Advantages of Continuous Monitoring of Partial Discharges in Rotating Equipment and

Claude Kane Stephen V. Carney Chemicals & Energy Division Cutler Hammer Predictive Diagnostics Manager Electrical Maintenance 5421 Feltl Road, Suite 190 US Steel Group Minnetonka, MN 55343 Clairton Works - MS #42 Phone: 952-912-1358 400 State Street Fax: 952-912-1355 Clairton, PA. 15025 Email: [email protected] Phone:[412]-233-1409 Fax: [412]-233-1356 Igor Blokhintsev [email protected] Cutler Hammer Predictive Diagnostics 5421 Feltl Road, Suite 190 John Pozonsky Minnetonka, MN 55343 Cutler Hammer Engineering Service & Systems Phone: 952-912-1334 130 Commonwealth Dr Fax: 952-912-1355 Warrendale, PA 15086 Email: [email protected] Phone: 724-779-5815 Fax: 724-779-5828 Email: [email protected]

Abstract -- For over fifty years companies have been Advances in technology are allowing new approaches to performing partial discharge measurements on electric maintenance. These include reliability-centered equipment in order to obtain information as to the quality of maintenance, predictive maintenance, condition monitoring insulation on operating equipment. Partial discharge activity and expert systems. Trend setting organizations are is a well-accepted indicator of insulation deterioration. Consequently, trending of partial discharge activity provides increasingly taking advantage of the convergence of these equipment owners the opportunity to reduce forced outages new technologies to implement proactive maintenance and increase productivity of plant equipment. Traditionally, programs to improve their company's bottom line. periodic testing is performed, but advances in technology allow One such condition based maintenance technology is the the continuous monitoring of partial discharge activity. measurement, monitoring and analysis of an electrical phenomenon know as Partial Discharge (PD), a well-known Index Terms -- Partial discharge, motors, generators, and accepted precursor or indicator to the failure of switchgear, continuous monitor insulation systems in higher voltage electrical equipment. Although it has been successfully applied to equipment I. INTRODUCTION rated as low as 2,300 volts, customer value is low. Value is created at voltage levels of 4,000 volts and above. Last year US industry spent over $600 billion on plant and equipment maintenance. According to industry experts, at II. TRADITIONAL APPROACH OF PD least $200 billion was wasted. Other industry statistics MEASUREMENTS suggest that 80% of all failures of plant and equipment Traditionally, PD measurements have been performed on occur on a random basis and only 20% of the failures are equipment on a periodic basis, one to four times per year. age related. This means that 80% of failures have not been The periodicity is based on the analysis of the most current predicted with current test and maintenance practices and reading, trend analysis, availability of resources, etc. therefore have not been prevented. Can the situation be Typical monitored equipment includes motors, generators, improved? switchgear, cables, bus duct and . In today's competitive environment, increasing demands are Some of the deficiencies in performing periodic PD being placed on the management of physical assets. Many measurements as well as periodic off-line tests such as things have been done to improve the efficiency of insulation resistance and over potential testing are: maintenance work through the use of computerized maintenance management systems (CMMS), but they are 1) Time duration between tests are typically six to twelve more focused on corrective and preventative (periodic) months (if on-line) and frequently much longer for off-line measures. tests. Many problems will manifest themselves in a much shorter period of time. 2) During the performance of a periodic test a problem is organizations that are being driven to do more with less are detected, one starts to ask the following questions: looking for ways to increase reliability with fewer · When did the problem start? resources. The only true way to reduce maintenance costs · How fast is it degrading? Not only the velocity of is to take the "work" out of it, which is the essence of change, but also the acceleration of change. condition-based maintenance. · How fast will it continue to degrade? III. CONTINUOUS MONITORING OF PD ACTIVITY The true answer is - one does not really know! In a periodic mode one will need, at least, two consecutive tests to A continuous monitoring system will overcome all of the establish a rate of defect development which could mean deficiencies listed above and will be an effective condition another year of testing. Therefore, one does not have based maintenance tool: Advantages include: enough qualitative and quantitative data to make a proper judgment at the time of problem detection, so arbitrary 1. Finding a problem quickly. The monitor will identify a decisions are made. problem in its earliest stages of defect development. This will provide one with sufficient information as to 3) In many cases PD activity is unstable. It may not be the growth rate and the severity of the defect. active today, but active tomorrow. Wide variations of PD activity are frequently observed based on data from 2. Providing information as to which phase the defect is in continuous monitoring. Many factors can affect PD activity. and generally what type of defect such as: A few include: · Corona · Voltage · Surface Discharge · Load · Void type of defect (Insulation Delamination) · Temperature · Conductive tape deterioration (Slot Discharge) · Humidity · Loose connection that is arcing · Vibration It is even possible to localize even further the location of the · Pressure defect. As more and more data is collected in conjunction with other dynamic factors, much is being learned. Correlation 3. Since no labor is required to perform the tests, of the certain dynamics with PD activity provides additional continuous monitoring allows the use of limited insight and can improve diagnostics. resources to finding solutions to problems instead of finding problems. Most accepted PD standards focus on off-line partial discharge tests. OEM’s test labs commonly have their own 4. Reducing unnecessary maintenance because the test specifications and acceptance levels vary from OEM to monitor will be constantly testing and will have OEM and are specific for specific insulation and equipment. accurate data on which to base decisions. Most PD standards dealing with on-line PD measurements avoid the stating of specific levels or quantities to determine 5. Collecting more accurate data as tests are conducted acceptability limits. Instead, trending is stressed. under real operating condition

In order to perform proper trend analysis, one should make 6. Requiring no outage to perform the test, therefore there sure all dynamics are nearly identical at each testing is no loss of asset productivity. interval. This is very difficult, and sometimes impossible, to accomplish and can be time consuming. At this time, 7. No introduction of infant mortality failure patterns via knowledge on correction factors to account for the change more invasive testing procedures. in dynamics is more academic than practical. Another reason for the avoidance of establishing levels is that even a 8. Reduction of forced outages and increased safety of very low level of PD activity that is trending upward will be personnel. One will always be aware of conditions indicating the development of a defect. Conversely, a high and/or problems. level of PD activity that is stable should be of concern, but at least one knows that the problem or defect is not getting 9. Correlation of other dynamics such as temperature, worse, which should provide some level of comfort. humidity and load current to PD activity, which provides additional insight for diagnostics. There will 4) Time consuming. Taking periodic readings is labor be no need to go to several sources and gather the intensive and frequently requires an expert to take and information. analyze the data to extract valid information. Maintenance 10. Provides the opportunity for remote diagnostics. The expert does not need to come out to the field for basic diagnostics. A site visit by an expert will be the exception and not the rule. This can be done my emailing data to an expert or perhaps have a modem connected to the monitor so the expert can dial-in and upload the information for analysis and even provide a special test if necessary.

11. Evaluation of a piece of equipment is based on its own history and not by comparison to other equipment. This will make the detection of subtle problems easier.

12. Easily monitor worsening conditions so one can defer repairs and allow time to plan an outage. Figure 1. 13.,8 kV switchgear, PD, different magnitude scales IV. PD DYNAMIC EXAMPLES Figure 2a shows partial discharge intensity (PDI) and As it was stated above, dynamics will significantly affect temperature correlation for cubicle 2. One can observe partial discharge activity. This complicates establishing PD there is a negative correlation between these two quantities. trend and some times makes trending based on periodical Figure 2b shows the correlation of PDI to humidity for measurements (one – four times a year) almost impossible. cubicle 14. (Quantities are normalized to their own The following examples show frequent PD behavior that maximums.) Lack of information on dynamics in both cases one can find in the field. would cause wrong conclusions to be made on PD trend. In the second case, trend is also complicated by the instability Data was collected utilizing a continuous partial discharge of PD. monitor developed for MV equipment. The monitor is an intelligent stand alone device and has the following features:

· 15 PD signal and 1 noise channel · 75 dB of signal dynamic range · frequency band of 1 –20 MHz · RS-485, 4-20mAmp interfaces and two alarm contacts · up to two years of on board data storage · measurement is triggered by event, schedule, time Figure 2a. 13.8 kV switchgear, mainly negative temperature correlation interval or operator request · Recording of dynamics such as temperature, load current and humidity

The monitor acquires PD data in the form of phase-resolved pulse height PD distribution (PRPHD).

Case 1. Trend. 13.8 kV Switchgear Line-up.

The line-up has 15 breakers equipped with Radio Frequency Voltage Sensors (RFVS) type sensors. Several cubicles are Figure 2b. 13.8 kV switchgear, mainly humidity correlation exhibiting partial discharges. Figure 1 illustrates a flat projection of PRPHD on the magnitude and phase plane for Figure 2b shows significant unstable PD activity starting in all breakers at a selected time. A pulse is represented on the May and the PD activity shows a correlation with humidity. plot by a dot. Pulse repetition rate is color-coded. If periodic PD tests were performed it would be a hit or miss proposition due to the wide fluctuations. Additional on-line field tests were performed and it was determined the cause of the PD activity was a voltage (VT). In late August, the VT was replaced and as can be observed, the PD activity is no longer present. V. NOTES ON MEASUREMENT OF PD IN ROTATING transducer does not affect the measurement of the EQUIPMENT connected temperature monitoring equipment and only passes the high frequency PD signals to the PD instrument. As stated earlier, in order to make a good decision as to the Over 500 machines, primarily motors, have been tested over state of a piece of equipment, one needs as much valid data the past five years with impressive results. RTDs were used or information as possible. This means having multiple for both the initial survey/evaluation and for on-going sensing points, which can easily be done in switchgear since periodic measurements and data trending. RTDs are sensors can be placed in every cubicle, if necessary. currently very effective in trending of machine PD activity In the case of rotating equipment, sensors have traditionally when used with an analyzer or continuous monitor that can been placed at the line terminal of the machine. Due to effectively reject noise and process PD data. signal attenuation, all sensors have their own "zone of sensitivity". Some of the factors that affect this "zone of As an example: The effects of signal attenuation discussed sensitivity" are the physical design of the machine, the type above may cause errors in the evaluation of the stator of sensor used and frequency band of the PD measuring winding insulation condition, if the sensors located at the device. The typical sensors used on the line terminals of a machine line terminals were the only ones used for machine are 80 pF coupling or RFCT on surge assessment. Figure 3 shows PD test results of three 13.8kV ground. If these are the only sensors used, only a motors of a similar design at the same facility. All motors very small portion of the machine is being monitored. On a have permanent radio frequency current transformers large turbine generator, it may be only the ring bus and the (RFCT) sensors placed on the surge capacitor grounding first bar and on a large hydro the ring bus and first two to conductors. Temporary PD sensors were also connected to three coils. As one can observe, only a very small the existing RTDs in each machine. Figure 3 (top) shows percentage of the winding is truly being monitored. The maximum PD magnitudes recorded from the line terminal sensors located at the line terminals will not detect any sensors. Based on these results, one would conclude that developing defects that are deeper in the winding. Motor 1 is in a good state and the Motors 2 and 3 have moderate level of discharges. The data from three RTDs One could argue that the further away from the line that showed the highest reading for each motor are shown terminals, the voltage is lower, and therefore PD is not on Figure 3 (bottom). Conclusions based on this data, are prevalent. Another argument is that the condition of the the same as above for the Motors 2 and 3. The conclusion winding found at the line terminals is indicative as to the is different for the Motor 1. It has a high level of PD at the overall condition of the winding, such as a slot discharge. zone of RTD01 and is the first candidate for additional Let's address these issues. testing and internal inspection. RFCT Sensors One of the major factors that affect PD activity and the destructive capability of PD activity is the voltage. Assume the monitoring a 13.8 kV (~8 kV phase to ground) machine that has 75 coils / phase, therefore there is a 107-volt drop per coil. This means after three coils, the voltage is still at 7,679 volts to ground. PD will still exist. Motor1 Motor2 Motor3 One can assume the overall condition of the machine can be RTD Sensors determined from just a small sample of the overall machine. In some cases this is true, but there also are many localized defects that occur in the machine that will not be detected from the line terminals.

What can be done? Simple. Use detectors that are Motor1 Motor2 Motor3 embedded in the windings of the machine. This could become very expensive and would be difficult and Comparison or PD Magnitudes from Line Terminal Sensors and RTD expensive to add to existing machines. But not all is lost. Sensors Most machines already have RTD detectors embedded into Figure 3 the winding by the manufacturer and these temperature detectors can be used for partial discharge measurement [3]. In most cases, the sensitivity of the RTDs is comparable to Special PD transducers have been developed that connect to that of the line couplers. This is especially true in motors the existing RTD wires and is installed in series at the RTD and small generators. However, in some cases (specifically older large turbine generators, 500 MW and greater) terminal block located on the frame of the machine. The sensitivity can be much lower. But, there is no question the RTDs do provide additional information so a more valid decision can be made as to the condition of the overall insulation integrity of the machine. Over 500 motors and generators have been successfully tested using RTDs.

Use of the existing embedded RTDs as a PD detector is proven technology that will detect PD that is occurring deep in the winding that may not be seen from the line terminals. Figure 4b. 13.8 kV, 8,000 HP Synch motor, PD variations are both Another item that significantly affects signal attenuation is temperature and humidity related the frequency band of the analyzer. The higher the frequency detected, the quicker the signal attenuates. Since Case 3. Trend. Two 13.8 kV identical 36MVA generators. the signal attenuates faster, the "zone of sensitivity" of the sensor/analyzer combination is reduced. The reason for This example examines two identical turbine generators. higher frequency measurement under field conditions is to The generator shown in Figure 5a operates as a peaking reduce noise. Making measurements in the field can be unit. As can be seen, there is positive correlation of PD difficult at times due to the many sources of noise. Sources activity with temperature and load. The generator shown in of noise include, radio stations, control electronics, Figure 5b is base loaded and virtually no correlations are microprocessor based relaying and metering packages, etc. evident. The higher the frequency measured, the less noise needs to be dealt with, but the sensor will have a much smaller "zone of sensitivity."

VI. PD DYNAMIC EXAMPLES (CONTINUATION) Figure 5a. PDI correlates to both temperature and load.

The above discussion was centered around PD measurements on rotating equipment and signal attenuation aspects. Now, let's look at the effects and correlation of PD activity and specific dynamic parameters.

Case 2. Trend. Two 13.8 kV motors of the same manufacturer.

This case examines two 13.8 kV synchronous motors. Both motors are outdoors and drive compressors. Figure 4a shows that this motor has a slight temperature correlation Figure 5b. PDI is stable. Correlation to both temperature and load is not evident. with PD activity. Most of the variations are load dependent. Figure 4b shows this motor has correlation of VIII. WHAT CAN BE DONE ON CABLES? PD activity to both temperature and humidity. On-line monitoring or testing of cables is an unique application. The effectiveness of performing PD tests on in service cables presents unique situations. Two types of sensors are typically used. On solid extruded cables such as XLPE, HMWPE and EPR, typically the drain shield of the cables is placed though a Radio Frequency Current Transformer (RFCT). This is difficult to do on Paper Insulted Lead (PILC) since the terminations on in a pothead and there are no drain shields. For PILC cables, one can Figure 4a. 13.8kV, 28,000 HP Synch. Motor. PD variations are slightly insulate the pothead from ground and then ground the load and temperature related. pothead with and external ground cable and passing this ground cable through an RFCT. Anther option is to install 80 pF coupling capacitors, phase to ground (one per phase). Experience has shown that in the solid cables, typically locations for PD is found in the terminations and where g ?is the resultant – Gamma (commonly represented as splices. Unfortunately, experience has also taught us the a percent of initial current), I0 is an initial current, C0 is the water trees in these type of cables do not product PD until initially measured value of the bushing insulation right before failure. Consequently, effectiveness of a capacitance C1 , and DC and Dtand are the changes in system to detect water trees in the insulation of a cable capacitance and dissipation factor, respectively, during system is minimal. Also, the length of cable that can be monitoring. monitored is limited by the cable insulation and design. Typically, one can monitor 1,000 - 1,500 feet of XLPE and Fig. 6 illustrates a simple schematic of the measuring HMWPE cables and only about 300 feet of EPR type cable. circuit. EPR cable has shown to have high losses to the high frequency signals produced by a PD event. UA

UB

In PILC cables, PD events can easily occur within the main UC part of the cable insulation structure. Typical effective distances approach those of XLPE cables. C1A1A C1B1B C1c PD Sensors IA IBB IC IX. CONTINIOUS MONITORING ON LARGE D II POWER TRANSFORMERS K S Depending on the data analyzed, bushing failures account from 10% to 35% of all power transformer failures. On-line K - Current Balancing Unit S - Current Summer monitoring of the bushing dielectric characteristics has been successfully utilized for several decades. The original concept is quite simple: The current that can be obtained Figure 6. Simple System Schematic from the test or potential tap of a bushing contains both reactive and active components representing respectively Bushings dielectric properties can vary significantly with bushing capacitance and power loss. Three bushings are temperature, especially if defects exist. Some examples are normally available on transformers or transformer group shown in Figure 7. (including CTs and CCVTs). Assuming that at least one bushing is in a good condition, a comparison between the three currents can be made and diagnostic conclusion be Tan Delta v.s. Temperature derived on this comparison. It would be highly unusually 100 that more than one bushing would deteriorate at the same rate at the same time. Normally, only one of the three bushings would be in poor condition at the same time. 10

The main difference between on-line and off-line 1 measurements is the voltage source. In on-line test one has Tan Delta [%] three separate voltage sources (three-phase system) that can behave somewhat different over a time. This has an effect 0.1 on the measurement accuracy of a single measurement. An 0 20 40 60 80 100 on-line test is continuous in nature and large amounts of Temperature [C] Good Some contamination data are available for statistical processing. In the author’s Severe moisture contamination Semi-conductive sediment North American experience the accuracy of 0.1% is easy to achieve. This accuracy is sufficient for almost any practical Figure 7 application. The tand of good and dry insulation normally has none or a The deterioration of bushing insulation is accompanied with very slight U-shaped temperature dependency. Severe an increase or decrease in its dissipation factor and/or an moisture contamination causes almost an exponential rise of increase in its capacitive current under the operating tand with temperature. If there is semi-conductive sediment voltage. Dissipation factor changes are usually more settling in the bottom of the inner porcelain surface, tand sensitive to initial defects, while capacitance changes tend may show a slight negative temperature effect even causing to indicate a developed defect. The relative value of the negative dissipation factor reading at elevated temperature. current change can be expressed in the form:

DI g = = (D tand )2 + (DC )2 Io Co Consequently temperature dependency is a very important X. SUMMARY diagnostic parameter of bushing insulation. On the other hand, temperature dependency some times can cause This paper reviewed the features and advantages of a problems with interpretation of the data obtained at continuous monitoring system for partial discharge activity different temperatures in off-line tests. Doble provides a on a wide variety of equipment utilizing both traditional PD guideline for this situation. Temperature correction detectors and winding RTD’s. A continuous monitoring coefficients vary with bushing design, year of system can save the user a considerable amount of time and manufacturing and so determining the temperature money and provides more accurate data allowing for more correction factor is a “scientific” problem. informed decision. Some of the advantages of a continuous monitoring system include: When there is a change in g, there is both a magnitude change and a vector change. Monitoring of the magnitude · No labor is required to conduct tests. change provides information as to the rate of change and the · More efficient deployment of personnel. Labor is freed critically of the unbalance. The vector change provides the up to fix problems and finding solutions instead of user information as to which bushing is deteriorating and looking for them. the cause. As shown in Figure 8, there is a step change in · More accurate data as tests are conducted under real o the slope of g as well as a vector at 330 . The vector change operating conditions. · Finding a problem is not left to chance (like interval testing). One will know when a problem started and how rapidly it is progressing, thereby allowing a better outage/service decision. · Unnecessary maintenance is reduced because the monitor is constantly testing providing more accurate data on which to base decisions. In effect, achievement indicates that the power factor of the "C" Phase bushing is of a predictive-based maintenance is reached versus changing. interval/chance maintenance. Figure 8 · Prioritization of which equipment receives service first.

The main obstacle in practical on-line electrical The advantages of using the existing RTDs in rotating measurement of PD on transformers is there is a high level equipment as a PD sensor in conjunction with the more of radio frequency noise in these high voltage/ extra high traditional sensors that are installed at the machine line voltage substations. To date there is no know technology terminals were also discussed. Use of the information that is effective in the electrical measurement of PD at obtained from both types of sensors provides a higher voltages of 69 kV and greater on a continuous basis. There quality of information so better decisions can be made. does exist one method using a specialized analyzer that can be used on a periodic basis. Utilizing of this analyzer, the following defects have been uncovered. REFERENCES

Particularly, the following problems have been identified in [1] ASTM, Standard D 1868-93 (Reapproved 1998),“Standard the transformers already tested: Test Method for Detection and Measurement of Partial · The source of critical PD was detected in a 750 kV Discharge (Corona) Pulses in Evaluation of Insulation Systems.” –ASTM, 1998, West Conshohocken, PA. autotransformer of a nuclear power plant. These PD were caused by a progressing creeping discharge across the 750 kV bushing insulation that was confirmed by [2] IEEE P1434 “Trial-Use Guide to the Measurement of Partial internal inspection. The problem was associated with Discharges in Rotating Machinery, 1998. water penetration through loosed top sealing of the 750 [3] Z. Berler, I. Blokhintsev, A. Golubev, G. Paoletti, A. kV lead. Internal inspection has shown that a really Romashkov “RTD as a Valuable Tool in Partial Discharge catastrophic failure had been prevented. The sketch of Measurements on Rotating Machines”Proceedings of 67th defects found at inspection is presented in Fig. 2 and Annual International Conference of Doble Clients, March 27- picture of a barrier in Fig 3. 31, 2000, Watertown, MA · A source of PD was located in a 500-kV bushing of a 300 MVA autotransformer; [4] IEEE Guide for Partial Discharge Measurement in · Defective bus bar isolators were found on 13.8-kV side Power Switchgear—IEEE Std. 1291-1993 of a large GSU transformer. · Arcing contacts on No Load Tap Changers. [5] IEEE Guide for Partial Discharge Measurements in Liquid Filled Transformers and Shunt Reactors—IEEE C57.113-1991

[6] IEEE Recommended Practice for the Detection of Partial Discharge and the Measurement of Apparent Charge in Dry-type transformers—IEEE Std. C57.124 (R2000)