finding the best one-section fit curve
%=====================================================================
% find the best fit curve for 2KVA XFMR automatic short-circuit data %
% (R)
%
%
% Newobj.m is used to find the best exponent fit for R array without
% DC point
% Matlab fmin function used as object function
% take the exponent as input parameter, then do
% Linear Regression with (fh/f1)^expo up to the points
% specified by N1;
% the error was the return value so fmin
% can find the optimal exponent value.
% autofit22.m call the fmin (will use Newobj.m )
%
% Usage: change the N0 and N1 to decide how many point you want to be
% used in the fitting.
%======================================================================
clear all;
close all;
global r f N0 N1;
f = dat1(1:N, 1)';
r = dat1(1:N, 2)';
%---------------------------------------------------------
% Without DC value: R start from 60 Hz
%---------------------------------------------------------
%------------------------------
% Case 1: expo = 2
%------------------------------
disp('------Without DC value: R start from 60 Hz, expo = 2')
expo = 2;
for i=1:N0
h(i) =(f(i)/60)^expo;
end;
55
f11 = f(1:N0);
r11 = r(1:N0);
[p,s]= polyfit(h,r11,1)
err11 = getfield(s,'normr')
y0 = polyval(p,h);
plot(f11,r11,'r.:',f11,y0),grid; % r: real data; y0: fit data
xlabel('f (Hz)');
ylabel('Rac');
legend('Test data','Fit data',2)
title('One section Square fit for 2 kVA XFMR (60 - 1680 Hz) ( 27
points) ');
axis tight;
%----------------------------------------------
% Case 2: expo = optimal output of the Newobj
%----------------------------------------------
output2 = fmin('Newobj',0.5,2) % Newobj: without DC point
% all points N = 25 points
% N0 %how many points are included from 60Hz
disp('------Without DC value: R start from 60 Hz, expo = best')
expo = output2
for i=1:N0
h2(i) =(f(i)/60)^expo;
end;
f12 = f(1:N0);
r12 = r(1:N0);
[p2,s2]= polyfit(h2,r12,1)
err12 = getfield(s2,'normr')
y2 = polyval(p2,h2);
figure;
plot(f12,r11,'r.:',f12,y2),grid; % r: real data; y2: fit data
xlabel('f (Hz)');
ylabel('Rac');
legend('Test data','Fit data',2)
title('One section optimal fit for 2 kVA XFMR Rac (60 - 1680) Hz ( 27
points) ');
axis tight;
56
List D.2 Object function used in finding best one-section fit curve
%================================================================
% Object Function for finding the best one-section fit curve
%
%================================================================
global r f N0 N1;
r4 = r(1:N0);
[p4,s4] = polyfit(h4,r4,1);
err = getfield(s4,'normr');
List D.3 Program for finding the transition point of the two-section fit and the best
Curves.
%================================================================
% Find the best transition points and fit curves for 2KVA XFMR Short
% Circuit test data ( AC winding resistance R )
%
% R array has 25 point in total, without the DC point
%
% change the variable "expo1" can set if the first section fit
% is using 2nd order or a optimal value.
% the total fit error is in array err
%================================================================
clear all;
close all;
global r f N0 N;
f = dat1(1:N, 1)';
r = dat1(1:N, 2)';
k = 1;
for N0 = 3: N-1; % N0 is global, the number of points used
for fitting
57
%expo1(k) = 2;
for i=1:N0
h(i) =(f(i)/60)^expo1(k);
end;
f1 = f(1:N0);
r1 = r(1:N0);
[p,s]= polyfit(h,r1,1);
err1(k) = getfield(s,'normr'); % the error of firs part square fit
pp_a(k) = p(1); %keep the results
pp_b(k) = p(2);
for i=N0+1:N
h2(i-N0) = (f(i-N0)/60)^expo2(k);
end;
h22 = h2(1:N-N0);
r2 = r(N0+1:N);
[p1,s1] = polyfit(h22,r2,1);
err2(k) = getfield(s1,'normr');
pp1_a(k) = p1(1); %keep the results;
pp1_b(k) = p1(2);
x = f(3:N-1);
hx = x/60;
plot(hx,err,'r.:'),grid;
xlabel('Harmonic order');
ylabel('Total Error');
axis([3 27 0 0.06]);
title('Total Error when the transition point moves from 3 - 27 ');
figure;
plot(x,expo1,x,expo2),grid; % r: real data; y0: fit data
xlabel('f (Hz)');
ylabel('exponent');
legend('firt part','second part',2)
title('exponent found ');
axis tight;
58
List D.4 firstpart.m
%====================================================================
% Object function used for find the best fitting curve for the points
% group from 1 -> N0
%
%====================================================================
function [err] = firstpart(expo)
global r f N0 N;
r1 = r(1:N0);
[p1,s1] = polyfit(h,r1,1);
err = getfield(s1,'normr');
%======================================================================
% Object function used for find the best fitting curve for the points
% from N0 +1 -> N
%
%======================================================================
function [err] = secpart(expo)
global r f N0 N;
r1 = r(N0+1:N);
[p1,s1] = polyfit(h,r1,1);
err = getfield(s1,'normr');
59
APPENDIX E
clear all;
close all;
global r f N0 N1;
60
P = dat1(1:N,10)'; % Power (W)
r1 = dat1(1:N,13)';
%--------- Data 12_3 ------------------------------
dat3 = dat12_3;
r3 = dat3(1:N,11)'; % different column with data12_1
%--------- Data 12_4 ------------------------------
dat4 = dat12_4;
r4 = dat4(1:N,11)'; % different column with data12_1
%--------- Average value of these 3 data set -------------
r = (r1 + r3 + r4)/3;
%---------------------------------------------------------
% Without DC value: R start from 60 Hz
%---------------------------------------------------------
%------------------------------
% Case 1: expo = 2
%------------------------------
% all points N = 25 points
%how many points are included from 60Hz upward
disp('------Without DC value: R start from 60 Hz, expo = 2')
expo = 2;
for i=1:N0
h(i) =(f(i)/60)^expo;
end;
f11 = f(1:N0);
r11 = r(1:N0);
[p,s]= polyfit(h,r11,1)
err11 = getfield(s,'normr')
y0 = polyval(p,h);
plot(f11,r11,'r.:',f11,y0),grid; % r: real data; y0: fit data
xlabel('f (Hz)');
ylabel('Rac');
legend('Test data','Fit data',2)
title('Square fit for Rac (fh/f1)^2 from 60 Hz to 2940 Hz ( 25 points)
');
axis tight;
%-----------------------------
% Case 2: expo = output2
%-----------------------------
output2 = fmin('Newobj',0.5,2) % myobj: without DC point
% all points N = 25 points
% N0 %how many points are included from 60Hz
disp('------Without DC value: R start from 60 Hz, expo = best')
expo = output2
for i=1:N0
h2(i) =(f(i)/60)^expo;
end;
f12 = f(1:N0);
r12 = r(1:N0);
[p2,s2]= polyfit(h2,r12,1)
err12 = getfield(s2,'normr')
61
y2 = polyval(p2,h2);
figure;
plot(f12,r11,'r.:',f12,y2),grid; % r: real data; y2: fit data
xlabel('f (Hz)');
ylabel('Rac');
legend('Test data','Fit data',2)
title('Optimal fit for 10 kVA Rac (fh/f1) ^ expo from 60 Hz to 2940 Hz
( 25 points) ');
axis tight;
%--------------------------------------------------------------
% With DC value: Rdc = 30 ohm, R start from 60 Hz
% (total 26 points)
% expo = output2 ( the best value fmin found)
%--------------------------------------------------------------
% insert the DC point to f, R array
rNew(1) = 30; %R dc = 30 Ohm
rNew(2:26) = r(1:25);
fNew(1) = 0;
fNew(2:26) = f(1:25);
%----------------------
% Case 1: expo = 2
%----------------------
% Linear Regression with (fh/f1)^2 up to N1 points)
%N1 ;
f3 = fNew(1:N1);
expo = 2;
for i=1:N1
h3(i) = (f3(i)/60)^expo; %use the fit results
end;
r3 = rNew(1:N1);
[p3,s3] = polyfit(h3,r3,1)
err21 = getfield(s3,'normr')
y3 = polyval(p3,h3);
figure;
plot(f3,r3,'r.:',f3,y3),grid;
xlabel('f (Hz)');
ylabel('Rac');
legend('Test data','Fit data',2)
title('Linear regression for Rac (fh/f1)^2 including DC value ');
axis tight;
%----------------------
% Case 2: expo = output2
%----------------------
output2 = fmin('Newobj2',0.5,2) % myobj2: including DC point
62
f4 = fNew(1:N1);
disp('------With DC value: R start from 0 Hz, expo = best')
r4 = rNew(1:N1);
[p4,s4] = polyfit(h4,r4,1)
err22 = getfield(s4,'normr')
y4 = polyval(p4,h4);
figure;
plot(f4,r4,'r.:',f4,y4),grid;
xlabel('f (Hz)');
ylabel('Rac');
legend('Test data','Fit data',2)
title('Linear regression for Rac (fh/f1)^expo including DC value ');
axis tight;
List E.2 Program for finding the transition point of the two-section fit and the best
Curves.
%================================================================
% Find the best transition points and fit curves for 2KVA XFMR Short
% Circuit test data ( AC winding resistance R )
%
% change the variable "expo1" can set if the first section fit
% is using 2nd order or a optimal value.
% the total fit error is in array err
%================================================================
clear all;
close all;
global r f N0 N;
63
r4 = dat4(1:N,11)'; % different column with data12_1
%--------- Average value of these 3 data set -------------
r = (r1 + r3 + r4)/3;
k = 1;
for N0 = 3:22; %N0 is global, the number of points used for fitting
f1 = f(1:N0);
r1 = r(1:N0);
[p,s]= polyfit(h,r1,1);
err1(k) = getfield(s,'normr'); % the error of firs part square fit
p_a(k) = p(1); %keep the data;
p_b(k) = p(2);
for i=N0+1:N
h2(i-N0) = (f(i-N0)/60)^expo2(k);
end;
h22 = h2(1:N-N0);
r2 = r(N0+1:N);
[p1,s1] = polyfit(h22,r2,1);
err2(k) = getfield(s1,'normr');
p1_a(k) = p1(1); %keep the data;
p1_b(k) = p1(2);
err
err1
err2
expo1
expo2
x = f(3:22);
plot(x,err,'ro:'),grid;
xlabel('f (Hz)');
ylabel('Total Error');
axis manual;
axis([0 2700 0 140]);
title('Total Error (transition point moves from the 3rd - 22nd(300 Hz-
2580 Hz) 10 kVA');
64
figure;
plot(x,expo1,x,expo2),grid; % r: real data; y0: fit data
xlabel('f (Hz)');
ylabel('exponent');
legend('firt part','second part',2)
title('exponent found ');
axis tight;
r1 = r(N0+1:N);
[p1,s1] = polyfit(h3,r1,1)
errSec = getfield(s1,'normr');
65
APPENDIX F
66
Appendix G
Test Transformer:
(1) 2 KVA single phase, dry type,4 winding 120/240 Volt transformer Square D cat. No.
2S1F
(2) 10 KVA amorphous steel core single-phase oil filled distribution transformer
7200-120/240-V
Lab Equipment
Computer Hardware
Computer Software
67