CONTINUOUS ON-LINE PARTIAL DISCHARGE MONITORING OF GENERATOR STATOR WINDINGS

B. A. Lloyd, Member, IEEE S. R. Campbell, G. C. Stone, Fellow, IEEE Iris Power Engineering Inc., Etobicoke, Ontario, Canada M9C 1B2 PE-388-EC-0-2-1998 A paper recommended and approved by the IEEE Electric Machinery Committee of the IEEE Power Engineering Society for publication in the IEEE Transactions on Energy Conversion. Manuscript submitted August 12, 1997; made available for printing March 2, 1998. (Pages 11311137) Abstract: The use of on-line partial discharge (PD) system can be programmed to collect data under various measurements as an indicator of stator winding insulation machine conditions, thus providing a series of consistent condition, is well known. Trending results from on-line PD data points for trending. This paper describes such a system, measurements over time, often allows maintenance personnel and presents the results from several generators where it has to become aware of developing stator insulation problems in been used. motors and generators, and take corrective measures before II. NEED FOR AUTOMATED PD MONITORING failure occurs. However, partial discharge measurements on operating rotating machines can be influenced by conditions Stator winding insulation deterioration is usually a relatively such as humidity, temperature, terminal voltage, and load. slow process, where the time between when significant PD Thus, to provide reliable data which can be trended, it is can be detected and when winding failure may occur, may important to take readings under similar machine operating be several years. Thus measurements which are done once conditions. In remote plants, or on pump storage units which every six months or so, are often adequate for detecting are frequently cycled, it can be both time consuming and stator winding problems. However, there are several costly to ensure that personnel are available to take PD situations where continuous and automated monitoring may readings under ideal conditions occur. This paper describes be advantageous: the development of a automatic partial discharge monitoring 1. There is a trend for hydrogenerator plants to be svstem which is permanently connected to relevant machine remotely operated. Since such power plants are often sensors and can be configured to collect PD data when located in remote sites where it is expensive to transport suitable conditions are met. The operating principles of the test personnel to the site, an automated PD monitoring system are described, and some results are presented from system allows measurements to be made, without data collected to date on 30 hydrogenerators. sending personnel to the plant. This can reduce overall test costs. I. INTRODUCTION 2. The stator winding PD activity is often affected High voltage stator windings often produce partial discharge by motor/generator operating conditions, including (PD) as a symptom of many conditions, including loose winding temperature, load, voltage, etc. [2]. Although windings, overheated insulation, and contaminated windings these operating condition-dependent results help [1-5]. Various methods have been developed to measure the in interpreting the type of deterioration occurring, PD activity in a winding. The most popular methods are they often make it difficult to trend PD data over those which can be implemented during normal operation time. To obtain trendable data, (which is important of the machine, since costly outages can be avoided. In for interpretation) the machine must be tested under research conducted over the past 20 years, PD measurement the same load, temperature and voltage conditions. methods have been developed which make an on-line For critical units which are subject to frequent load PD test easier to perform and interpret, allowing noncycling, or pumped-storage generating units, it is specialized plant personnel to reliably perform the test [2,3]. often difficult for the plant operators to recreate exact In hydrogenerators, this method depends on high voltage operating conditions from test to test. An automated PD capacitors installed in the winding as well as specialized monitoring system avoids this problem by continuously electronics to eliminate the electrical interference (noise) measuring the motor/generator operating conditions, common to operating machines. Once the PD is separated and then taking the PD measurement only when the from the noise, standard electronic pulse counting techniques desired operating conditions occur. can be used to accurately determine the magnitude, number 3. Collecting PD data at various temperatures and load and phase position of the PD activity. points can itself be useful in the data interpretation The development of instrumentation which can be used process. For example, if a winding is loose, an increase by non-specialized plant personnel to reliably perform a PD in generator load (and thus the forces on the stator test, together with recent advances in computer technology, bars) will cause the PD activity (i.e. slot discharge) has made it feasible to, develop a system which can be to increase dramatically. A continuous PD monitoring permanently installed in a power plant, and to monitor system can collect data at specific time intervals, and automatically and continuously the PD activity. Such a

under selected operating conditions, thus providing multiple points of compatible trend data. Utilizing such data maintenance engineers have a more complete picture of the windings condition. 4. Several expert systems are being developed which continuously monitor all the sensors in a generator (for example temperature, vibration, current, voltage, hydrogen gas pressure, etc.) to determine if any problems are occurring in the generator. This provides plant operators, as well as maintenance personnel, with warnings that a problem may be developing, and also gives advice on corrective action. Such expert monitoring systems include the Generator Expert Monitoring System (GEMS) for turbine generators and the Advanced Condition Monitoring System (ACMS) for hydrogenerators [6- 8]. Since partial discharge activity is an important source of information with regards to stator winding insulation, it is desirable that generators equipped with such expert systems have ready access to PD data. An automated PD monitoring system can fulfill this function by making PD data available on acontinuous basis to such expert systems. The need for a continuous and automated PD monitoring system was first recognized in the early 1980s when Ontario Hydro constructed a prototype system to investigate the effect of operating conditions on PD. In the late 1980s other researchers also developed a non-commercial quasicontinuous system [9]. However, the computer technology and computer communication systems available at the time made the earlier systems relatively unreliable and/or relatively expensive and difficult for utility personnel to use. With the advent of superior computer technology, a monitoring system which can be used by non-specialized plant staff has been developed and is described below. III. SYSTEM ARCHITECTURE To measure the machines operating conditions and to record PD data, each motor or generator monitored requires a Data Acquisition Unit (DAU) to beinstalled near the machine. The DAU measures the PD activity from all the PD sensors installed on the machine; separates the PD from noise; and counts the number, magnitude, and phase position of the PD. In addition, the DAU has digital and analog inputs to measure the motor/generator operating information such as load, temperature, voltage, and breaker status. The DAU contains a DOS based 80386 computer for control of the PD and analog sensors, and to communicate with a central system controller via an Ethernet Local Area Network (LAN). Up to eight DAUs in a plant are serviced by one system controller, over a LAN (Figure 1). The System Controller is a PC running a dedicated monitoring application. The System Controller acts as a file server for storing data sent from the DAUs, and is also responsible for configuring and controlling the DAUs. In addition, via the System Controller,

a user defines, for each DAU, the generator operating conditions (i.e. triggers) for which a PD measurement should he made, and how often the measurement is made, assuming the conditions occur for a long time. There is also a facility for continuously measuring the PD activity at each DAU, and for raising an alarm if the PD activity is above a defined threshold. In addition, there is a program for displaying and trending any saved PD data. The System Controller can be remotely accessed via a Wide Area Network (WAN) or modem. This allows personnel in engineering offices to define or change trigger conditions and alarm levels, as well as to download test results for display on a local computer. IV. HARDWARE DESIGN The electronics used in the DAUs is based on that of portable instruments in use for many years [2,3]. This ensures that PD test results will be trendable no matter which method is used to collect data. The input electronics is based on four single channel pulse height analyzers which can be sequentially scanned through a range of magnitudes to determine the number and magnitude of the input pulses. Four channels are needed to measure simultaneously both polarities from two sensor inputs. Associated logic determines if the incoming pulse is noise or PD, according to the type of sensor used [2,3]. To determine the pulse count, the content of a counter is incremented each time a pulse occurs in the present magnitude window. A counter is associated with PD pulses from each sensor and polarity. Counters are also associated with various noise conditions. Once every one-one hundredth of an ac cycle (167 microseconds for 60 Hz), the contents of the counter are downloaded to the DAlTs memory. The computer associates the present ac phase position with pulse count and magnitude. The analog bandwidth is a minimum of 350 MHz. Most machines are equipped with 6 to 12 sensors, to help locate where in the stator winding the PD is occurring. A multiplexor within the DAU, controlled by the DALTs computer, is used to sequentially scan through all the PD sensor inputs. The multiplexor also has one open channel which is used to completely isolate the DAU from the sensors installed in the high voltage winding, to minimize the potential for damage to the electronics during transients. That is, unless the signals from a sensor are actually measured, the sensors are not connected to the electronics. Inaddition, a test input is permanently attached to one of the multiplexor channels to verify the proper operation of the DAUs PD collection electronics. Most installations were made in 1996 and 1997, but one installation has been operating since February 1995. After the physical effort of installing the DAUs and LAN, it generally only takes 3 to 4 days at site to commission the system. Figure 3 shows approximately a years worth of data that has been acquired from one site, collected with the

in Figure 3 and 4, (he machine in Figure 5 shows little or no temperature dependence of PD. This machine is a large pump storage motor/generator which sees rapid changes in load. Figure 5 shows the two dimensional pulse height analysis for the machine at more or less constant load, but for two very different temperatures. In this case there is little difference in the PD activity. Figure 6 is an example of a machine which shows very

system in Figure 1. As can be seen from Figure 4, there is a substantial decrease in PD activity with any increase in stator winding temperature. This particular generator has an asphaltic mica insulation system, which is known to be delaminated. This behaviour is expected since the asphalt tends to expand as the temperature increases, closing up the voids and thus reducing the PD activity. This example shows the importance of comparing PD results only when the operating conditions are similar. In contrast to the temperature sensitivity of the generator

load dependent behavior that can be automatically detected by the continuous monitoring system. This machine is a large pump storage generator which is suffering from slot discharge. In this case, the PD monitor has collected data at no load, while the machine is still hot, and again at full load, at approximately the same temperature. The significant difference in PD between the no load and full load condition is usually a result of the change in bar forces with load and indicates that the bars may be loose in the slot. Conventional analog to digital converters are incorporated in each DAU to measure the generator/motor voltage, power, and stator winding temperatures. The electronics can be configured to measure any type of analog signal (from transducers, current loops, RTDs and/or thermocouples) and to yield digital data for the DAU computer. Calibrated,

a complete scan of the PD sensors. For example, a PD measurement on a generator may be taken once every 3 months when the load is above 225 MW and the stator winding temperature is above 85 C and once every 3 months when the load is above 225 MW and the stator winding temperature is between 50 C and 55 C. Utilizing such a trigger, the effect of temperature on the PD can be investigated. a module to control communication over the LAN to each of the DAUs connected to the System Controller, as well as to a remote computer. A proprietary NetBIOS-based remote procedure call (RPC) protocol was developed to provide communication between the DAU and System Controller. This protocol allows for 3 distinct communication channels with the DAU: a command channel for sending command and configuration data to the DAU; a data channel to allow the DAU to send PD data records to the System Controller; and an emergency command channel to allow the controller to gain a DAUs attention, while a PD test is in process. An Ethernet communication protocol is used to ensure future compatibility to proposed IEEE standards for generator station monitoring system communication [8]. The DAU software is a DOS based application. The software accepts commands from the System Controller via the RPC protocol described above. Once a configuration has been loaded into a DAU, it will run autonomously monitoring the machine sensors for appropriate trigger conditions and only communicating again with the System Controller when PD data has been collected. This reduces network traffic and allows one System Controller to service many DAUs VI. HYDROGENERATOR EXPERIENCE The operating conditions and PD data can be displayed on any network computer. The display module has a wide variety of possible output graphs including: conventional pulse height analysis plots with families of curves based on sensor identifier, time of test, pulse polarity, machine load, etc. conventional 3 dimensional pulse phase analysis plots of the number of pulses per second versus magnitude and phase position of the 50/60 Hz voltage pulse phase analysis graphs which plot the pulse magnitude versus the phase position (either linear or polar mode) on a 2 dimensional graph, with the pulse repetition rate encoded as the color of a data point (Figure 2) trend plots of PD activity indicators (such as the peak PD magnitude or the total discharge activity, NQN) versus time, for a user-defined set of machine operating conditions (Figure 3) trend plots of PD and generator operating conditions versus time

i -rerature compensated reference voltages are used to detect any drift in the A/D electronics. V. SOFTWARE DESIGN The System Controller has most of the critical software to run the monitoring system. The software modules are written in the C++ programming language and are 16 bit Windows applications. Because each successful PD test of 6 coupler pairs results in over 800kB of data, the PD test result data is compressed and stored in binary large object fields of a Paradox database. The controller consists of several distinct software modules: a supervisory program to configure and accept data from me DAUs a module to control the sensitivity of the PD measurements in each DAU, as well as the time to scan the PD sensors. a module to allow the user to input the particular operating conditions of the motor or generator which cause each individual DAU under what operating condition ranges

 Since a tremendous amount of data can be acquired in theory, it is critical for powerful display techniques to be available for the user, to reduce the data to meaningful results. A variety of filters, based on date and machine operating conditions have been developed. Multiple operating conditions can be ANDed to interactively provide selectable data access patterns to extract meaningful results. About 30 DAUs have been installed to date in utilities around the world, mostly on hydrogenerators. VII. CONCLUSION A practical, automatic, continuous, PD monitoring system has been developed to measure, store, and display the PD activity in generator siator windings. Since the first installation on a generator more than two years ago, the system has been proven to be robust and reliable. From the data collected thus far, several interesting features of PD, as a function of machine operating conditions, have been found. It is expected that as more data is collected on more machines, and correlated with machine condition, as determined by visual inspection, other relationships may be established. VIII. BIOGRAPHY Dr. Greg Stone obtained his degrees in Electrical Engineering from the University of Waterloo. Canada. From 1975 to 1990 he worked primarily at the research Division of Ontario Hydro where he contributed to the development of on-line partial discharge tests. He joined Iris Power Engineering in 1990, and is currently Vice President of Global Education Services. He has won several IEEE awards for the development of on-line PD tests, has chaired a number of IEEE standards working groups, represents Canada on EEC 2J and 15E, and is a past President of the IEEE Dielectrics and Electrical Insulation Society. He is a Fellow of the IEEE and a registered professional engineer in Ontario, Canada. Mr. Blake Lloyd graduated in Electrical Engineering from the University of Waterloo, Canada. From 1985 to 1991, he was employed in the Electrical Research Division of Ontario Hydro where he was responsible for developing software for advanced measurement, testing, anil diagnostic monitoring equipment for rotating machines. He joined Iris Power Engineering in 1991, and is the Vice President of Research and Development. His responsibilities include new product development and design, and he has been an architect of several on-line monitoring products. He has been an TREE member since 1985. Mr. Steve Campbell graduated from Sheridan College, Toronto in 1981, specialising as an Electronics Technologist. After graduation Mr. Camptx11 worked with Ontario Hydro Research Division, where he was responsible for design

and construction of high speed trigger circuits, system design of computer-controlled, high speed digitizers and measurements of high speed, high voltage pulses in large motors and generators. He is presently Vice President of Strategic Planning for Iris Power Engineering.

IX. REFERENCES [1] J. Johnson and M. Warren, Detection of slot discharges in high voltage stator windings during operation, Trans. AIEE, pt. II, p. 1993,1951. [2] M. Kurtz, J.F. Lyies, and G.C. Stone, Application of partial discharge testing to hydrogenerator maintenance, IEEE Trans. PAS, vol. PAS-103, pp. 2148-57,1984. [3] S.R. Campbell et al, Practical on-line partial discharge tests for turbine generators and motors, IEEE Trans. EC, p. 281, June 1994. [4] J.E. Timperley and E.K. Chambers, Locating defects in large rotating machines and associated systems through EMI diagnosis, CIGRE, no. 11-311.Sept.l992. [5] B.A. Fruth and V. Green, Benefits of partial discharge monitoring. EPRI Motor and Generator Conference, Orlando, Nov. 1995. [6] R.A. Halpern and D.J. Stanton, Long term reduction of liquid cooled generator operation and maintenance costs, EPRI Motor and Generator Conference, Orlando, Nov. 1995. [7] G.S. Klempner, A. Kornfeld, and B. Lloyd, The generator expert monitoring system (GEMS) experience with the GEMS prototype, EPRI Utility Motor and Generator Predictive Maintenance Workshop. December 1991. [8] A. Roehl and B. Lloyd, A developing standard for integrating hydroelectric monitoring systems, EPRI Motor and Generator Conference, Orlando, Nov. 1995. [9] E. Colombo and G. Tontini, On-line PD measurement on the stator winding, IEEE International Symposium on Electrical Insulation, Baltimore, June 1992.