Features and Benefits of UHF Partial Discharge Monitoring Systems for GIS
N. Achatz, J. Gorablenkow, U. Schichler B. Hampton, J. Pearson
SIEMENS AG Diagnostic Monitoring Systems Ltd. Power Transmission and Distribution United Kingdom High Voltage Division, Germany dms@dmsystems.co.uk uwe.schichler@siemens.com
Abstract: This paper describes the features and benefits of ultra high frequency (UHF) partial discharge monitoring (PDM) systems for gas-insulated switchgear (GIS). The main features of two high-performance PDM systems are described in detail including explanations of the data displays, and the interpretation of partial discharge data by expert systems. The benefits of such monitoring systems are summarized by evaluation of statistical data. An outlook for future PDM system features is given.
INTRODUCTION
GIS have been in operation for more than 30 years and they have shown a high level of reliability with extremely low failure rates. This is the result of quality assurance during the complete design and manu- facturing process as well as during the erection and on- site commissioning. The return of experience shows that some of the in-service failures are related to defects in the insulation system. Many of these defects can be detected in service by continuous PD monitoring based on the UHF technique. Different PDM systems are nowadays available and all systems operate on the same principle of UHF technology, which has not changed over the last decades after the first experiments were made 25 years ago [1]. The difference between them lies in their PD detection sensitivity and the system features. The user-friendly display of the PD data, the accuracy of automatic PD identification and of course the reliabilty of the PDM system equipment itself are important facts which have to be taken into account.
PRINCIPLE OF UHF PD DETECTION
The principle of the UHF technique was developed in the UK more than 20 years ago, and is now well-known worldwide after being adopted by many leading switchgear manufacturers and utilities. However it may be recalled that the current pulse which forms the partial discharge has a very short risetime, which recent measurements have indicated can be less than 70 picoseconds. The rising edges of these pulses excite the GIS chambers into multiple resonances at frequencies of up to at least some GHz. Although the duration of the current pulse is less than a few nanoseconds, the microwave resonances persist for a relatively long time, typically a few microseconds. They may readily be picked up by UHF couplers fitted either inside the GIS chambers, or over dielectric apertures in the chamber wall. Whether internal or external couplers are used, the UHF signals can be amplified and displayed in different ways where their characteristic patterns reveal the nature of any defect that might be present in the GIS. More recently, they may be analysed automatically with good accuracy. With this early warning of any impending breakdown, utilities can take appropriate action to virtually eliminate enforced outages of their GIS.
Extensive investigations in laboratories have confirmed that PD detection using UHF technology results in higher or at least the same sensitivity as detection by PD measurement as set out in IEC 60270. For PDM system application at GIS substations a sensitivity verification procedure was published by Cigr based on a detection level of equivalent to 5 pC [2].
Figure 1: 420 kV GIS with PDM system
UHF PDM SYSTEMS FOR GIS
Different PDM systems are nowadays applicable for PD monitoring of GIS in service. The design and main features of two high-performance PDM systems [3-7], which are used worldwide, are described in the following subchapters. PDM system: type A
The main components of this PDM system are the following (Fig. 2): UHF couplers - antenna for UHF signals PD converter box - signal conversion from UHF to low frequency, noise rejection central processor unit - monitoring of data stream data concentrator - share/fit together of data stream operator unit - data display, archiving, remote control expert system - identification of PD defect
The PDM system monitors more than 120 UHF couplers simultaneously, enabling the measurement of phase resolved partial discharge (PRPD) charts and PD short term trends (STT) as well as the presentation of all results as history trend diagrams [5-7]. In addition, an automatic noise suppression and an expert system for computer based identification of the PD defect is integrated. The PDM system output is connected to the substation control system to give warning and alarm signals in case of any critical PD activity.
Figure 2: PDM system hardware arrangement
Three single phase UHF/LF-converter which acts as an amplifier and peak detector are the main parts within the PD converter box. In the frequency range examined, the sensitivity limit for UHF signals is set at -80 dBm. The dynamic range of the UHF/LF-converter input covers all signals between -80 dBm and 0 dBm (Fig. 3). Furthermore the detection of fast transient signals caused by switching operations, and a high voltage phase angle detection, are achieved.
Figure 3: UHF/LF-converter ouput versus signal input left: input (dBm), right: input (mV)
For the automatic identification of a PD defect it is necessary to measure and store the PD activity as a phase resolved pulse sequence (PRPS). So the aquired PD pulse amplitude, time of occurrence and the phase angle are available for further analysis. All diagnostic systems based on pattern recognition use the com- ponents feature extraction, classification algorithm and reference database for evaluation of the PRPS data. The applied redundant PD diagnosis system (RDS) has a redundancy in its classification algorithms, and this system was tested with 1247 UHF data sets coming from industrial conditions. The three classifiers are: L2- distance, neural networks and fuzzy logic. This RDS was continuously updated and improved. The actual version of RDS is based on a hierarchical concept and the extraction of -, u/- and u-features. The diagnosis result is that 96.6 % of the PD defects are identified in the correct way and only 3.4 % of the PD data sets are recognised as unknown or wrong [8].
For a risk assessment based on PD diagnostics many factors have to be taken into account. These include not only the knowledge about the type and location of a PD defect, but also exact knowledge of the GIS design, the experience of the GIS manufacturer and the utilities operating and grid conditions. An additional factor for risk assessment on moving particles is the occurrence of PD bursts, which are an indication for the breakdown probability. Fig. 4 shows PRPD patterns from moving particles with PD burst indication on top of the chart.
tag rail for phase resolved PD bursts
Figure 4: PRPD of particles with PD burst indicator [5]
PDM system: type B
This PDM monitoring system, currently installed in more than 45 GIS worldwide, consists of the following basic parts (Fig. 5): UHF couplers - to take the UHF signals from the GIS. Optical Converter Units (OCUs) - each OCU con- tains circuitry to detect and process the UHF signals from one 3-phase set of couplers. The UHF data is then transmitted via an optical fibre to the equipment cabinets, which are often located in the Relay Room. The equipment cabinets contain the electronics to receive and handle the streamed data from the OCUs, and the PC and Control Unit for data storage and display.
Optical Converter Unit Equipment cabinets UHF coupler
Figure 5: General arrangement of the radial PDM
The features of the UHF discharge pulses most useful for interpretation purposes are their amplitude, point on wave, and the intervals between them. These parameters enable typical defects such as fixed point corona, free metallic particles and floating electrodes to be identified. Other defects also have their own distinctive features [3,4]. The UHF data may be displayed in any way which reveals the characteristic patterns typical of the defects causing them, as, for example, that of corona in the 3D pattern of Fig. 6.
Figure 6: Busbar corona, streamers and leaders
A pattern recognition system developed by the PDM manufacturer enables PD signals to be identified automatically, and with high accuracy. This system is especially important to utilities who may be monitoring many hundreds of UHF couplers on their GIS, since it removes the need for an expert to routinely examine the PD data patterns.
The three main areas of soft computing used in pattern recognition tasks are artificial neural networks (ANNs), genetic algorithms and fuzzy logic. These approaches have been combined with more traditional statistical techniques to form a highly-effective hybrid classifier. The stages in this, illustrated in Fig. 7, are to:
remove noise from the raw input data extract complementary features, and form the input vectors for a number of classifiers combine the classifier outputs using a trainable combination function monitor the overall process by a control system that can bypass and modify any of the preceding stages, before making the final classification.
input noise removal feature extraction other classifiers fuzzy logic ANNs genetic classifier combination function control system output
Figure 7: System for classifying PD signals
The following performance figures were obtained for the ANN component working in isolation, having been trained using a database of approximately one million exemplars. The modular ANN was trained on 80 % of the database, and then tested on the remaining 20 %. The results of the tests are shown in Table 1.
Signal type Accuracy of identification Free Particle 97.5 % Chamber Corona 98.3 % Busbar Corona 98.1 % Floating Component 99.6 % Insulation 97.0 % Mobile phone signals 98.0 % Radar 98.6 % Various non-PD signals e.g. switching, lights noise, motor noise 98.5 %
Table 1: ANN performance using a test set of data
degrees cycle no BENEFITS OF PDM SYSTEM APPLICATION
In general the benefit of PDM system application can be described by the successful detection of PD defects during GIS operation and the prevention of related breakdowns. A lot of information is available nowadays for the statistical evaluation of PDM systems. Two different databases are available:
A) One type of PDM system continuously taking PD data from 363 UHF couplers, which are located at six GIS with rated voltages from 245 kV to 550 kV (one single GIS manufacturer). The PD detection sensitivity was equivalent to an apparent charge of 5 pC and often much better. PD data of a seven years period and 223 bay-years are available in total. Only one defect was found in service. A floating electrode defect was detected by the PDM system and confirmed by visual inspection during the repair work. For all GIS with PDM system application no in-service breakdown has occured so far. The PD defect rate for the available data basis can be calculated as follows:
Number of UHF couplers 363 Number of bay-years 223 Number of UHF coupler-years 1356 Total number of defects (critical PD) 1 Defects per 100 bay-years 0.45 Defects per 100 UHF coupler-years 0.074
B) Since 1996 a PDM system manufacturer has been monitoring data from its PDM systems, which now contain more than 3000 UHF couplers (different GIS manufacturers, GIS with rated voltages from 132 kV to 800 kV). The results from more than 1000 bay-years of in-service operation show that outages were obtained to remove 47 defects that would almost certainly have caused complete breakdown of the GIS, had no action been taken. Over the same period, breakdown occurred for unknown causes in only 2 GIS that were being monitored.
An in-service breakdown in a GIS typically takes more than a week to repair, and the costs of this, together with the consequent circuit disruption and loss of supply in a single outage, usually far outweigh the initial cost of high-performance PDM systems. A detailed life- cycle-cost calculation based on typical or type-related failure rates and outage costs for GIS substations can confirm the above mentioned general statement. Several utilities recognise the evident benefits of being able to prevent most of their GIS breakdowns, and are installing high-performance PDM systems to their major GIS [9].
FEATURES OF FUTURE PDM SYSTEMS
Both described PDM systems have been updated and improved continuously in the past according to the return of experience and customer requirements. The technical features of future PDM systems will be as follows: application of a standard UHF front-end with high dynamic range and intelligent noise suppression monitoring of more than 200 UHF couplers by use of PDM electronic board modules application of standardized PDM system equipment with high reliability and an expected lifetime of more than 15 years user-friendly software with flexible data display including trend diagrams and customized reporting PD identification with high accuracy and support to the risk assessment based on PD diagnostics behaviour more like a black box with submission of PD alarms to the substation control system only in case of service relevant PD activity.
CONCLUSION
The principle of the UHF technique was developed more than 20 years ago, and is well-known worldwide. Different PDM systems are nowadays available for PD monitoring of GIS in service. The difference between them lies in their PD detection sensitivity and their PDM system features. Several utilities recognise the benefits and are installing high-performance PDM systems to their major GIS in order to detect PD activity at an early stage. The application of standardized and high-performance PDM equipment will enhance the reliability of future PDM systems.
REFERENCES
[1] Hampton, 21 years on, UHF monitoring comes of age, CIGRE Study Committee 15, Gas Insulated Systems Symposium, Dubai, UAE, 2001 [2] Cigr TF 15/33.03.05, Partial Discharge Detection Systems for GIS: Sensitivity Verification for the UHF method and the Acoustic method, Electra, No. 183, 1999 [3] Hampton, Newlands, Detecting and Locating Par- tial Discharges in metal clad Equipment, 4 th Int. Symposium on Gaseous Dielectrics, USA, 1984 [4] Pearson, Hampton, Sellars, A continuous UHF monitor for gas-insulated Substations, IEEE Transactions on Electrical Insulation, Vol. 26, 1991 [5] Schichler, Gorablenkow, Experience with UHF PD Detection in GIS Substations, 6 th ICPADM, Xian, PR of China, 2000 [6] Gorablenkow, Huecker, Schichler, Application of UHF Partial Discharge Monitoring and Expert System Diagnosis, IEEE ISEI, USA, 1998 [7] Schichler, Gorablenkow, Diessner, UHF PD Detection in GIS Substations during on-site Testing, 8 th DMMA, Edinburgh, UK, 2000 [8] Aschenbrenner, Kranz, Schichler, Hierarchical Diagnoses Systems for Partial Discharge Identi- fication and Risk Assessment in GIS, 12 th ISH, Bangalore, India, 2001 [9] Yoon, Behrmann, Pietsch, Pearson, Further Results of the Partial Discharge Monitoring System at Labrador and Ayer Rajah 400 kV GIS Sub- stations, 13 th CEPSI, Manila, Philippines, 2000