A
Report on 10th Thomas Alva Edison Memorial Lecture
PES
Distinguished Lecturer (DL) Dr Sivaji Chakravorti, Professor of
Electrical Engineering, Jadavpur University, Kolkata 700 032, India landed
in Delhi in the morning of November 20, 2007 (Monday).
For
the year 2006,
Prof. Chakravorti
presented the techniques to identify impulse faults in transformers. Impulse tests are done to
assess the ability of winding insulation to withstand the surge voltages in
use. According to Standard - IEC 60076, fault
identification is based on incident voltage and winding current waveforms
recorded at calibrating level and BIL during the impulse test. It requires
specialized knowledge - often unshared. The major faults can be identified
easily but it could be problematic in the case of minor faults. To determine
the location of fault along the length of the winding, the winding length is
divided into three equal parts. Therefore the fault location is to be
identified within one-third of the winding length to provide the necessary
information and interaction in order to make unambiguous fault diagnosis.
He
classified various faults into series faults occurring between the discs or turns and shunt faults occurring between
the winding and earthed components like tank, core, etc. The fault
classification time-domain parameters are (i)
difference between the area under the BIL current wave and the calibrating
current wave, (ii) deviations in the times of occurrence of the respective peaks
in the two current waves under comparison, (iii) average absolute errors of the
two normalized current waves under comparison, (iv) distance vectors calculated
for calibrating and BIL levels for voltage and current, and (v) average
peak-to-peak oscillation magnitude of the two current waves under comparison.
Moreover, the fault classification frequency-domain parameters are (i) area under the transfer function curves at the
calibrating and BIL levels, (ii) first resonant frequencies of the two transfer
function curves, and (iii) second resonant frequencies of the two transfer
function curves. Reasonably accurate fault identification is achieved
within 33% of the winding length with the clustering technique using 3 fault
classification parameters.
In
latter part of the seminar, an advanced technique is presented to identify the fault within 10% of the
winding length. To achieve this goal more detailed information is required
about the frequency components. It is preferable if the current response is
decomposed into low and high frequency components. The methodology adopted for
this purpose is to decompose the current response into low and high frequency
components, extract the time-frequency domain parameters out of the decomposed
components and the use pattern classifier based on time-frequency domain
parameters for fault location.
This
seminar covered various types of impulse
fault diagnosis techniques in transformers and their usefulness in power
system applications. Various modeling techniques of impulse faults in
transformers namely Fourier Transform
and Wavelet Transform in time-frequency
domain Analysis were briefly discussed.
The simulated results in EMTP and analog models were presented and
correlation was made with test results from the field measurements.
This presentation was of interest to all power engineering
professionals, who came from academic institutions as well as power utilities
and equipment manufacturers. In all 58 persons attended the event.
Earlier in the
daytime Prof Sivaji Chakravorti gave a talk on ‘Condition Monitoring of Paper-Oil Composite
Insulation’ in the Lecture Hall of Central Electricity Authority (CEA),
