By Sergey Agou, Vladimir Tsvetkov, Gabe Paoletti and Alexander Golubev
In recent years, a number of new methods of on-line insulation assessment have been developed for MV and HV AC equipment. These methods are slowly gaining the recognition of the electrical industry. The attractiveness of these methods is that they can be applied on-line. In other words, they can be used to assess the equipment's condition while it is in operation, without the need for a shut down.
These methods include On-line Partial Discharge Analysis (PDA), On-line Vibro-acoustic Analysis for Rotating Machines (VAA), Dissolved Gas Analysis for Transformers (DGA) and On-line Power Factor Monitoring (PFM), to name a few.
Operational decisions are often based on assessment results, which emphasizes the importance of the accuracy and reliability of the applied methods. In order to reduce the risk of erroneous diagnosis, one should look into applying several methods based on different principles to diagnose the suspected problem.
This article briefly discusses experiences and utilization of such an approach.
Partial Discharge
Partial Discharge (PD) is an electrical discharge that only partially bridges the insulation between conductors. A transient gaseous ionization occurs in an insulation system if the voltage stress exceeds a critical value, and this ionization produces partial discharges1.
Partial Discharge Analysis is a method of detecting, filtering and analyzing partial discharge in insulation systems. This method has quickly evolved in recent years due to the efforts of several companies using affordable and portable digital signal processing equipment.
Partial discharge analysis relies on measurements of very low magnitude electrical discharges occurring inside insulation systems. Analysis of these discharges reveals insulation deterioration processes including severity, type and location. The challenge is to identify these minute discharges in the presence of electrical background noise of a similar nature and of much higher amplitudes. A good example would be a live on-line assessment of the insulation of a transformer operating in a substation switchyard. Modern advanced partial discharge analysis instrumentation [1] utilizing digital signal processing is capable of detecting a 20-30 picocoulomb discharge in a 750 kV transformer [2]. Identifying such a low discharge is a significant achievement -- as a comparison, a new switchgear lineup leaving a factory floor typically has to have a PD level of less than 100 picocoulombs at an operating voltage to pass quality control test.
Partial discharge analysis can be applied to various types of MV and HV equipment, including switchgear, motors, generators, power transformers, substation instrument transformers (PTs and CTs), terminations and splices of cables, iso-phase buses, etc.
Partial Discharge Monitoring as an Arc Preventive Measure in Switchgear
In switchgear, a multi-channel PD analyzer measures both internal and external PD, to reject the external noise component. Partial discharge analysis instrumentation also detects similar phenomena such as surface tracking, corona and arcing. Here are some examples of problems that are detected in switchgear:
- Surface tracking (for example, surface contamination of an isolator shows up as 'treeing')
- Partial discharge between a high-voltage conductor and insulation (for example, in a void between bus and insulation)
- Arcing and sparking (for example, in case of a conductor under floating potential in a defective cable termination)
- Corona (for example, visible surface discharge or discharge from a sharp electrode into air or corona in the air gap between MV bus and the support window that leads to the deterioration of insulation)
If the partial discharge process is not suppressed, it can develop into a full-scale insulation failure event followed by arcing and complete failure. Timely detection of insulation degradation through escalating partial discharge levels prompts corrective actions. Examples of these actions are predictive maintenance or measures to alter operational conditions or environmental conditions to stop the destructive process and prevent failure. If a suspicious level of PD is detected in a system component, such as CT, PT or cable splice, that component is to be inspected, tested off-line and replaced if found defective.
In switchgear, elevated level of partial discharges can indicate a problem in:
- Bus & Supporting Structures
- Potential Transformers
- Current Transformers
- Cable Terminations & Splices
- Points where a grounded component is located close to a high voltage conductor
- Circuit Breaker Insulation structure
The following problems can also be detected:
- Surface contamination or moisture (on bus insulation)
- Conductor under floating potential problem
- Arcing contacts of breaker
- Arcing Primary Stabs
Factors that lead to partial discharges in switchgear and influence the speed of problem development are primarily environmental: humidity, pollution, corrosive environment, presence of conductive or non-conductive airborne particles, etc. This leads to a limited number of predictive maintenance solutions that involves correcting environmental conditions.
Switchgear is a rather simple object to analyze: the insulation system is subject to insignificant levels of mechanical vibration and relative movement; electrical potential and its distribution through the conductor is equal. These factors allow the results of a periodical PD analysis to be very conclusive. This is why Partial Discharge Analysis can be the sole method of detecting a developing problem in switchgear.
Partial Discharge Monitoring in Rotating Machines
In rotating machines (motors and generators), the situation is more complex. Partial discharge can be detected:
- Between the outer surface of the coil and the stator iron (Slot discharges)
- In internal voids of the ground wall insulation (delamination)
- Discharges on the end turns (Corona)
- Discharges adjacent to coil copper (internal deterioration of original copper/insulation bond)
- In the output bushings (generators) or bus supporting structures
Some of these parts are moving, vibrating and experiencing mechanical and electrical load variation. Their aging is influenced by the mode of operation (base load or start/stop), temperature stress, etc. The electrical field distribution in the windings of a rotating machine is complex and has to be taken into consideration when using PD analysis methods. Also, in rotating machines, physical parameters of insulation material, wedging system condition and its age, cooling gas type and other parameters, can significantly influence partial discharge level and its interpretation. The large number of variables complicates the task of partial discharge analysis of medium voltage rotating equipment [2] and calls for the use of supplemental on-line methods to increase the validity of assessment. In addition, most failures of the rotating machine windings are the result of the mechanical forces applied to the electrical winding. These forces are most prevalent in the end-turn and slot area, in contrast to the area of the higher electrical stress that may be at the termination end of the winding.
Detection of Loose Wedging Condition in Rotating Machines
One of the important factors that needs to be diagnosed in rotating machines is the condition of the wedging system. Loose winding, if left unattended, deteriorates and causes a lot of problems that can eventually lead to a phase-to-phase or a phase-to-ground fault. Besides the fault, loose windings often have to be replaced at a significant cost. Detecting the problem in time allows inexpensive corrective actions to be taken, such as timely rewedging, epoxy injection, or 'clean, dip and bake' for motors.
Slot discharge is one of the signs of this problem, making it possible for the problem to be detected with partial discharge analysis technology. However, this type of discharge will only occur if a part of the winding that is vibrating is under electrical potential higher than the PD inception voltage, typically over 2 kV. If the part of the winding that is loose and vibrating has a lower voltage (which is true for 50 per cent of a 4300 V motor winding), the Partial Discharge Analysis method will not detect anything. This can lead to erroneous assessment and operational decisions.
This creates the need to use another method, Vibro-acoustic Analysis for Rotating Machines. This method uses vibration sensors-accelerometers, which are temporarily attached to the motor casing to capture a mechanical vibration signal [3]. The vibration signature is processed by specialized computer software to identify signs of shock type vibration (attribute of a periodical shock-type process).
Loosening of stator core clamping pressure or of a winding fastening are accompanied by shock vibration processes that can be detected by processing the vibration signal collected off a motor stator or rigidly connected elements (like motor casing). The repetition frequency of shock vibration pulses (if they are present in the vibration signal) is equal to the electromagnetic force frequency of 120 Hz (this is due to the fact that a vibrating winding bar will produce this mechanical vibration signal by interacting with the bottom of a stator slot with a period equal to the doubled power frequency). If shock-vibration pulses are present, the high-frequency part of the signal spectrum (kHz range) will contain a number of peaks with frequencies multiple of the electromagnetic force frequency of 120 Hz.
Besides spectrum analysis, it is beneficial to use a cepstrum (a spectrum of a spectrum) analysis. The presence of a cepstrum function maximum at the 120 Hz frequency indicates the presence of shock vibration pulses with the repetition frequency of 120 Hz.
Vibration is measured using accelerometers; each one is attached to a magnetic base to stick to the surface for better contact. Measurements are performed in 6 points of a rotating machine casing. Measuring points are located symmetrically along the perimeter of the frame, close to the horizontal plane of symmetry.
If a shock-type process is detected, it is possible to distinguish between the problems in the core and in the windings. For this purpose, a set of two measurements is required: at a full load mode and at a reduced load. Best results could be achieved when the stator current in the 'reduced load' measurement mode is 50 per cent or less.
The cepstrum analysis yields a set of relative coefficients representing the condition of the stator for each point. The results are displayed in numerical and symbolic format with colors assigned to parts of core and winding. If these coefficients are in the range of 0.85 - 1.0, the condition of the core is 'normal'. If the shock vibration processes are absent ,the corresponding part of the winding or the core will show in dark green.
If the relative coefficients are in the range of 0.6 - 0.85, the condition of the core is classified as 'moderate'. The shock vibration processes are present, but their level is relatively low. A bright green color is assigned to that part of the core.
If the relative coefficients are in the range of 0.45 - 0.6, a 'warning' condition is detected. The shock vibration process levels are significant. Yellow color is assigned to these parts of the stator.
If the relative coefficients are less than 0.45, the condition of the unit is critical. The shock vibration processes are dangerous and the failure of motor could follow. Red color is assigned.
In cases when partial discharge analysis method indicates that the windings are loose, the Vibro-acoustic method confirms the diagnosis. Fig. 1 shows an integrated partial discharge parameter for 8 motors tested at one facility. The higher the parameter value, the higher is the discharge. The discharge signature has indicated a slot type discharge, which is typical for loose winding condition. Fig. 2 shows Vibro-acoustic distribution for the same group of motors. The correlation is obvious. In this case, two analysis methods that use different approaches yield similar conclusions. Re-confirmed diagnosis leads to an operational decision with a high level of confidence.
Dissolved Gas Analysis, Partial Discharge Analysis and Vibro-acoustic Analysis in Transformers
A single on-line assessment method frequently cannot provide conclusive data about the assessed unit. Let's look at the Dissolved Gas Analysis (DGA). This method detects the consequences of a partial discharge or of a heating process occurring in a transformer by analyzing chemical composition of produced gases. In a real-life scenario, when a maintenance engineer gets 'bad' DGA results, he starts sending oil samples to different labs for analysis confirmation. In many cases, a discrepancy between the diagnoses from different labs prompts sending oil to more labs for analysis. Then, in some cases, an expert consultant is also called before the final decision is made. We can conclude that although the DGA method was used for many years and, in many cases, proves to deliver valid results, it is still not reliable enough to draw direct conclusions and provide an indisputable basis for operational decisions, especially if the recommendations involves a large investment (rewind) or extended outage. In some cases, 'in spite of periodic off-line electrical tests, and on-line gas-in-oil analysis' transformers still fail without warning [5].
Here, again, the solution is to use more than one method. For transformers, the on-line Partial Discharge Analysis test is a great tool to resolve uncertainties of the DGA method. PDA has been used on many occasions to get to the root cause problem in a transformer where suspicions were aroused by DGA results.
A typical question in cases when a combustible gas is detected is 'Are the partial discharges in the winding insulation or in the core?'. If they are in the winding insulation, and steadily increasing, a very costly decision to take the transformer out of service and to rewind can be made.
What if they are in the core gap? In this case, the problem is far from critical and the cure is not as costly. The problem is that, as per known field experience, the DGA method alone cannot distinguish between the two. PD analysis allows us to verify the fact that the detected gases are a result of partial discharges, and in many cases, this method helps identify the type of discharge and criticality.
For example, in our practice, a customer's decision to rewind a transformer with high gassing was altered when PDA results proved the problem to be in the core. The savings was substantial, since the customer was about to invest in the removal of the unit from service, transportation to a facility for an internal inspection, not to mention the cost of the outage and lost production.
Another parameter of a transformer that requires assessment is looseness of the winding or of core clamping. This looseness can be a result of rough handling of a new transformer during transportation. Looseness also gradually develops during transformer operation/ aging, and some looseness has been attributed to over-drying of the transformer insulation system. The progress of this parameter can be accelerated due to a shock-type electrical load -- periodical (a furnace-type load) or single (short circuit or a lightning strike).
Loose winding clamping can lead to problems with insulation due to vibration and abrasion. If this problem is detected at an early stage, it can be fixed during a scheduled outage. Proper clamping tightness makes the transformer more impervious to short circuits on the load side.
DGA and PDA methods can detect this problem, but with a significant uncertainty factor. The proper technology to confirm or reject suspicions is the Transformer Vibro-acoustic Analysis (VAA) [2]. The functionality of this method is similar to VAA for motors, although signal processing is different. This technology also uses accelerometers attached to a magnetic base that pick up vibration velocities emanated by windings and core. The accelerometers are temporarily attached to the tank form the outside. Then, the spectrum of vibration signal is processed and a set of diagnostic rules is applied to recognize certain groups of odd and even harmonics based on the amplitude pattern.
The Transformer Vibro-acoustic Analysis method is based on the fact that a compressed magnetic core or the winding represents a mechanical system where the main resonant frequency depends on its rigidity. Under the application of magnetostriction and electrodynamic forces, the core and winding will vibrate creating oscillations of several frequencies, and the highest magnitudes will be observed at the resonant frequencies. Vibration of windings and core is transferred via transformer oil to the tank walls where the signal is measured. The closer the vibrating element to the tank, the higher its contribution. So for better discrimination of the contributing elements, several sensors and several modes of transformer operation can be used. Piezoelectric accelerometers, with the frequency band of 10-1000 Hz, are used as sensors.
Theoretically, the spectrum of vibration for an ideal transformer has to contain primarily three odd harmonics, multiple to electromagnetic force frequency, i.e. to the doubled power frequency (120 Hz for power frequency of 60 Hz). Magnetostriction forces in the core and electrodynamic forces in the windings create the 1st harmonic. The 3rd and the 5th harmonics are created by the saturation of the magnetic core and lead to the symmetrical distortion of vibration signals. In real transformers the vibration of separate turns in a winding and separate laminations in the core superimpose the additional higher frequencies onto the main frequency.
The analysis of the recorded data and evaluation of the clamping pressure is performed by software. The software converts recorded signals into spectra using the Fast Fourier Transformation, detects significant harmonics and calculates the rms value of vibration velocity. The vibration parameters for each sensor location are processed in three 'families' of frequencies: odd (1st , 3rd , ..., 7th ), even (2nd , 4th , 6th ) and fractional (1/2, 3/2,..., 9/2). The following coefficients are calculated (in relative units) for each sensor location:
- the coefficient of winding compression;
- the coefficient of core compression;
- the coefficient of compression along leakage flux paths;
- the generalized compression coefficient.
These coefficients represent the residual clamping pressure relatively to the initially specified one. This technology also uses color coding to present analysis results.
Detecting Problems in High-Voltage Transformer Bushings
Uncertainty is a dilemma for most offline tests as well. Let's take the well-known Doble Power Factor test for the bushings, as an example. It has been recognized, that 'even after millions of tests, we can't predict the rate of degradation'[3] of the bushing insulation.
This means that an operational decision is still made with out solid results and involves taking some kind of a risk. Several vendors are now offering systems for on-line power factor monitoring of the bushings. These systems are more accurate, of course, but can their use guarantee the complete safety of the bushing?
There are several insulation destructive processes that can lead to bushing failure. One of them is related to bushing insulation degradation and can be captured through measuring power factor, preferably on-line or, sometimes, off-line. Another is partial discharge through oil sediment in the oil-filled bushings or through the migrated ink. Although this problem cannot be detected with off-line or on-line monitoring of the bushing power factor, it can be identified with PDA method.
Conclusion
Various on-line assessment methods (Partial Discharge Analysis, Vibro-acoustic Analysis for Rotating Machines and for Transformers, Power Factor Monitoring, Dissolved Gas Analysis) rely on different techniques. Integrated application of these methods increases the reliability of diagnosis and can serve a solid foundation for operational decisions.
Reference:
[1] Z. Berler, A. Golubev, A. Romashkov and I. Blokhintsev, 'A new method of partial discharge measurements', IEEE 1998 Conference on Electrical Insulation and Dielectric Phenomena, Atlanta, GA, USA, October 1998.
[2] A. Golubev, A. Romashkov, V. Tsvetkov, V. Sokolov, V. Majakov, O. H. Capezio, C. Rojas, V. Rusov 'On-line Vibro-Acoustic Alternative To The Frequency Response Analysis And On-line Partial Discharge Measurements On Large Power Transformers', TechCon'99 Annual Conference of TJ/H2b, February 18-19, 1999, New Orleans, LA
[3] Vladimir A. Tsvetkov 'Two Methods of On-line Stator Winding and Core Diagnostics for Large Electric Machines.' IRIS 1999 Conference
[4] Paoletti, G, Golubev, A, 'Partial Discharge Theory and Technologies related to MV Electrical Equipment.' IEEE- IAS Annual Meeting’99 -- To be published in the Nov/Dec 2000 issue of the IEEE Transactions on Industry Applications
[5]. W. McDermid, A. Glodjo, J. C. Bromley 'Analysis of Winding Failures in HVDC Converter Transformers' -- EIC/EMCW'99 -- Cincinnati, OH, October 26-28, 1999
Sergey Agou, Vladimir Tsvetkov, Gabe Paoletti and Alexander Golubev, are with Cutler-Hammer Engineering Services and Systems.