Bring sense of Peace and Security to your equipment with Eaton Partial Discharge diagnostics capabilities

Transcription

1 Bring sense of Peace and Security to your equipment with Eaton Partial Discharge diagnostics capabilities Monitor insulation condition, predict and prevent insulation problem before they will harm your business Eaton Corporation

2 Partial Discharge diagnostics in Eaton Corp. Continuous Partial Discharge monitoring and Remote diagnostic as invaluable tools in keeping medium voltage equipment running safe and productive. This white paper covers the basics of Partial Discharge and Eaton s approach to Partial Discharge (PD) diagnostics on a wide range of Medium Voltage (MV) apparatus such as motors, generators, substations, small transformers, cables and other critical equipment. Industry strives to extend the life of the main process equipment while decreasing maintenance costs without sacrificing personal or environmental safety. Despite advances in material science, design and manufacturing that made MV equipment more reliable, failures continue to happen, leading to production process interruption, costly repairs and even accidents involving human safety and environment contamination. New technologies are emerging globally to meet this challenge. We will show why Eaton s approach to insulation diagnostics and preventative care is a valuable tool to improving reliability and productivity in many industries, such as Power Generation and Distribution, on and offshore Gas and Oil production, Refineries, Treatment Plants and other manufacturing plants that use critical MV equipment to run safe and profitable businesses. Introduction to Partial Discharge diagnostics Insulation degradation Electrical insulation is subjected to electrical and mechanical stress, elevated temperatures and temperature variations and to other environmental conditions, especially in outdoor applications. In addition to normal operating conditions, there are a host of other factors that may trigger accelerated aging or deterioration of insulation. Switching or lightning surges can cause discharges in areas with already or previously stressed insulation. Mechanical strikes during breaker operation or starts/stops of rotating machines can cause micro cracks and voids. Excessive moisture or chemical contamination of the surface can cause tracking. A defect in design or manufacturing could also contribute to the onset of PD. Both normal and accelerated aging of the insulation produce the same phenomenon Partial Discharge. Partial Discharge PD is a localized electrical discharge that does not completely bridge the electrodes, essentially a small spark inside insulating material or along insulating materials surface. PD is a leading indicator of an insulation problem. It is not, in itself, a unique cause of failure, however quickly accelerating PD activity can result in complete insulation failure. Discharges in air gaps are a typical type of PD. Since air has a lower permittivity than insulating materials, an enhanced electrical field forces the voids to flashover, resulting in PD. Energy dissipated during repetitive PD will carbonize and weaken the insulation. PD happens in voids and cavities filled with air in poorly cast transformers, in epoxy spacers and supports, in air gaps in the main insulation in rotating machines, in gaps between a high voltage (HV) conductor and insulation or between an insulator and ground. These voids can be represented as tiny capacitors that are charged by an applied AC voltage to a breakdown voltage before discharging through the air gap (Figure 1). Voids in insulation HV electrode Like tiny capacitors, charged by AC power to a breakdown voltage Ground electrode Solid Insulation Figure 1. Representing PD in air voids as small capacitors in bulk insulation Partial discharge in air voids is frequently seen leading up to insulation failure in rotating machines. These voids can be the result of poor epoxy impregnation of windings during the VPI process. PD activity can carbonize the insulation in these voids making them conductive. It will then stop sparking in those particular air voids, possibly resulting in a decrease of PD activity during the first Eaton Corporation

3 few years of operation. Then, typically, new voids will emerge due to the thermal and mechanical stresses on the insulation while operating. Delamination of wound insulation creates voids in the depth of bulk insulation and next to the HV conductor. The carbonization process will continue in new voids and eventually will fully degrade the insulation to a critical failure, as shown in the picture (Figure 2). The consequence of not addressing PD activity in this stator was a very costly repair of that stator (replacing both the coils and core). Figure kV generator failure result of not paying attention to growing PD activity in the voids of delaminated insulation PD characteristics can be different depending on where and how the sparking occurs: Contaminants or moisture on the insulation s surface induce electrical tracking or surface PD. Continuous tracking can grow into a complete surface flashover, stopping the whole production line and resulting in a very costly recovery. Figure 3. Tracking along the surface of a 34.5 kv PT provoked by high humidity and airborne chemicals Corona discharges in large air gaps, from sharp edges of a HV conductor for example, produce ozone that aggressively attacks insulation, facilitating a flashover during periods of overvoltage. Figure 4. Corona discharge and ozone attacking ring bus insulation in 13.8 kv steam turbine generator In rotating machines it is common to have substantial PD activity for long periods without significant weakening of the insulation, while in healthy switchgear (SWG) one should expect very little or no PD activity. If PD activity is significant, it will eventually deteriorate insulation to a complete failure. However, compared to rotating machines, maintenance and repair on switchgear is possible without a full outage, in many cases. Higher voltage produces higher intensity PD, thus PD detection in equipment with higher voltage (13.8 kv and up) is more critical. PD activity parameters, such as magnitude, pulse rate and long term trend variation are important tools for determining the insulation s condition. Partial Discharge measurements Partial discharges are very short electrical pulses with rise times from a fraction of a nanosecond to several dozen nanoseconds. Measureable phenomena always accompany PD. That is: Electromagnetic pulses Light emissions Ultrasound waves Electro-Chemical reactions Any of above can be used, but electrical PD sensing is most commonly used for PD detection because it allows for the most consistent quantification. PD pulses have a very broad frequency spectrum ranging from DC up to the GHz frequency range. Different PD analyzers are designed to use different frequency bands, depending on the application and noise environment. PD measurements can be roughly divided into several major groups and subgroups: Off-line PD testing, widely used in the design, manufacturing and maintenance of MV and HV equipment in a HV lab, on factories test floors for design and acceptance testing, and at the installation location during periodic testing. Off-line tests are commonly performed with rated line to ground AC voltage applied to one phase while the other two phases are grounded. Off-line tests Eaton Corporation 3

4 are governed by international standards and performed using a relatively low frequency (LF) range of 20 to 800 khz. Using a LF range allows for even large object to be considered as compact because PD pulse attenuation is low in that range. Low and High frequency cutoff filters are commonly adjustable which helps to minimize noise impact on the measurements, but these filters bring inconsistency if measurements are done with different frequency settings. Even with two devices which are both formally compliant with IEC60270 standard [1], are calibrated to that standard and which measure absolute values, test results can be significantly different with different PD test equipment. On-line PD testing is a rather young technology which has emerged because of progress made in the electronics and computing. Temporary sensors Ultrasound and simple electrical PD sensors can be used for online PD survey measurements. These tests are typically less expensive but with very low sensitivity and poor repeatability of measurements. Permanent PD sensors These provide consistent measurements with much higher sensitivity, but are more expensive and require an outage for installation. Periodic measurements These measurements use a portable PD analyzer to connect with either temporary or permanent sensors to perform PD measurements as needed. But PD activity is highly variable and depends heavily on operating conditions such as system voltage variations, temperature, humidity, and load. PD can disappear and reappear randomly based on these parameters. Low readings from a periodic test can result in a false sense of security. Continuous PD monitoring Continuous monitoring is much more certain. Because it is continuous it is able to capture the PD under all conditions, both during times of high PD as well as when there is very low PD. This advantage can be illustrated by choosing different dates on an actual continuous monitoring trend taken in a SWG section with a Partial Discharge source. You can see that walk-in measurements performed at different times can easily lead to incorrect conclusions (Figure 5). Actual test data shows strong step changes of the PD activity as measured by a permanent RFCT sensor placed on a short feeder to a motor [2]. Several motors with high PD activity were connected to the bus and as they were switched ON and OFF, they produced changes in PD that could be interpreted incorrectly by someone performing just periodic testing. This could easily result in wrong maintenance decisions and wasted time and money. By having a continuous trend and knowing the dates and times the step changes occurred, this information can be compared with the switching schedule which would then lead to a definitive conclusion that the PD is coming from those external motors. This would prevent initiating a costly outage for unnecessary maintenance on the SWG. Continuous monitoring also eliminates human error in the measurement process. Maintenance activity and operating conditions of an object are clearly reflected in continuous PD data trend revealing the entire history of the equipment. This facilitates a very practical and effective approach to planned maintenance. Case study #1 PD shows the motor history Motor history can be reflected by the PD activity trend, with slow deterioration of the insulation and step changes due to failures and maintenance efforts (Figure 6) Figure kv motor trend of PD activity on phase C coupling capacitor: 1- original stator; 2- reversing star point; 3 input leads A and C nearly touching and then separated again; 4 stator replacement. As can be seen, multiple maintenance efforts have failed to bring long term improvement [3]. Evidently a new maintenance strategy is needed in order to bring lasting improvement. Case study #2 Intermittent PD activity in 13.8 kv switchgear line PD monitor was detecting random PD activity above alarm levels [2]. Closer analysis of the accumulated data in a zoomed area of the trend revealed an unusual pattern: that PD activity was typically detected only by the 1 PM reading (Figure 7). Figure 5. - Low PD activity conclusion; - High PD activity conclusion; - Strong upward trend conclusion. Eaton Corporation 4

5 Figure 7. PD activity trend with a zoomed area showing periodic spikes around noon. Humidity and temperature sensors were located in an air conditioned substation building and did not reflect outdoor conditions. The multiple, permanent sensor system revealed that the main incoming bus section fed from the outdoor transformer and bus run was the location of the PD activity. Using this information, an outage was planned and during the inspection a bad gasket was found which was allowing water penetration in an outdoor section of the bus duct (Figure 8). Figure 8. Signs of PD activity in bus supporting insulation. It turned out that the midday sun was heating the bus enclosure, vaporizing the accumulated water and creating a high humidity environment very conducive to PD. This was the cause of the lunch time PD phenomenon seen in Figure 7. No PD activity has been detected in the substation since the repair. It would be impossible for periodic PD measurements to provide this level of information which was only available through continuous monitoring. Remote Monitoring Zoomed area from above Continuous PD monitoring solves many problems, but still does not save equipment if the alarms are not addressed and the data not analyzed on a regular basis. The erroneous approach to the PD monitor as a relay, similar to a protection relay, can give a false sense of security if no alarms are seen on the display (or vice versa, too much worry with too many alarms). PD diagnostics is an intricate technology and requires experts to set alarm thresholds, analyze data and make recommendations. This lends itself to the need for remote monitoring (RM). Contemporary communication technology provides a way to connect remotely to equipment regardless of where it might be located. The internet has brought continents together and connections can now be made to any device that is equipped with the necessary communication interface. Old landline phone modems are being replaced with cell modems and network connections which allow inexpensive communications around the world. Eaton has developed a RM system that offers multiple ways to connect to PD monitors located at customers' sites. Cell modems provide two-way communication which allows for full service. Using cell modems, we are able to change setting, perform special measurements remotely, download historical data and edit the memory. This significantly increases the value of the monitoring. A network connection provides one-way communication. This method pushes data from a monitor to our dedicated Eaton Remote Monitoring server through a LAN without sacrificing plant network security. However, because this method only provides one-way communication, when configuration or setting changes need to be made someone onsite must make those changes (Eaton's experts can provide guidance only). The following two case studies illustrate the difference RM can make when it comes to detecting and mitigating PD. No RM case Without RM, alarms can go unnoticed and preventable failures can occur. In this case, a feeder termination failed causing approximately $1 million of lost production in the blink of an eye. The customer did not install the necessary RM hardware to allow Eaton to monitor alarms and they themselves were not watching for alarms shown on the LED display. Alarm threshold Feeder termination failed Figure 8. PD activity exceeded Alarm levels on several occasions but no one noticed. The PD pattern shows distinctive PD activity on phase A before the failure and no PD activity after repair. Eaton Corporation 5

6 Motors, Generators, Switchgear or Busductswith partial discharge monitors (PDM) Remote Monitoring Back Office Alarms Acknowledgement Escalation PDM Cell Modem Secure Web Portal EVPN Secure Database PDM Cell Modem AT&T Provider Before failure After repair Figure 9. Phase resolved PD patterns before and after feeder termination failure/repair. PD activity was reliably detected by the monitor and the feeder and phase are easily identifiable from the data and yet the customer received NO VALUE from this information because alarms were not noticed. With RM case A water treatment plant with high concentrations of corrosive and conductive chemicals in the air installed a PD monitor on a disconnect switch feeding a transformer. PDM PDM Cell Modem Expert Analysis & Recommendations Figure 12. Cell modem based RM system structure. Problems and solutions in PD diagnostics NOISE, FREQUENCY BAND AND ATTENUATION On-line PD monitoring must operate in a noisy industrial environment which makes off-line technology useless because standard frequency range is overwhelmed with different kinds of noise (Figure 13). 1 us/div ms/div ms/div Figure 10. Shows trend of PD activity exceeding alarm level and PD patterns before and after repair. Alarms were detected remotely, data analyzed, and the location and phase of PD identified. The customer was notified and corrective measures were taken quickly. A critical failure was prevented us/div ns/div 2 ms/div RFCT 50 mv/div 1 ms/div Figure 11. Evidence of PD found after RM system received alarms and the owner notified. The Remote Monitoring service includes automated monthly reports and secure access to a customer WEB portal with data and graphic information. Experts are available for consulting as needed. The RM system structure for a cell modem connection is shown in Figure mv/div 2 ms/div Figure khz radio noise on HV transformer; 2 20 khz noise on motor feeder terminations; 3 70 khz noise in switchgear; 4 thyristor switching in excitation system; 5 thyristor pulses are higher than PD pulses; 6 combination of switching pulses and digital noise The selection of a frequency range for signal acquisition is critically important [2]. While noise is the challenge at the low end of the spectrum, the high end of the frequency range imposes limitation on the sensitivity of PD sensors. PD pulses experience attenuation while propagating from the place of origin to the location of a PD sensor. Attenuation is negligible in the low frequencies (below a few hundred khz) which is why standard Eaton Corporation 6

7 off-line tests utilize this frequency range when noise is not present. For higher frequencies the attenuation is significant, and the higher the frequency the steeper the PD pulse attenuation. Frequency band In off-line PD measurement the highly referred standard is IEC (IEEE STD C ) [1]. It covers many Partial Discharge measurement methods. For wide band instruments the standard recommends frequency bands with lower cut-off 30kHz f 1 100kHz, upper cut-off frequency f 2 500kHz and bandwidth 100kHz f 400kHz. Narrow band instruments should have a midband frequency 50kHz f m 1MHz and a bandwidth 9kHz f 30kHz. These two methods are very different and only proper calibration can make them compatible with each other to some extent. PD can also be detected by instruments providing very high bandwidth (VHF) or by frequency selective instruments such as spectrum analyzers. Adequate coupling devices must be used in the different frequency ranges. No recommendations are given in the standard for either measuring methods or bandwidth/frequencies to be used, as these methods and instruments, in general, do not directly quantify the apparent charge of PD pulses. These no recommendation techniques are what people use for on-line PD measurements where noise impact in the low or standard frequencies makes PD measurements impossible. Most popular VHF bands are 1 to 20 MHz and ultra high frequency (UHF) bands above 200 MHz. Calibration is also possible in those frequency ranges, but it is impossible for the equipment to determine the PD s absolute values due to attenuation at those high frequencies [4,5]. PD pulse shape and the shape of the calibration pulse may affect the calibration and the test results to such an extent that attempts to measure absolute values are useless. Performing simultaneous measurements by standard LF analyzers and VHF and UHF equipment can bridge the gap between those measurements, but they will still give different results for different PD pulse shapes and different test arrangements. The large variety of sensors in the higher frequency range does not simplify the situation: different sizes of coupling capacitors with direct connection to High Voltage bars, Radio Frequency Current Transformers (RFCTs) over cable shield grounds, and various types of antennas. This sensor variety makes PD measurements in the higher frequencies relative and not comparable to each other. However, comparisons to other similar equipment in the same voltage class or data trending on the individual sensors are still valid. Attenuation Partial discharge pulses attenuate while propagating from the place of origin to a PD sensor. In general, low frequency attenuation is insignificant and even large objects, like a hydro-generator stator, can be considered as a single point. In the higher frequencies, attenuation can be very significant and propagation can be very complicated, affecting magnitude and pulse shape as well. Signal attenuation creates two important consequences: The entire extended object, like a stator winding or switchgear line-up, can not be assessed reliably for PD with sensors installed in one or only a few locations. With some accuracy, the location of PD can be identified based on the signal magnitude distribution in multiple sensors placed along an object. Even in the 1 20 MHz range, a pulse originating at one end of the slot of a stator may be attenuated up to 10 times traveling just a couple meters to the opposite end of the slot in a medium voltage motor stator. In large steam generators Stator Slot Couplers are widely used by competitors but are only sensitive to a small portion of one stator bar about 5' from the sensor [6]. A very simple example is when a PD signal propagates along a homogeneous switchgear line. You can see in Figure 14 that the PD signal shows significant attenuation (approximately 30-50%) per vertical section which is typically just a three to four foot distance. In the example below, a PD pulse originated in vertical section #2 traveling in both directions attenuating along the line (Figure 14). mv PD signal attenuation along homogeneous SWG line Vertical section # Figure 14. Attenuation of a PD pulse while propagating from point of origin in section #2 to other sections (sensors) in the line. In a typical switchgear line with T connected bus ducts, PTs, CPTs and Tie breakers, the attenuation pattern can be much more complex [2]. In rotating machines, propagation of a PD pulse along the winding is affected by crosstalk between turns and coils and the attenuation patterns are even more complicated. One thing is certain, that the PD pulse at the place of origin has a higher magnitude and better shape Eaton Corporation 7

8 than far from the place of origin. That naturally leads to using multiple PD sensors distributed along an object for simultaneous detection of PD pulses as a solution to attenuation. Multiple sensors, attenuation & sensitivity Using multiple sensors has two advantages. First, the sensor that is closest to the PD site will show higher magnitudes compared to the rest of the sensors, so having multiple sensors can be used to locate a PD source with some accuracy and, possibly, direct corrective measures before a failure occurs. Second, using multiple sensors provides better, more reliable coverage of the whole object under the test. It is impractical to place several dozen PD sensors in a switchgear line or in a rotating machine, but by reasonably placing just sensors you can significantly improve the detection of PD activity. Attenuation in the higher frequency bands raises the question of the validity of calibration in the sense that is defined in the standard for LF bands [4,5,6]. In the standard band, a calibration pulse can be injected anywhere along the object in the system and it will be the same as injecting it directly at the PD analyzer s input, but in the higher frequencies, a sensor will experience a different response depending on whether the signal was injected next to the sensor (maximum magnitude response) or at some distance from the sensor. So, sensor sensitivity should be defined as the sensitivity to a PD signal injected next to the sensor. By injecting a signal at multiple points throughout an object, one can better understand what actual attenuation can be expected and how many sensors are needed to reliably measure PD activity in that object. It is also worth noting that a variety of PD generators used for calibration can produce significantly different responses in a PD measuring system, even if all of them formally comply with the LF standard. Pulse shape (and hence frequency spectrum) and frequency response of the test instruments will affect the resulting sensitivity. In the same way, PD pulses of various shapes will produce different response in a PD analyzer. All of this makes PD measurements in the high frequency (HF) range relative by nature. It is best to pick a technology (frequency band, test equipment, methodology, etc.) and stick with it [10] so that consistency is achieved over the years and data is comparable to each other. Simultaneous calibration of LF and HF PD analyzers using the same calibrator can bridge the gap between the two technologies, but not a very solid bridge. Eaton Partial Discharge Technology InsulGard TM Several years ago Eaton Corporation introduced the InsulGard TM Partial Discharge monitor with 15 multiplexed channels for PD sensors. Thousands of monitors have been manufactured since that time and installed throughout the world. It s simplicity coupled with a reasonable price made this monitor attractive for customers wanting to monitor the condition of Medium Voltage insulation in various applications, such as switchgear, transformers, switches, bus ducts, cable terminations and splices and rotating machines (from 1000HP motors to 700 MVA steam turbine generators). The InsulGard is an intelligent monitoring device with a simple design and internal signal processing which allows for the reliable detection of PD, even in high electrical noise environments. It is not perfect, but much more capable compare to most of competing technologies. Figure 15. Front and back view of the InsulGard monitor. Door mount version shown with the sensor interface board mounted on the back side (typical for switchgear applications). There are multiple ways to communicate to the InsulGard including RS485, USB and Ethernet interfaces, a 4-20mA interface, and dry contact relays for Alarms. The InsulGard can accommodate many different Partial Discharge sensors manufactured both by Eaton and by many other manufactures. Temperature, Humidity and Load Current channels provide a means to record the local operating parameters which provides helpful information to better understand the deterioration processes in the insulation. The internal structure of the monitor is not much different from other metering devices with microprocessors. Eaton Corporation 8

9 Memory Processing Display Keyboard InsulGard Filter & Conditioning Analog Inputs Alarm Warning Status 4-20 ma Interface Communicatio n Board Noise Channel 15 PD Channels Dynamics sensors Temperature Current or Voltage Humidity or Voltage Dry contacts to annonciation 4-20 ma PD Sensors Capacitor USB interface Ethernet interface RS485 interface Landline or Cellular modem RFCT RTD Figure 16. Internal structure of the monitor. One of the most important features of the monitor is its ability to compare the PD magnitudes seen by the sensor under measurement to the signals that appear simultaneously in the rest of the sensors. If that sensor sees a magnitude higher than the rest, it means that the PD source is closest to that sensor. That pulse will be assigned to that sensor only and the lesser magnitudes seen by the other sensors removed from those sensors. This provides valuable information on the location of the PD source and prevents detecting the same PD activity in different sensors. Noise reduction is achieved using the InsulGard by selecting the proper frequency range, utilizing microprocessors to analyze each PD pulse shape and implementing digital filters for PD pattern recognition. In addition, each InsulGard comes with analyzing and reporting software which is useful for both simple reporting and also for in-depth analysis by experts. Technical data PD Measurement: Number of PD Channels 15 Magnitude Dynamic Range 68 db Frequency Bandwidth 1 20 MHz Phase-Resolved Pulse Height Distribution: 21 Magnitude Windows (3.23 db each) 24 Phase Windows (15 each) System Power in Monitored Equipment: 3 20 Hz and Hz (including VFD applications) Synchronization Internal (by AC power) or External Equivalent PD Pulse Repetition Rate <367,300 pulse/second Calculated PD Parameters: Partial Discharge Intensity (PDI) Maximum Pulse Magnitude (Qmax) Pulse Repetition Rate (PPS and PPC) Trends of PDI, Qmax and PPS/PPC Noise Channel allows using a designated sensor for additional noise suppression. Continuous Watch Function watches for repetitive high magnitude pulses between measurements. Auxiliary Inputs: Temperature - uses 100 Ω Platinum RTD. Current or Voltage and Humidity or Voltage Internal Data Memory allows for up to 1000 days of data storage at default measurement schedule Self-test and Self-calibration occurs each time the unit powers up and before each measurement. Interfaces: C-form Dry Type Relays: Device Status signals loss of power and device malfunction. Warning based on preset PDI or Qmax levels Alarm based on preset PDI or Qmax levels and on preset upward trend of PDI or Qmax 4 20 ma isolated interface RS-485 optically-isolated interface USB interface for connection to PC Ethernet port interface Communication Protocol: ModBus RTU and Binary The InsulGard monitor can come in weather proof enclosure for outdoor applications, as a door mount unit typically used for indoor substations or just on a back panel which can be mounted in any cabinet or into an explosion proof enclosure for hazardous environments. Examples of InsulGard installations are shown in Figures 17 a, b and c. Eaton Corporation 9

10 Figure 17a. InsulGard mounted in a weatherproof NEMA 4X enclosure on transformer/disconnect switch combination at a water treatment plant. Figure 17c. InsulGard mounted on a back panel which can be mounted in any enclosure. Typical sensors The InsulGard PD monitor can use any PD sensors with a reasonable sensitivity in the 1 20 MHz frequency range. In many installations, existing sensors from other manufacturers can be used which saves time and money during the installation. Coupling capacitor Integrated Partial Discharge Sensor (IPDS) is a 80 pf capacitor. Eaton manufactures a whole family of IPDSs for different voltage class applications 5 kv, 7 kv, 18 kv, 27 kv and 38 kv. Embedded protection is what differentiates Eaton couplers from other manufactures. Eaton IPDSs can be safely disconnected from the monitor while the equipment is running. The AC voltage on the coax cable from the Eaton coupler will never exceed 15 VAC, while competitor s couplers can have nearly full phase to ground voltage on an ungrounded coax cable. The IPDS is the main type of PD sensor used by the InsulGard because of its high sensitivity and low susceptibility to industrial noise. The IPDS family is presented in Figure 18 (bottom view with holes for brackets and groove for protecting electronics). Figure 17b. InsulGard mounted on the door for an indoor substation application. Figure 18. Family of IPDS sensors for 5 kv, 7 kv, 18 kv, 27 kv and 38 kv applications. Eaton Corporation 10

11 Radio Frequency Current Transformer (RFCT) sensor The RFCT is a solid ferrite core doughnut shaped sensor which is typically placed on the cable shield grounding wire (Figure 19). Figure 19. RFCT sensor with 0.75" opening which comes with a standard 65' coax cable lead. RTD PD sensor In rotating machines, stator windings are commonly equipped with a number of Resistive Thermo Detectors (RTDs). These RTDs can be used as inexpensive PD detectors by utilizing them as antennas throughout the stator. The RTD PD sensor board extracts the high frequency component of PD pulses, leaving intact the DC component which is used for temperature monitoring. The RTD PD board looks and acts like a simple terminal block for six RTD interconnections. RTD wires can pass through the block, as normal, to a temperature metering system while small RFCT sensors extract the PD pulses overlapping the DC signal and send that data to the InsulGard. Additionally, HF filters cancel any noise which may come from the temperature metering system. Six sets of RTD wires are connected to the bottom side of the sensors while the wires connected to the topside of the sensor board continue on to a temperature metering (Figure 20). One RTD PD sensor board has terminations for up to six coax cables carrying PD signal to a PD monitor. temperature, humidity, load current and system voltage, therefore it is important to have simultaneous measurements of those parameters along with the PD measurements. Knowing the dependence of PD parameters on auxiliary parameters allows for better diagnostics, since different types of PD have different dependencies. The InsulGard monitor has inputs that can accommodate these auxiliary sensors. For temperature measurements, it is preferable to have one of the stator winding RTDs brought to the monitor, but this is not frequently allowed because that RTD cannot be used for anything else, so it must be a spare. The second best option is to use an RTD in the cooling air exhaust or use the supplied external temperature sensor placed in a location that closely follows the stator temperature. The last choice is to just put a temperature sensor in the termination box, but that is better than no temperature reading at all. Humidity sensors are typically placed next to the external temperature sensor. External temperature and humidity sensors are shown in Figure 21. Figure 20. RTD PD sensor board. There was a lot of discussion about using such a simple and inexpensive PD detection method [6,7,8,9], but our experience has shown the value of using these sensors distributed throughout the winding to limit attenuation and pick up signals in the winding depth. In large machines there can be dozens of RTDs embedded in the winding all over the stator. If more than six RTDs are available, adding additional RTD PD sensor boards can significantly increase the coverage of the PD detection system. Operating Parameter Auxiliary Sensors PD depends on many parameters, mainly on Figure 21. Temperature and Humidity sensors. Figure 22 shows an example of PD which has a significant dependence on temperature. In this case, it points to a problem with the semi-conductive layer inside slots and/or the grading layer in the endwinding area. Figure 22. PD increases more than 10x based on temperature. A primary or secondary Current Transformer (CT) provides load current information, typically in rotating machines (Figure 23). Eaton Corporation 11

12 Figure 23. Small CT on the secondary wire of a primary CT (left) and Split Core primary CT over the incoming feeder cable (right). InsulGard system applications. Rotating machines application Motors and small generators The typical sensors used for PD detection in motors and generators are the IPDS couplers and the RTD PD sensor board. If there are 12 RTDs available in the unit to be monitored, it is recommended that two RTD PD sensor boards be used for best results. a) Installation utilizing a welded ground bar and the brackets included with the sensors. Humidity sensor is attached to one of the coupler's brackets. Figure 24. Typical PD sensors in a motor or small generator. The RFCT sensor on the cable shield ground is optional. Actual sensor configuration and installation can be significantly different from one installation to the next based on the design of the termination box, the number of RTDs and the space available for the sensors. Knowing the specifics of a given installation before showing up on site can significantly reduce the amount of time, money and effort spent on the installation process. Examples of IPDS installations are presented in Figures 25 a and b. b) IPDS installation between ground and HV buses. Figure 25. IPDS installation options. RTD sensor boards should be installed at the first termination block after the stator winding, which is typically located on the motor frame. That will decrease the attenuation of PD pulses and the level of induced noise. One or two RTD sensor boards should be used, depending on the number of RTDs in the winding. Examples of RTD sensor board installations are presented in Figure 26. Figure 26. Examples of RTD sensor board installations. RTD PD sensor board replaces the standard RTD termination strip. A very important and unique feature of the InsulGard PD system is its ability to work on Variable Frequency Drives (VFDs). In a VFD application, one of the IPDS couplers is used for synchronization. Additional filtering is used to fight heavy pulse noise typically produced by VFDs. Eaton Corporation 12

13 Large generators In large machines, it is difficult to cover the whole stator winding as reliably as is possible in smaller objects, but utilizing multiple sensors (up to 15) significantly improves the coverage compared to only three sensors in line termination area or even compared to six Slot Couplers in the stator winding. A typical set of PD sensors that an InsulGard uses for a larger machine consists of three coupling capacitors in the Iso-Phase bus head next to the generator line terminations and two RTD PD sensor boards which utilize up to 12 RTDs embedded in the stator winding. Figure 27. IPDS-27kV sensor connected to HV bus in iso-phase bus next to the generator terminals. the PD pulse direction of propagation feature that an SSC can provide (from end-winding area or from slot area). In our opinion, it is a small loss to achieve a more reliable and much more inexpensive PD detection system using the already available RTDs. The iso-phase bus is frequently a source of high magnitude sparkling between the metal tip of a support insulator and a HV current carrying conductor. This type of PD sparks to a conductor under floating potential. Compared to measuring PD in the stator, that type of sparking is considered external noise and can mask other, internal sources of PD. Various technologies have attempted to fight this particular type of external PD activity, but not a single one has been 100% successful. Using two sets of coupling capacitors is the most common approach by Eaton s major competitors. They most typically utilize the time of arrival of the PD pulses, trying to distinguish pulses coming from inside a generator verses those pulses coming to the generator from the bus. They have had moderate success, but in several cases with high PD activity on the iso-phase bus it has failed. Eaton uses a different approach based on internal digital filtering that recognizes a PD pattern specific to a HV conductor sparking to a conductor under floating potential. The InsulGard is able to filter that specific spark out, leaving untouched the other sparks which do not meet the filters criteria. The efficiency of this filter is not 100% because it depends on a very specific pattern, but it has proved to be more efficient than time of arrival and less expensive since it requires only one set of coupling capacitors. Figures 29 illustrate how the InsulGard s firmware filters eliminate that particular PD pattern from the PD data. Figure 28. A set of two RTD PD sensor boards installed in a separate enclosure next to the hermetically sealed glands of a hydrogen cooled 500 MVA generator. Using already embedded RTDs for PD sensing is much less expensive compare to Stator Slot Couplers (SSCs). SSCs can only be installed during manufacturing and/or rewinding or during a full outage with the rotor removed. For the InsulGard monitor, there is no difference between an SSC and an RTD sensor because it is able to use both. In fact, simultaneous measurements on SSCs and RTDs in the same locations proved that both sensors have essentially the same PD pulse magnitude and shape, so existing RTDs can fully replace SSCs for a drastically reduced cost [8]. There is one slight disadvantage, though, to using RTDs over SSCs. The InsulGard monitor can't use a) Raw unfiltered PD data with sparking in dozens of support insulators along more than 100 meters of iso-phase bus b) Same data with Firmware filters applied Figure 29. Effect of digital filtering of external PD activity coming to the generator from iso-phase bus Eaton Corporation 13

14 The fact that RTDs are not affected by high magnitude sparking on the iso-bus makes Eaton s IPDS/RTD combination of sensors an especially effective PD detection system. If there are more than 12 RTDs in the winding of a large machine, proper RTDs must be chosen based on the following two criteria: the chosen RTDs should be in the parts of the winding with high electrical stress (not close to neutral areas) and they should be evenly distributed to cover the winding as uniformly as possible. The task of choosing the proper RTDs for PD detection requires pre-engineering before the PD system is installed, based on information obtained from stator winding drawings. Hydro generators Because of the large diameter of the stator in Hydro turbine generators, using two sets of coupling capacitors is recommended. Each set should be located on opposite sides of the ring bus, typically next to the connection of a winding parallels jumpers to the ring bus. The remaining 9 channels of the InsulGard should be connected to wisely chosen RTDs with two RTD PD sensor boards used. Substations The online PD diagnostics that was first applied to rotating machines was successively applied to medium voltage substations and distribution equipment. Eaton is an OEM for medium voltage substations which was very helpful in extending Partial Discharge technology to that equipment. Multiple sensors placed along a switchgear line-up not only provide enough sensitivity to cover the whole SWG line, but are also helpful for localization of a source of PD activity. PD signals attenuate and are detected with a much lower magnitude by sensors distant to the source; however the sensor closest to the source will see the highest magnitude which allows for easy source location. Narrowing down the source to a specific location and phase allows for focused corrective measures without an expensive inspection of the whole substation. Typical sensors installed in a substation are IPDS coupling capacitors and RFCTs. The number of InsulGard monitors, the number and type of PD sensors and their locations depend on a number of factors, including size and rating of the gear, criticality of the circuit (incoming main breakers, ties, critical load, etc.) and available budget. ground is depicted in Figure 30 which allows for reliable detection of PD with reasonable sensitivity. This setup tracks activity along both halves of the substation and also allows for monitoring the feeder terminations and feeder splices all the way to the other end of the feeders depending on distance and cable type. In this case, two InsulGard monitors are necessary to service the substation because 24 channels are required. 138/13.8 kv RFCT RFCT RFCT RFCT RFCT RFCT Figure 30. A sketch of a double-ended substation with IPDSs on the bus and RFCTs on the feeders. An example of an IPDS installation is presented in Figure 31. An example of an RFCT sensor installed over three phase shield grounding braids is shown in Figure 32. Figure 31. A set of IPDSs installed on the load side of a breaker The general rule of thumb is to put one set of IPDS couplers in every third section of a typical SWG lineup. There are two extremes when designing a PD detection system for switchgear, too little and too much. The economical approach is where you minimize the amount of PD equipment to such a degree that it creates blind spots because of attenuation and lack of sensors. The alternative is to go overboard with the sensors, installing them everywhere. This provides great sensitivity and the highest level of protection but can be overwhelming when trying to analyze the data. A good middle Eaton Corporation 14

15 Figure 34. Configuration window - displaying existing setting and allows for changing settings before sending those settings to a monitor. Figure 32. An RFCT sensor installed on 3-phase cable shield grounds. Disconnect switches and small transformers are other popular applications, especially when in contaminated outdoor environments typically found in places such as refineries, chemical plants and water treatment plants. A combination of three IPDS couplers and RFCT sensors on both the incoming and outgoing feeders can provide a perfect tool for diagnosing the insulation condition of such equipment. InsulGard software Monitors come with software package. SW allows communicating to a monitor, downloading data, changing settings and controlling measurements. Number of analyzing tools allows for data presentations, for creating one-touch simple reports and for deep analysis for expert usage as well. Here are few screen shots of the software: Figure 35. Trend of PD Intensity in six coupling capacitors (can choose to graph from 1 to 15 sensors simultaneously). The graph shows significant PD activity, then gap in data because the customer had problems at their network and failed to send data (one way connection) and finally bursts of PD activity that lead to outages and lost production. Customer was notified about severity and probable locations for PD activity way ago, but has taken only cosmetic measures that haven't changed PD activity. Figure 36. Distribution of PD intensity along the sensors for peak PD activity - phase of PD activity and location along the line is evident. Example of more sophisticated analysis - correlations of PD parameters have shown high positive correlation of PD magnitude to Load current in the motor and negative to the temperature - typical for lose winding -lost or loosen wedges, shrunk insulation. Figure 33. Initial window allowing for choosing particular object to work with. Figure 37. Correlation charts showing high correlation of PD values to Load and temperature Eaton Corporation 15

16 Looking at those charts and also using Assist table with useful hints, one should anticipate predominance of PD magnitudes on negative half wave of HV power. Need to look further at Phase Resolved PD Distribution (PRPDD) patterns. Scales - magnitude (mv), Phase angle (deg.) and number of pulse for given magnitude and phase angle bin - color coded from black to yellow. 2D presentation of 3D distributions allows for compact and visual presentations of the data. Polarity predominance definitely can be seen in PD patterns in several sensors. Figure 38. Phase Resolved PD Distribution Different representation of PD pattern - 3D graph and another graph - PD pulse counts vs. magnitude for RTD01 sensor also shows that predominance. Figure 40. 2D graph for Pulse Repetition Rates versus PD Magnitudes. Red curve is for positive PD pulses between 180 and 360 phase degrease of system voltage. So conclusion is "Slot Discharges, re-wedge or rewind the stator during scheduled outage before it fails. One click simple report is available with InsulGard software: InsulGard Status Report Date of Report: 9/1/14 Customer: YYY Plant Location: XXX Plant area: ZZZ Equipment ID: 1B-Motor Basic Analysis Report from 4/7/14 to 7/7/14, 361 measurements, 1 alarms Status: Based on 3 month Q trend, Chan.# 7(RTD04) alarm expected within 3 months Average Partial Discharge Intensity (PDI). This chart represents the average of the PDI as a percentage of the Alarm Point. PDI is the calculated amount of energy in the discharges. The horizontal line represents the average value of the selected time period and the vertical line indicates +/- one standard deviation. Figure 39 3D presentation of PRPDD Figure 41. Title head of the report and one of the graphs presenting averaged PD Intensity in comparison with Alarm Thresholds. Remote Monitoring software Remote Monitoring contract comes with a number of useful options. MyInsulGard.com Web page and automated Monthly reports are available by secure connection to RM server with customer ID and password. Customer portal also give access to more clearly presented information. Visit MyInsulgard.com with ID= CP and password = password, kindly allowed by Ingredeon Inc. for presentation purpose. Eaton Corporation 16

17 Statistical distribution for calculated PD parameter % % Percentage 80.00% 60.00% 40.00% 20.00% 0.00% 75% More PD parameter (Total PDI) Figure 42. Remote Monitoring customers' portal Figure 44. Accumulated percentage distribution for motors >10 kv to illustrate the meaning of the status value. Monthly report gives general information on insulation condition in a monitored object: Comparing Eaton PD technology to others'. With all respect to our colleagues and competitors, Eaton PD technology has features that differentiate it from others and gives Eaton advantage in continuous on-line PD monitoring: 80 pf couplers in line termination area is very popular sensors, however Eaton s IPDS is safer and more reliable: for 15 kv class capacitors we do not have PD up to the maximum test voltage of 36 kv. PD values in certificate are taken at 25 kv at that voltage most competitors couplers caps sink in PD. No need to disconnect IPDS sensors for HiPot testing compare to competitors. Figure 43. Top part of monthly report (courtesy of Ingredeon Inc.). Analysis technology and data presentation are continuously improved. New statistical approach to PD data allowed for better understanding of condition of large pool of monitored motors and generators. That approach gives a powerful tool for asset management to priorities and plan maintenance /repair /replacing activity and save time and money. In statistical approach each motor from Remote Monitoring pool is compared to the rest of the motors of the same voltage class. Monthly Averaged PD Intensity of a motor is compared to standard distribution that calculated on annual basis. Monthly Insulation condition Status shows percentage of the motors from the pool that have PD activity less than the motor in question. For example 75% means that 75% of the motors have lower PD activity, while 25% have higher. Green, Yellow and Red zones (<60%, 60%<=90% and >90%) gives simple answer on what is the motor conditions are compare to a large number of motors and plan maintenance accordingly. Eaton unique technology allowing to work with motors feed by Variable Frequency Drives in wide frequency ranges from very low frequencies of few Hz to 400 Hz and above. Adjustable filters can provide reliable synchronization on background of heavy pulse noise from VFDs. Additionally, RTD sensors inside winding depth are not affected by VFD noise and providing reliable information on insulation condition even in VFD application why coupling capacitors are capable to detect only high PD activity above VFD noise levels. Time of arrival techniques on large generator with huge sparking in iso-phase buses may be less effective than Eaton PD pattern recognition. Eaton s FW filters recognize that specific patterns and discard that sparking as external noise in most cases. Many competitors are using 3 high capacitance coupling capacitors in line termination area and sometimes additional one - on neutral (Rotating Machines). Have to deal with noise issues in online measurements. Most applications are in lab and on test floor in off-line measurements. Can be very adaptive, using multiple frequency ranges, filtering, pulse recognition, special data presentation. Noise rejection and PD signal analysis: since people are using high speed digital acquisition, it opens doors to a various complicated Eaton Corporation 17