Workshop 2001, Alexandria, Virginia, December 3 & 4 , 2001
PD DIAGNOSTICS – ITS HISTORY AND FUTURE
A. Bolliger, E. Lemke
LEMKE DIAGNOSTICS GMBH, GERMANY
LEMKE DIAGNOSTICS AG, SWITZERLAND
HV TECHNOLOGIES, INC: USA
1. INTRODUCTION
A significant trend in the development of electrical power apparatus is the increase of the power and
size of the units. This requires severe demands on increased reliability [1;2]. Today's high voltage
insulation technology therefore requires modern testing procedures. In this respect increasing
attention is being paid to the development of predictive diagnostic tools. Against this background,
there is no doubt that the recognition of partial discharges (PD) is of great importance, because PD
phenomena can be regarded as the forerunner for ageing phenomena in electrical insulation.
Despite the recent progress in PD diagnostics, we should remember that the basis for this has been
established over a long-term historical development. Because of the great amount of existing
publications, however, it seems impossible to report in detail on the complete chronological
development of partial discharge technology. This presentation will therefore only feature some
examples.
2. HISTORICAL REVIEW
The very beginning of partial discharges recognition goes back to the year 1777, in which
LICHTENBERG reported on novel results of experimental studies [3] during a Session of the Royal
Society in Göttingen. Using VOLTA's "Elektrophor" the "Harzkuchen" this instrument showed fantastic
dust figures like stars and circles (Fig. 1(. It lasted more than 100 years until it was clarified, that the
dust figures represented dielectric surfaces discharges appearing as electrical discharge channels.
Fig. 1: Dust figures produced by surface discharges under positive (left) and negative polarity (right)
of the applied voltage, presented by Lichtenberg in 1777 [3]
In 1873 MAXWELL published "A Treatise on Electricity and Magnetism" [4]. His theoretical work is of
fundamental relevance for both, the design of instrumentation for electrical PD detection and the
development of physical models for better understanding of the very complex PD phenomena.
Fig. 2: Experimental set-up for demonstration of the existence of electromagnetic waves
Hertz 1886
In the year 1896, HERTZ demonstrated with an impressive experiment, according to Fig. 2, the
hypothesis of MAXWELL on the existence of electromagnetic waves and their propagation in space
and time. In principle, his experiment can be regarded as the first application of the inductive field
coupling mode, nowadays used for in the Lemke Probe LDP-5 [5].
Both electrical and non-electrical procedures are used for PD recognition. Due to the limited time I will
concentrate only on some highlights in the development of electrical PD detection. These methods
were of highest technical importance in the first decades of the 20th century, forced by the practical
application of electrical power and the newly developed HV equipment
transmission and distribution of electricity.
for the generation,
The first measuring device used for the electrical detection of PD events was the loss factor bridge
according to SCHERING, developed in 1919 [6] and applied for this purpose in 1924. One year later,
in 1925 SCHWAIGER recognized the radio frequency character of corona discharges [7]. This finding
can be considered as the basis for the introduction of radio interference meters for evaluation of the
noise level of corona discharges. This RIV test is still widely used, especially in North America. In
Germany,this kind of instrument has been first used by DENNHARD in 1937 [8].
An essential progress in PD detection was achieved when electron beam oscilloscopes were
available. In 1928, LLOYD and STARR used two pairs of perpendicular deflection plates inside the
BRAUN tube [9] for displaying PD events. Here, one pair was subjected to the instantaneous test
voltage, whereas the other pair has been connected to a capacitor, used for the accumulation of the
generated corona charge. This early approach, called parallelogram method, allowed an excellent
wide-band measurement of corona discharges in wire-plane arrangements, used for the simulation of
HV overhead transmission lines.
In 1928 BYRSTLYN introduced a simple equivalent circuit for the assessment of PD losses under AC
stresses [10]. His approach "Funkenstrecke mit Vorkondensator" has been systematically investigated
by GEMANT and PHILIPPOFF by means of oscillograghic techniques in 1932 [11]. In this way, they
could explain the sequence of discharge events per cycle of the applied AC voltage.
It should be noticed, that the previous mentioned parallelogram method can be regarded as the
predecessor for the integrating bridge, used by DAKIN and MALINARIC in 1960 [12]. This tool is
nowadays also applied, in particular for physical PD studies (Fig. 3), as reported in [13]. The
integrating circuit has been modified by our Dr. Lemke in 1976 in order to study so-called pulse-less
discharges at high sensitivity [14]. Some selected measuring examples are shown in Fig. 4.
Fig. 3: Parallelogram method for investigation characteristic types of PD
Izeki and Tatsuta 1984 [13]
Fig. 4: Measuring circuit for sensitive detection of PD in the low frequency region and chosen test
results for air, SF6 and oil, Lemke; Hauschild; Nagel 1975
At the early stages, an essential progress in development of sensitive PD detectors has been
achieved by application of narrow-band amplifiers based on resonance circuits, as reported by
ARMAN and STARR in 1936 [15]. In 1954 the first portable PD detector was commercially available,
designed by MOLE [16]. Basic characteristics of PD calibrators have also been submitted by him in
1970 [17].
It seems noticeable, that up to the 1980's commercially available PD detectors used only a measuring
frequency band below 1 MHz. One of the first commercially available wide-band PD detectors has
been developed by the co-author Dr. Lemke. The applied measuring principle was based on an active
(electronic) integration of the wide-band pre-amplified PD pulses. The upper cut-off frequency of the
wide-band pre-amplifier was about 10 MHz. Under this condition, PD events could not only be
detected but also located, as reported in [18]. Additionally, an effective noise rejection could be
achieved by this non-conventional measuring principle, as shown in Fig. 5.
Fig. 5: Features of a non-conventional wide-band PD detector for noise rejection Lemke 1975 [14]
Today, there is no doubt that essential advantages arise, if a high frequency range of the origin PD
pulse frequency spectrum is used for PD recognition. In 1966 BAILEY estimated the duration of origin
PD pulses in cavities of solid dielectrics as short as some nanoseconds [19]. This has been confirmed
by practical measurements carried out by FUJIMOTO and BOGGS in 1981 [20] and by BOGGS and
STONE in 1982, applying high speed oscilloscope technique up to 1 GHz bandwidth [21]. Records of
typical origin PD pulses are shown in Fig. 6. Today the UHF-technique can be considered as a
substantial tool for PD diagnostics, in particular for gas-insulated switchgear (GIS) and power cable
accessories.
Fig. 6: Oscilloscopic records of PD pulses from trees (left) and floating particles (right) using ultra
wideband detection up to 1 GHz
Boggs and Stone 1982 [21]
In modern PD diagnostics, not only individual PD pulses are of interest. Much more information can be
achieved by means of the so-called PD pattern analysis (PDPA). So the
occurrence of sequences of PD pulses versus the phase angle of the applied AC test voltage as well
as the statistics of the pulse height distribution may give additional information. The first sophisticated
computer-based system in this respect has been developed by TANAKA and OKAMOTO in 1978 [22].
Their approach according to Fig. 7 provides the following three major types of statistical distributions:
1. The discharge rate versus the apparent charge
2. The discharge rate versus the measuring time
3. The discharge rate versus the phase angle.
Fig. 7: Minicomputer based PD measurement system for PD pattern recognition Tanaka Okamoto 1978
[22]
As well known, sensitive PD measurements may be disturbed by electromagnetic noises. Hence, a lot
of work has been done in order to reject external interferences. In 1973 OKAMOTO et al. reported on
the noise suppression in case of PD testing of 500 kV transformers [23]. Also in 1973 PRAEHAUSER
discussed the capability of the balanced PD bridge for elimination of external noises [24]. In 1975
BLACK presented a pulse discrimination system for discharge detection in noisy environments [25] as
shown in Fig. 8. Later on, numerous approaches have been adopted, such as the
- averaging technique,
- cross correlation technique,
- filter technique based on passive, active and adaptive filters
- pulse discriminator technique and windowing
Fig. 8: Pulse discrimination system for noise rejection
Black 1975 [25]
3. FUTURE TRENDS
3.1. General aspects
There is a growing trend, to use the HV lab, non-destructive PD measurement also for predictive
diagnosis tests under on-site condition. The availability of even more powerfull computers has allowed
to development of so called experts systems for PD Analysis, PD Statistics and PD Diagnosis, used
for lab applications, on-site applications or for permanent PD monitoring on HV apparatus. Results of
this recent development as well as the discussions within the relevant CIGRE working groups may
give ideas on future tasks in the development of PD diagnostics , such as:
1.
Continuous revision of the existing standards for PD measurements with respect to both,
the improvement of the reproducibility of PD tests performed in different laboratories and
the possibilities of digital PD measuring technique.
2.
Development of more powerful noise rejection procedures in order to discriminate
electromagnetic interferences significantly from PD events to be detected.
3.
Improvement of the reliability of monitoring systems used for long-term diagnostics, which
must be better than those of the monitored HV apparatus.
4.
Further development of sophisticated expert-systems including multiplexed data and
simultaneous processing technology for quick recognition of dangerous PD faults.
5.
Design of HV apparatus with built-in ultra-wideband PD couplers and development of
advanced PD sensors in order to perform more informative PD diagnostics under service
conditions.
3.2. Example of a Digital Partial Discharge and Monitoring System:
Concept of a computer-based PD monitoring System
A. Hardware
Due to the enormous wide range of applications of PD measuring systems the developed
instrumentation can be flexible composed according to the particular measuring situation. Hence, the
new developed PD monitoring system contains different package modules, as evident from Fig. 9.
Fig. 9: Block diagram of the on-line PD monitoring and warning system LDWD-6
B. Available Software of a state-of-the-art PD Detector
·
Program “PD Analysis”
This program covers the following functions:
-
Evaluation of the captured PD pulses in compliance to the relevant national and international standards
(IEC, VDE, AEIC, IPCEA, ASTM, ANSI, VDE), where the following main PD quantities are evaluated:
apparent charge q
pulse repetition rate n
pulse repetition frequency N
phase angle φi and time ti
-
average discharge current I
discharge power P
quadratic rate D
inception / extinction voltage Vi / Ve
Replay of the above listed PD quantities using an operation panel similar to an audio or video player.
The following display modes are selectable:
1. Conventional phase resolved presentation of the PD pulses like an oscilloscope, where either a
linear time scale (Fig. 10a) or the elliptical scale (Fig. 10b) can be selected. Besides continuous
replay mode, which shows again the PD events appearing during the real-time PD measurement,
individual snapshots can be made at different cursor positions.
2. Conventional time and voltage dependent presentations of standardized PD quantities, such as
q, D and P. Example for this are shown in Fig. 10d and 10e. Changing the start and stop position
of the cursor, the interesting time intervals can be selected accordingly.
3. Sophisticated presentation of the phase-resolved PD pattern according to Fig. 10c, where an
impression on the PD activity is obtained by classification of the pulse number using a colour
code.
4. Phase resolved three-dimensional
presentation, where different PD parameters can be
combined, such as the PD magnitude, pulse number and testing time (Fig. 11).
Fig. 10a Snapshot of phase-resolved PD
pulses using the linear time scale
Fig.10b Snapshot of phase-resolved PD
pulses using the elliptical time scale
10c Record of the phase-resolved PD pattern
10d) Record of the PD level versus the testing time
10e) Record of the PD level versus the test voltage
11a) PD pulse number vs. phase angle and pulse charge
11b) PD charge magnitude vs. phase angle and test time
3c) PD charge magnitude vs. phase angle and test voltage
11c) PD charge magnitude vs. phase angle and test voltage
Fig 11 Typical PC screen shots of three-dimensional presentations obtained by the program
· Program “PD Statistics”
This program covers the following functions:
-
Evaluation of fundamental statistical parameters of the stored phase and polarity resolved PD data,
which supports the identification and classification of PD faults.
-
Presentation of the following graphs:
1. Distribution function of the peak, average and mean values of the PD pulse magnitude versus the
phase angle of the applied test voltage (Fig. 12a).
2. Distribution function Hn (q), which represents the total number of PD pulses versus the PD pulse
magnitude (Fig. 12b). Here the positive and negative pulse numbers are displayed separately.
3. Distribution functions Hn (q) for interesting time intervals, displayed like waterfall diagrams (Fig. 12c).
The time intervals can be chosen by setting the cursors accordingly.
4. Summary of the fundamental statistical parameters „standard deviation, skewness, kurtosis
and cross correlation“ separated for the positive and negative half cycle (Fig. 12d). This parameters can
be considered as a „Fingerprint“ of the detected PD failure.
12a) Statistical distribution functions of phase-resolved PD quantities
12a) Statistical distribution functions of phase resolved PD quantities
12b) Statistical distribution of the total PD pulse number Hn (q)
12c) Statistical distribution functions Hn (q) for five measuring intervals
12d) Calculated fundamental statistical parameters
Fig. 12: Typical PC screen shots obtained by the program window „PD Statistics”
· Program “PD Diagnosis”
As well known, a common mathematical modelling of all PD failures is not available up to now. Only for
exceptional cases a mathematical model exists, which is suitable to describe a subclass of PD problems.
Therefore, an automatized diagnosis system for the identification and classification of PD failures is limited to the
recognition of specific symptoms in PD data records.
In this context it must be noted that the characteristic feature extraction of the PD data record is cut out for the
key position in the quality of the diagnosis result [30]. In the scientific field of the PD fault recognition exits a wide
range of formulations about the suitability of different features to be extracted. In the software program „PD
Diagnosis“ of the LDWD-6 a combination of two independent feature detectors is realized. The Fourier
correlation coefficient of the phase resolved PD pulses is normalized to the number of the test voltage periods.
In order to describe the phase resolved PD pulse distribution, only a limited number of coefficients of the Fourier
series is necessary [30], which is used for the feature extraction array. Additionally, the variation of the
coefficients versus the test periods is inserted to the feature pool. Furthermore, the classical statistical operators
[31] of the derived histogram functions of the PD frequency distribution are included into the feature extraction
matrix.
After the extraction, the two resulting feature arrays are subjected by a classification schedule. The classification
is effected by means of comparison of feature extraction arrays of the actual measured PD data with feature
objects of all existing PD failure records, stored in a reference database. As the classification result, the qualified
probability of the class membership of the classified object array related to already identified PD faults is
evaluated and, after a mutual coincidence check, displayed on the computer screen of the device LDWD-6. As a
result the following two graphs are displayed:
1. PD failure classification (Fig. 15). Here the results of the actual PD measurement are compared with
PD failure types, already stored in the reference databank of the LDWD-6. If desired, the results of
the actual PD measurement can be added to the existing reference databank.
Fig. 15: PC screen shot of the panel „Classification“ obtained by the program window “PD Diagnosis“
3.3. Complex Discharge Analysis System for on-site PD on cables
Fig. 16:
Voltage shape of the CDA test voltage (a) and recorded signals at a test shot (b)
Lemke; Schmiegel; Elze; Rußwurm 1995 [26]
The new developed diagnostic tool ensures not only the measurement of the standardized PD
quantity "apparent charge" but also the location of the PD site in power cables. Because the
procedure bases on the complex analysis of PD events during the tail time t2, i. e. when the cable
capacitance is discharged, the test method is named: Complex Discharge Analyzing (CDA).
This procedure is characterized by the following benefits:
1.
Low voltage stress, because a test level of 2 * U0 seems sufficient for PD recognition.
2.
Time saving, because in general already 5 shots at each voltage level are sufficient, with
respect to the statistics of PD events.
3.
Low power demand, because of the comparatively long charging time of several seconds.
4.
Low weight, because of the low power demand. Hence, non transportation problems.
5.
Mains-independent power supply is possible, because of the low power demand.
6.
PD fault location, because the wideband PD measurement is capable for the reflectometry
method.
Concept of a CDA PD analysis system
Fig 17 Impulse reflectogram of a single PD-Impulse
Fig 18 PD-Faults on XLPE-Cables
3.2. Traveling Wave sensor technique for HV cable accessories
As already reported in [27], in case of PD diagnostics on extruded EHV cables there is only one
chance for sensitive PD recognition in the accessories, if they are equipped with PD couplers. For this
purpose besides conventional capacitive and inductive sensors, so-called travelling wave sensors
(TWS) are used. The latter ensure an ultra-wideband PD detection. Due to this, the detection of PD
pulses as low as 1 pC can be realized, even if the noise level exceeds several thousands of pC. The
in practice well proved principles for PD detecting using TW Sensors is schematically shown in Fig. 19
and Fig. 11; for more details see the references [28] and [29].
Fig. 19:
TWS-technique for PD recognition and location in cable joints.
Upper: Test arrangement
Lower:
Characteristic records at
1- cable joint
5 ns/DIV resolution
2- cables to be connected
3- TW sensors
Left- PD inside the joint
4- Casing of the joint
Middle- PD outside the joint, left
5- VHF/UHF PD probe
cable end
Right- PD outside the joint, right
cable end
Pommerenke; Krage; Lemke; Schmiegel 1975 [28]
Fig. 20:
Set-up for PD detection using Directional Coupler Sensors (DCS)
Pommerenke; Strehl; Kalkner 1997 [29]
4. SUMMARY
It was the intent of this contribution to the PD Seminar, to give a chronological review of the historical
development in the very complex subject of electrical PD detection. Considering future trends it is
noticed, that increased demands on the reliability of HV equipment require advanced diagnostic tools.
Economic aspects and the reliability of diagnostic tools have to be taken into account. Practical
examples, especially in the field of after laying tests of power cables, should underline that the
development is going on, but much more work has yet to be done in the future.
Both Lemke Diagnostics GmbH and HV Technologies, Inc. are committed to continued research and
development in the field of partial discharge testing.
5. REFERENCES
[1]
Dielectric diagnosis and its effect on insulation coordination. CIGRE Session Paris (1990) Joint
group discussion SC 15 and SC 33
[2]
Material reliability for network equipment. CIGRE Session Paris (1994), Panel 3.
[3]
G. Ch. Lichtenberg: Novi Commentarii Societatis Regiae Gottingae tom 8 (1777) 168.
[4]
J.C. Maxwell: A treatise on electricity and magnetism. Clarendon Press, Oxford (1873).
[5]
E. Lemke: A new procedure for partial discharge measurements on the basis of an
electromagnetical sensor. 5th. ISH Braunschweig (1988) paper 41.02.
[6]
H. Schering: Brücke für Verlustmessungen. Tätigkeitsbericht der Physikalisch-Technischen
Reichsanstalt (1919).
[7]
A. Schwaiger: Über die Entladungsvorgänge auf Isolatoren. Rosenthal-Mitt. H. 6 (1925).
[8]
A. Dennhardt: Ursache und Messung der hochfrequenten Störfähigkeit von Isolatoren.
Elektrizitätswirtschaft 34 (1935) S.15.
[9]
W.L. Lloyd; E.C. Starr: Untersuchung der Wechselstromkorona mit dem KathodenstrahlOszillographen. ETZ 49 (1928) 1276.
[10]
Bursteyn: Die Verluste in geschichteten Isolierstoffen. ETZ 9 (1928) 1258-1291.
[11]
A. Gemant, W. v. Philippoff: Die Funkenstrecke mit Vorkondensator. Z. f. techn. Physik 13
(1932) 9, 425-430.
[12]
T.W. Dakin; P.J. Malinaric: A capacitive bridge method for measuring integrated corona-charge
transfer and power loss per cycle. Trans AIEE PAS 79 (1960) 648-653.
[13]
N. Izeki; F. Tatsuta: Three types of partial discharges in electrical insulation. IEEE Intern.
Symp. in El. Ins., Montreal (1984) Conf. Record 327-331.
[14]
E. Lemke: A new partial discharge measuring procedure based on the avaluation of the
cumulative charge. Conf. on PD in Electrical Insulation Bangalore/India (1976) 1-22
[15]
A.N. Arman; A.T. Starr: The measurement of discharges in dielectrics. J. IEE 79 (1936)
81; 88-94.
[16]
G. Mole: The E.R.A. portable discharge detector. CIGRE-Session, Paris (1954) No. 105, App.I.
[17]
G. Mole: Basic characteristics of corona detector calibrators. IEEE Trans. PAS 89 (1970) 198204.
67-
[18]
E. Lemke: A new method for PD measurement of polyethylene insulated power cables. 3rd
ISH, Milan (1979) paper 43.13.
[19]
C.A. Bailey: A study of internal discharges in cable insulation. IEEE Paper No. 31, PP 66-363
(1966) 9 pp.
[20]
N. Fujimoto; S.A. Boggs; R.C. Madge: Electrical transients in gas-insulated switchgear. Trans.
of the March, 1981 Meeting of the Canadian Electric Association.
[21]
S.A. Boggs; G.C. Stone: Fundamental limitations in the measurement of corona and partial
discharge. IEEE Trans. EI 17 (1982) 2, 143-150.
[22]
T. Tanaka; T. Okamoto: A minicomputer-based partial discharge measurement system. IEEE
Intern. Symp. on Electr. Insul., Philadelphia, Conf. Record (1978) 86-89.
H. Okamoto; H. Kenji; J. Tomiyana: Partial discharge tests and noise suppression of 500 kV
transformers. IEEE Summer Power Conf., Vancouver (1973).
[23]
[24]
Th. Praehauser: Teilentladungsmessungen an
Brückenschaltung. Bill. SEV 64 (1973) 1183-1189.
Hochspannungsapparaten
mit
der
[25]
I.A. Black: A pulse discrimination system for discharge detection in electrically noisy
environments. 2nd. ISH Zürich (1975) paper 3.2-02.
[26]
E. Lemke; P. Schmiegel; H. Elze; R. Russwurm: Procedure for evaluation of dielectric
properties based on Complex Discharge Analyzing (CDA) IEEE International Symposium on
Electrical Insulation, Montreal, Canada (1996).
[27]
M. Ogino; M. Ichihara; A. Fujimori: Recent developments in Japan of insulation Diagnostic
technology for extra-high voltage XLPE cable lines. CIGRE Session, Paris (1994) paper 21103.
[28]
D. Pommerenke; I. Krage; W. Kalkner; E. Lemke; P. Schmiegel: On-site PD measurement on
high voltage cable accessories using integrated sensors. 9th ISH, Graz, Austria (1995).
[29]
D. Pommerenke; T. Strehl; W. Kalkner: Directional Coupler Sensor for Partial Discharge
Recognition on High Voltage Cable Systems. International Symp. on High Voltage, ISH
Montreal, Canada (1997
[30]
Gulski, E. Digital analysis of partial discharges. IEEE Trans. On Dielectrics and Elec.
Insulation, Vol 2, No 5
[31]
Hücker, T. UHF Partial Discharge Expert System Diagnosis. 10th ISH Montreal, Canada, 1997