Capacitor Failure Analysis (Eaton)


Author: Thomas M. Blooming, P.E., Eaton Electrical Asheville, North Carolina

1.0 Introduction

A steel processing plant was experiencing unexplained capacitor failures and fuse operations in an automatically switched capacitor bank. The plant rolls and galvanizes sheet steel for the automotive industry. Any problem that interferes with production schedules affects the bottom line. With increased productivity demands, the plant cannot afford to devote man-hours to recurring problems. Plant personnel need to solve problems as they occur rather than continue to replace failed equipment or restart shut-down processes.

Maintaining an acceptable power factor is important to the plant because the utility rate structure includes a penalty for low power factor. The accounting department did notice a reduction in the electric bill when the capacitors were added, proving that they are definitely contributing to the bottom line.

Due to the variable loading on one of the plant’s 480 V buses that needed power factor correction, plant engineers chose an automatically switched capacitor bank with four variable steps. When capacitors and fuses in the bank began to fail the electric bill increased and plant processes were affected.

2.0 Electric Power System

2.1. System Description

A simplified one-line diagram showing the portions of the power system relevant to this paper is shown in Figure 1.

Capacitor Failure Analysis  F1

Figure 1. One-Line Diagram of the Power System

The steel processing plant is served at 13.2 kV at the end of a radial overhead distribution line. This line has a relatively low short-circuit MVA for this voltage level. The short circuit MVA at 13.2 kV is 55 MVA with an X/R ratio of 2.99. From the metering, there are four transformers serving various parts of the plant. These transformers range from 1000 to 3000 kVA.

One of the transformers, a 13.2 kV-480Y/277 V, 1500 kVA delta-grounded wye with a 5.6% impedance, serves the 480 V bus where the automatically switched capacitor bank is installed. It is this bank that has been experiencing capacitor failures and fuse operations.

The capacitor bank contains two 50 kvar fixed steps and four switched steps of 50 kvar each for a total of 300 kvar. The capacitors which make up each of the steps are 16.67 kvar, three-phase cells. All of the kvar ratings are at 480 V. None of the steps are configured as harmonic filters. Each 50 kvar group is protected by its own set of current limiting fuses. The steps switch in and out of service automatically based on the power factor correction control algorithm in the bank.

The variable steps in the bank are switched by means of electro-mechanical contactors. The control algorithm switches steps in and out in order to maintain a target power factor. There is a time delay when switching, either adding or removing capacitors, to avoid hunting, the excessive switching in and out of a step.

The control algorithm also avoids switching in a step within one minute after it has been disconnected. This allows trapped charge to dissipate to less than 50 V before reconnecting them. This is done so that capacitors are not in power factor penalties. Release of system switched in when they have a trapped charge, capacity was not an issue on this particular which might lead to an excessive switching service. A multi-step, automatically-switched bank transient. was chosen because of the intermittent nature of

These two limitations allow short periods of time when the power factor criterion is not met. On balance, however, the overall power factor from a demand point of view is maintained above the set level.

The load on this 480 V bus includes four DC drives, served from two isolation transformers (two drives per transformer). These drives operate intermittently as the process demands. The average load on the main 1500 kVA transformer was 550 A with a maximum of 990 A during the measurements. The drives are the only significant harmonic sources on the bus. When the drives draw their maximum current, they can comprise about 40% of the bus load. This does not happen very often, however.

With the power factor correction capacitors, The steel processing plant benefits from voltage support in addition to cost savings due to reduction many of the loads on this particular bus.

2.2 Description of the Problem

The steel processing plant was experiencing problems with the automatically switched capacitor bank for some time before they investigated the problem. The problem was not discovered immediately because the bank is not checked regularly. The problem was first noticed in the electric bill. Permanent on-site monitoring may have detected the problem sooner.

The natural first action was to simply replace the blown fuses that were found. It was later noticed that some capacitor cells had also failed. These were also replaced. When the problems persisted a detailed examination was undertaken.

At the time of the measurements, some fuses were blown and some capacitor cells had failed. The fuses in variable Steps 1 and 4 were blown and one of the three 16.7 kvar (three-phase) cells in Step 3 had failed so Step 3 was supplying only 33.3 kvar rather than its nominal 50 kvar.

No obvious cause was observed during the measurements that were performed. Either the problem was due to cumulative effects over time or it was an intermittent problem that did not occur during the measurements.

The fact that failures did not occur during the measurements made further analysis necessary to determine the cause of the problem. If failures had occurred during the measurements, the measurement data at the time of the failures could have been analyzed and the cause may have been determined much sooner.

3.0 Power System Measurements

3.1 Harmonic Measurement Results

Possible causes for the capacitor failures and fuse operations included excessive harmonics and transients (overvoltages). Measurements were performed to quantify the harmonic voltages and currents in the capacitors in order to study whether harmonics were the cause of the failures. The power monitor used for these measurements would also catch transients if they were to occur. Measurements were also performed on other parts of the power system, including the DC drives which are known to cause harmonics, as part of a larger study effort.

The average values for the voltage total harmonic distortion (THD) and the rms, fundamental, and harmonic voltages at the capacitor bank with different kvar step configurations are presented in Table 1. All configurations also include the 100 kvar fixed step. All values given are three-phase averages. All harmonics are given in percent of fundamental.

Higher order even harmonics such as the 8th, 10th, 12th, 14th, et cetera are not normally reported, but they were in this case. This was done to investigate a possible harmonic resonance condition near those frequencies.

The average values for the current THD and the rms, fundamental, and harmonic currents flowing in the capacitor bank with different kvar step configurations are presented in Table 2. All configurations also include the 100 kvar fixed step. All values given are three-phase averages. All harmonics are given in percent of fundamental.

The average values for the current THD and the rms, fundamental, and harmonic currents flowing in several other important locations are presented in Table 3. All values given are three­phase averages. All harmonics are given in percent of fundamental. For the DC drives, all data is presented during periods of significant load. Time when the drives were not operating is not included in the drive data.

Capacitor Failure Analysis T3Table 1. Capacitor Voltage Measurement Summary

Capacitor Failure Analysis T2Table 2. Capacitor Current Measurement Summary

Capacitor Failure Analysis T1Table 3. Load Current Measurement Summary

The measurements show relatively high, but not unusual, levels of harmonics being produced by the pulse width modulated (PWM) drives. By comparison, the harmonics in the capacitor bank and in the transformer have much higher than expected levels of 11th and 13th harmonics relative to the harmonics injected into the system by the drives. This suggests a harmonic resonance condition. This phenomenon is explored further in Section IV, Harmonic Analysis.

3.2 Transient Measurement Results

During the course of the measurements there were only a few significant transients measured, none of which would be expected to cause problems. The highest voltage transient was 1.74 per unit. None of the transients with significantly high voltage lasted for more than 50 μsec.

The only voltage transients that had corresponding increases in current were some capacitor switching transients. Recall that the objective is to find the cause of the fuse operations as well as the capacitor failures. Therefore current is also of interest, not just voltage. One of the transients recorded is shown in Figure 7 and is discussed in Section VI.

4.0 Harmonic Analysis

IEEE Std 519-1992 [2] discusses the possible effects of harmonics on capacitors. Portions of Section 6.5 of this document are presented below:

A major concern arising from the use of capacitors in a power system is the possibility of system resonance. This effect imposes voltages and currents that are considerably higher than would be the case without resonance. The reactance of a capacitor bank decreases with frequency, and the bank, therefore, acts as a sink for higher harmonic currents. This effect increases the heating and dielectric stresses. The result of the increased heating and voltage stress brought about by harmonics is a shortened capacitor life.

Adding capacitors will cause the power system to be tuned to a certain harmonic. This is known as parallel resonance between the capacitors and the source (including the transformer) inductance. A parallel resonance presents a high impedance to injected harmonics at or near the resonant frequency. This should not be confused with series resonance, which is utilized in harmonic filters to present a low impedance to a certain frequency to remove that frequency from the system.

If the parallel resonant frequency is close to injected harmonic frequencies within the plant, voltages and currents at these frequencies will be amplified. This is more likely when the capacitor bank is a switched bank with multiple steps since there are several possible resonant frequencies. Resonance can result in increased harmonic problems and can lead to capacitor failures.

Calculations were performed to estimate the resonant frequencies of the power system with different levels of capacitance on-line. The resonant frequency of a system, at a transformer secondary, can be estimated with the following formula. h is the tuned harmonic of the system, XC is the capacitive impedance of all capacitors connected to the secondary bus of the transformer, and XL is the inductive impedance of the transformer (plus primary source inductive impedance, if available).

Capacitor Failure Analysis  E1

The information for transformer #3 is as follows: 1500 kVA, Z=5.6%, 13.2 kV-480Y/277 V. The short circuit MVA at the 13.2 kV level (primary of the transformer) is 55 MVA with an X/R ratio of 2.99. The resonant frequency calculations yielded the results shown in Table 4.

Harmonic impedance scans are shown in Figure 2. These scans show the impedance at a range of frequencies for three system configurations. The first configuration is without any capacitors or filters connected to the transformer secondary. The second configuration is with 150 kvar on-line, as was often the case during the measurements. The third configurationis with a 150 kvar capacitor bank replaced with a4.7th harmonic filter.Capacitor Failure Analysis T4

Table 4. Resonant Frequency Calculations

The impedance scans are performed without plant loads connected to the system for a worst case analysis. Connected loads tend to damp, and slightly alter, a system’s impedance scans by rounding off (lowering), and possibly slightly moving, the peaks in the plot. The purpose of the impedance scans is to identify possible system resonant frequencies. To allow these frequencies to stand out more clearly, the analysis is performed without connecting the plant loads to the system.

A high impedance at a given frequency means that any harmonic currents injected into the system at that frequency will cause greater voltage distortion than injected currents of the same magnitude at different frequencies. Harmonic resonance problems occur when harmonic currents are injected at frequencies with high impedances.

Figure 3 shows the possible magnification of harmonic frequencies due to the presence of a capacitor bank or a filter bank relative to having neither. The impedances of the system with the capacitor bank and with the filter were divided by the impedance of the system with neither. Again, without the presence of resistive loads to provide damping, this is a worst case analysis.

Capacitor Failure Analysis  F2

Figure 2. Impedance Versus Harmonic Frequency


Capacitor Failure Analysis  F3

Figure 3. Magnification Versus Harmonic Frequency

The presence of the capacitor bank clearly amplifies a range of harmonics. Characteristic harmonics of six-pulse drives include the 5th, 7 , 11 , 13 , 17 , 19 , et cetera, in decreasing amounts. But during the on-site measurements the capacitors and the main 1500 kVA transformer were carrying significantly more 11th and 13th harmonic current than 5th and 7th. This occurred despite much higher 5th and 7th harmonic current injections. This can be explained by the tuning of the system with the capacitor bank on-line. There is clearly some degree of harmonic resonance in this system.

Except for a small range of frequencies (due to the parallel resonance of the filter) the filter would tend to reduce the harmonic impedance relative to the system without any capacitors. The filter was tuned below the lowest characteristic harmonic frequency produced by the six-pulse drives to avoid amplifying any harmonic currents produced by the drives.

Figure 4 shows line-to-line voltage and total current into the capacitor bank with 150 kvar on-line, recorded during the measurements. These waveforms show what current and voltage waveforms can look like in a resonant condition. Note that there are additional frequencies riding on the 60 Hz waveforms, especially the current waveform.

With 150 kvar on-line, the calculations estimate a resonance at approximately the 11.1st harmonic. Frequencies near this harmonic may also be amplified. The current waveform shows a strong 11th and 13th harmonic components superimposed on the 60 Hz. The resonance can be identified in the waveform by counting the number of peaks due to the resonant frequency that occur within one 60 Hz cycle. This is somewhat less clear in this case because there are both 11th and 13th harmonics, but one can count 11 “dominant” peaks in one 60 Hz cycle.

Capacitor Failure Analysis  F4


Capacitor Failure Analysis  F5

Figure 5. Harmonic Current Spectrum from Figure 4

Figure 5 shows the harmonic spectrum calculated for the current waveform in Figure 4. It clearly shows the dominant and 13 harmonics despite the fact that the harmonic­producing load is generating more 5th and 7th harmonic current.

A detailed harmonic analysis studying how a harmonic filter could reduce harmonic levels and designing such a filter was not performed due to subsequent discoveries.

Although harmonics were not found to be the cause of the problems in the capacitor bank, the capacitors were causing a harmonic resonance situation. For this reason, or if harmonics become more of a problem in the future, it was recommended that if power factor correction was needed elsewhere in the plant where there were fewer harmonic-producing loads, it would be a good idea to move this capacitor bank to that area. It should then be replaced by a bank configured as a harmonic filter.

Another possibility, not investigated in this study, would be to “de-tune” the capacitor bank. This would not tune the bank to filter harmonics, but would tune it to avoid causing harmonic resonance. The addition of the de-tuning reactors would also reduce the transient overvoltages during capacitor switching.

5.0 Examination of Failed Equipment

5.1 Introduction

In cases like this, an analysis of the failed equipment often yields valuable clues and this case was no exception. Fuses which had cleared were x-rayed to determine the cause of their operation. This x-ray was sent to the fuse manufacturer for examination. A failed capacitor cell was examined by the manufacturer.

5.2 Capacitor Examination

The capacitor manufacturer found that the dielectric fluid in the failed capacitor was almost black from carbon deposits. Carbon deposits are caused by arcing which burns or breaks down the dielectric material.

The internal discharge (or bleed-off) resistors (required by the National Electric Code [3] to discharge capacitors rated 600 V and lower to 50 V or less within one minute) were found to have burned and disconnected connection tabs. It is not clear whether this was a cause or effect of the failure.

To check the discharge resistors in capacitors which had not failed, several of the good capacitors were disconnected from the system after they had been on-line. The voltages were then monitored to see whether the capacitors discharged properly. In every case, the capacitors discharged properly indicating that the discharge resistors were still connected and doing their job.

Several good capacitors were also removed from service in order to check their capacitance. In all cases the capacitance was very close to the expected value.

The manufacturer suggested two possible causes for the failures: excessive harmonic current draw and overvoltage conditions due to an intermittent connection. Excessive harmonic current could be due to motor drives or a resonant condition. An intermittent connection can leave a trapped charge on the capacitor which can result in more severe switching transients (higher overvoltages) when voltage is re-applied. This is why one should be careful when manually switching capacitor banks. When a step is manually switched off it should be left off for at least one minute for it to discharge to 50 V or less. This is discussed further in Section VII, Capacitor Switching Transients.

5.3 Fuse Background

The capacitor fuses in this case are current limiting fuses. Using current limiting fuses to protect capacitors is common at low voltages but is generally not done with medium or high voltage capacitors (4160 V and higher) due to the cost.

Current limiting fuses can clear in two ways: overload and short circuit, in the words of fuse manufacturers. Consulting power engineers also call these two events overcurrent and impulse energy (I2t).

The National Electric Code [3] defines an overload as follows:

Operation of equipment in excess of normal, full-load rating, or of a conductor in excess of rated ampacity that, when it persists for a sufficient length of time, would cause damage or dangerous overheating. A fault, such as a short circuit or ground fault, is not an overload.

An overload is a current that is typically “between one and six times the normal current level.” [4] A fuse will operate, or clear, if the overload is present for a certain length of time based on its time-current characteristic (TCC). If the overload is very short in duration, fuses are generally designed to ignore it. For example, motor inrush and transformer energization are normal system events which cause high currents for a brief time and should not cause a fuse to operate.

A short circuit is “an overcurrent which exceeds the normal full-load current of a circuit by a factor many times (tens, hundreds, or thousands) greater.” [4] Unlike an overload, a short circuit is often caused by a fault.

The National Electric Code [3] defines a current-limiting overcurrent protective device as follows:

…a device that, when interrupting currents in its current-limiting range, will reduce the current flowing in the faulted circuit to a magnitude substantially less than that obtainable in the same circuit if the device were replaced with a solid conductor having comparable impedance.

Current limiting fuses are designed to “limit peak fault current magnitude and reduce fault time duration for better equipment protection.” [5] They can interrupt a short circuit current in less than one-half cycle, before the current would have reached a natural current zero.

Current limiting fuse characteristics, when the current is high enough for them to operate in a current limiting mode, are described by their I2t values. I2t is a value that is proportional to energy (which would be I2Rt). Since the resistance, R, is constant within the fuse, the performance of the fuse is expressed in terms of the I (current) and t (time) variables. Often I2t is used interchangeably with energy, as will be done in the rest of this paper.

“There are two types of energy values – minimum melt I2t and let-through I2t. Minimum melt I2t is an indication of the amount of energy necessary to melt a fuse’s element. Let-through I2t is an indication of the amount of energy a fuse will let through to a fault before operating and clearing a current.” [5]

The type of fuse used to protect the capacitor bank is a full range current limiting fuse. This means that it has a TCC that allows it to operate for overloads as well as operate in a current limiting mode for high short circuit currents. It has separate elements to perform each of these functions.

Within the fuse there is an “M spot” which is made of an alloy that is designed to melt and clear for overloads but will not operate for short circuits. There are also several “weak spots” or “weak links” that are designed melt and clear for short circuits but not for overloads.

If there is a problem with excessive harmonics causing additional steady-state current, this would be expected to cause the M spot to melt and clear. If there is a problem with short circuits the weak spots would be expected to melt and clear.

5.4 Fuse Examination

As mentioned earlier, fuses which had cleared were x-rayed to determine the cause of their operation. This x-ray was sent to the fuse manufacturer for examination.

Figure 6 shows an x-ray of six of the fuses which cleared. In none of the six fuses did the M spot clear indicating that an overload was not to blame. In all of the six fuses one, two, or three weak spots cleared. If there had been a short circuit or a fault in the capacitor bank, all four weak spots would have cleared.

The engineer with the fuse manufacturer who analyzed the x-rays stated:

Note how the ‘M’ spots on the links are not melted. This suggests that the current was over 500% of the fuse’s rating. Now, not all of the weak spots are opened. This suggests an overload, not a short. Put the two together & you get something in the magnitude of 600%-800%. The harmonics should only add to the heating effects, not be the main concern.

Capacitor Failure Analysis  F6

According to the manufacturer, the 100 A current limiting fuses used to protect the capacitor bank had a minimum melt I2t of 5,000 A2sec and a peak let-through I2t of 11,000 A2sec. This means that for a short circuit that had an I2t of 5,000 A2sec, the weak spots in the fuse would start to melt and clear. All of the weak spots would not be expected to clear, however. For a very high short circuit, all of the weak spots would be expected to clear.

Because, in all the fuses x-rayed, only one to three of the four weak spots cleared, the I2t of the event which caused the fuses to operate was expected to be between 5,000 and 11,000 A2sec.

Based on this information it was now clear that it was transients which were causing the fuses to clear and, most likely, the capacitors to fail. Section VII, Capacitor Switching Transients, examines the cause of the transients and the unique situation that caused unexpectedly severe transients to occur.

6.0 Failure Analysis

6.1 Fuse Analysis

The measurements showed that the rms current in each of the fuses did not approach their 100 A ratings. Recall that each set of 100 A fuses protects a 50 kvar group of capacitors. The full load current of each 50 kvar group is 60 A. The fuse rating is 166% of the nominal full load current. When faster classes of fuses are used, they are often sized even higher.

The fuse rating is selected to allow for capacitor inrush currents (which can be much higher than full load) when each step is switched in. This prevents the fuse from operating during such normal system events.

If harmonics were causing excessive heating in the fuse the M spot should have cleared indicating a steady-state overload. This did not occur. Although the capacitors are sinking a very significant amount of harmonics, the harmonics were not the cause of the fuse operations.

If there were a fault within the capacitor cabinet, the current should be high enough to clear all of the weak spots in the fuse link. The available three-phase short circuit current at the 480 V bus is 21.9 kA and the available line-to-ground short circuit current is 24.6 kA, both considering only source and transformer impedance. Since all of the weak spots did not clear, a fault is not the likely cause of the fuse operations.

The approximate current which caused the fuse to operate was 600-800 A (600-800% of a 100 A fuse) according to the manufacturer. This current could be developed from a transient such as a capacitor energization.

The problem is that the measurement data also did not contain any transient events which would be expected to cause the fuses to operate. In fact, during the measurements there were no failures.

The transient waveform shown in Figure 7 is a capacitor energization when the 50 kvar Step 2 was energized with the base 100 kvar already in service. The steady-state currents before and after the energization were approximately 124 A and 180 A, respectively (60 A per 50 kvar group). The peak current in this event was -1480 A. This was the largest peak current recorded during the measurements.

The I2t associated with the 1480 A peak was 793 A2sec. Including the following positive peak increases the I2t to 1058 A2sec. These are both well below the 5,000 A2sec fuse rating for the weak spots to start to melt.

This type of event is analyzed in greater depth in Section VII of the paper, Capacitor Switching Transients. In Figure 7 it is also worth noting the resonance in the current waveform similar to that in Figure 4.

In summary, the measurement data did not reveal why the fuses had cleared.

6.2 Capacitor Analysis

Capacitors must be built to tolerate voltages and currents in excess of their ratings according to standards. The applicable standard for power capacitors is IEEE Std 18-1992, IEEE Standard for Shunt Power Capacitors. [6] Additional information is given in IEEE Std 1036-1992, IEEE Guide for Application of Shunt Power Capacitors. [7]

IEEE Std 18-1992 gives the following allowable contingency continuous overload limits.

  • 110% of rated rms voltage
  • 120% of rated peak voltage
  • 180% of rated rms current (nominal current based on rated kvar and voltage)
  • 135% of rated reactive power

It should be noted that capacitors are often fused below 180% of rated rms current so the 180% limit is not usually approached.

Short time overload voltages are specified in IEEE Std 18-1992 and IEEE Std 1036-1992 and are given below. These standards state that a capacitor may be expected to see 300 such overvoltages in its service life.

  • 2.20 per unit rms voltage for 0.1 seconds(6 cycles of rms fundamental frequency)
  • 2.00 per unit rms voltage for 0.25 seconds(15 cycles of rms fundamental frequency)
  • 1.70 per unit rms voltage for 1 second
  • 1.40 per unit rms voltage for 15 seconds
  • 1.30 per unit rms voltage for 1 minute
  • 1.25 per unit rms voltage for 30 minutes
  • An older standard, IEEE Std 18-1980 also included the following permissible overvoltages.
  • 3.00 per unit rms voltage for 0.0083seconds (½ cycle of rms fundamental frequency)
  • 2.70 per unit rms voltage for 0.0167seconds (1 cycle of rms fundamental frequency)

None of these tolerances were exceeded during the measurements.

7.0 Capacitor Switching Transients

7.1 Overview

A capacitor switching transient is a normal system event that can occur whenever a capacitor is energized. This transient occurs because of the difference between the system voltage and the voltage on the capacitor. A basic characteristic of capacitors is that the voltage across them cannot change instantaneously. If a capacitor is at zero voltage and system voltage is applied to it, the system voltage will be pulled down to nearly zero momentarily.

There will then be a capacitor inrush current as the capacitor charges. The voltage on the capacitor will then recover and overshoot the system voltage, and then oscillate around the system voltage. It is possible for this overvoltage to reach 2.0 per unit (twice the peak system voltage) if the capacitor is initially uncharged. System damping (resistance) usually keeps this overvoltage below the theoretical peak.

The capacitor voltage will continue to oscillate around the 60 Hz fundamental waveform, with the oscillation gradually getting damped out, usually within a cycle. The magnitude of the transient and its characteristic oscillation frequency will depend on the characteristics of the electric power system in question.

Capacitor Failure Analysis  F7

Figure 7. Measured Capacitor Energization Transient

The magnitude of the transient will vary based on two variables at the time of the switching.

These variables are the initial voltage on the capacitor (trapped charge, usually close to zero if the capacitor has been allowed to discharge) and the instantaneous system voltage at the time of the switching. The greater the difference between these two voltages, the greater the magnitude of the transient. The worst case transient will occur when the system voltage is at peak voltage and there is a trapped charge on the capacitor of peak system voltage at the opposite polarity.

Recall that the National Electric Code requires resistors to discharge capacitors rated 600 V and lower to 50 V or less within one minute. The control algorithm in the capacitor bank avoids switching in a step within one minute after it has been disconnected. So in normal operation there should be very little trapped charge on the capacitors when switching.

If the transient voltage is high enough the capacitor could fail immediately. If not, the cumulative effects of the transient voltages (greater than peak system voltage) may stress the dielectric to the point of failure over time. The transient currents will cause high I2t levels

Capacitor Failure Analysis  F8

Figure 8. Capacitor Energization Transient (Sim.) No Prior Charge on Capacitor, I2t=1,857 A2sec

7.2 Capacitor Energization Simulations

Capacitor energization simulations were performed for two reasons. The event that caused the capacitor failures and fuse operations did not occur during the measurements and the examination of the fuses indicated that transients were the likely cause. The information from the steel processing plant power system was used to simulate some capacitor switching events under different conditions.

Figure 8 shows the energization of a 50 kvar capacitor step with no trapped charge and with no other capacitor steps in service. The energization occurred at peak system voltage. This transienthad an I2t of 1,857 A sec.

Without any charge on the capacitors being switched into the circuit, the I2t values are below 5,000 A2sec, the minimum melt I2t value of the fuses used to protect the capacitors. This is, of course, an expected result. If this were not the case, the fuses would operate regularly for common system events.

Figure 9 shows the energization of a 50 kvar capacitor step with trapped charge and with no other capacitor steps in service. The energization occurred at peak system voltage. This transient had an I t of 5,661 A sec.

Capacitor Failure Analysis  F9

Figure 9. Capacitor Energization Transient (Sim.) Prior Charge on Capacitor (-300 V), I2t=5,661 A2sec


7.3 Back-to-Back Capacitor Switching

Another type of capacitor switching transient is called back-to-back switching. This is when a second capacitor is switched on in close proximity to a previously energized capacitor. In this case a fast transient occurs as the two capacitors share their charge with each other and come to the same voltage. Then there is another transient as the pair of capacitors cause the voltage to oscillate around the 60 Hz fundamental voltage, as described above, as if they were a single capacitor bank.

Figure 10 shows the energization of a 50 kvar capacitor step with trapped charge and with 150 kvar of other capacitor steps in service. The energization occurred at peak system voltage. This transient had an I2t of 5,178 A2sec. The time scale for Figure 10 is greatly zoomed in from that in Figures 8 and 9. This was done to better show the higher frequency initial transient.

Capacitor Failure Analysis  F10

Figure 10. Back-to-Back Capacitor Switching (Sim.) Prior Charge on Capacitor (-350 V), I2t=5,178 A2sec

 7.4 Trapped Charge

In both the simple capacitor energization and the back-to-back switching, when some trapped charge on the capacitors was assumed in the model, I2t values rose above the 5,000 A2sec which would cause the fuses to operate. In both cases, the I2t values did not exceed 11,000 A2sec which would be expected to cause all of the weak spots in the fuses to open. This was true even in the worst case scenario with the system voltage at its peak and a trapped charge on the capacitor of peak system voltage of opposite polarity.

It was known that the fuses operated due to I2t values between 5,000 and 11,000 A2sec based on how many weak spots in the fuses had cleared. The analysis showed that capacitor switching transients, with trapped charge on the capacitors, could cause I2t values in this range. The trapped charge could have occurred in three ways:

7.5 Correlation with Observations

After the various steps in the analysis, it was believed that the failures that were occurring were due to capacitor energization transients, most likely due to switching a bank with trapped charge. This had not yet been confirmed, however.

Plant personnel had reported that the contactor for some of the 50 kvar steps in the capacitor bank had been “chattering” occasionally, opening and closing very rapidly. This chattering did not occur at any time during the measurements so it was not able to be detected at that time. The plant electricians stated that the chattering was much more common during periods of high temperature, which was not the case during the measurements.

The chattering contactors would be a source of trapped charge on the capacitors. This would account for the transient overvoltages which damaged the capacitors and the transient overcurrents which caused the fuses to operate.

Once it was determined that the energization transients were most likely due to the chattering contactors, the contactors were replaced. The problems persisted, leading to further examination by plant electricians.

They reported that when variable Step 2 was brought on-line with variable Step 1 already on-line, the contactor for Step 1 would drop out and pick up approximately six to eight times within one minute. This would expose the capacitor to many switching transients. These would occur before the Step 1 capacitors would have had a chance to discharge. Some the re-energizations would be bound to occur when there was a large difference between the capacitor voltage (due to trapped charge) and the system voltage. This would lead to transient voltages and currents similar to those shown in Figure 10.

The next step was to replace the control board in the capacitor bank which monitored the power factor and determined which steps to bring on-line. Since a new board was ordered and installed there have been no capacitor failures or fuse operations in the bank confirming that the control board was the problem.

8.0 Summary

A steel processing plant was experiencing capacitor failures and fuse operations in an automatically switched, multiple step power factor correction capacitor bank. Initial impressions were that the problems were due to harmonics. This would not be unexpected in a system where harmonic sources, such as adjustable speed drives, are electrically close to power factor correction capacitors.

A preliminary assessment of harmonic resonant frequencies, in addition to the measured data, indicated that there was a resonant condition. The measured values were not high enough, however, to be expected to cause the fuse operations or the capacitor failures.

Examination of the fuses which had cleared indicated that low level transients, not harmonics, had caused them to operate. The measurements did not reveal any transients which would have caused equipment problems, but no problems occurred during the measurements so there were likely no significant transients to measure.


Simulations were performed to determine whether capacitor switching transients would have been able to cause the failures. The results of the simulations indicated that the capacitor switching transients could generate high enough I2t levels to cause the fuses to operate. This was only true if there were high levels of trapped charge on the capacitor step being switched in and the system voltage was near its peak at the time of the switching.

With high levels of trapped charge during switching, the capacitor voltages can also reach well over 2.0 per unit. These levels might not cause capacitors to fail immediately but could cause cumulative degradation of the capacitor dielectric, eventually leading to failure.


Even with worst case conditions, these t levels would not reach the peak let-through I2t of the fuses. The results of the simulations are therefore consistent with the fact that not all the weak spots in any of the fuses had cleared.

With this information, the plant electricians replaced the contactors that were believed to chatter occasionally. When the problems persisted, the electricians observed the operation of the capacitors and eventually replaced the control board in the capacitor bank. Since that time there have been no capacitor failures or fuse operations in the bank.



[1] “Electrical Transients in Power Systems,” Second Edition, Allan Greenwood, © John Wiley & Sons, Inc. 1991.

[2] IEEE Std 519-1992, “IEEE Recommended Practices and Requirements for HarmonicControl in Electric Power Systems,” © Institute of Electrical and Electronics Engineers, Inc. 1993.

[3] NFPA 70, National Electric Code, 1999 Edition, © National Fire Protection Association, Inc. 1998.

[4] “SPD Electrical Protection Handbook – Selecting Protective Devices Based On The National Electric Code,” © Bussmann, Cooper Industries 1992

[5] “Distribution System OverCurrent ProtectionWorkshop – Course Notes,” © Cooper PowerSystems, Inc. 1996.

[6] IEEE Std 18-1992, “IEEE Standard for Shunt Power Capacitors,” © Institute of Electrical and Electronics Engineers, Inc. 1993.

[7] IEEE Std 1036-1992, “IEEE Guide for Application of Shunt Power Capacitors,” © Institute of Electrical and Electronics Engineers, Inc. 1993.


Author’s Biography

Thomas M. Blooming, P.E. is a Senior Product Application Engineer for the Power Quality Division of Eaton Electrical. Tom received a B.S. in Electrical Engineering from Marquette University, an M.Eng. in Electric Power Engineering from Rensselaer Polytechnic Institute, and an M.B.A. from Keller Graduate School of Management. Tom works in the Power Factor Correction Group of Eaton Electrical (Power Quality Division). He handles application issues related to power factor correction capacitor banks, harmonic filters, static­switched capacitor banks, and active harmonic filters, as well as many power quality-related questions. Tom formerly worked in the Cutler-Hammer Engineering Services & Systems (CHESS) group and provided clients with electric power engineering expertise, focusing in the areas of power quality and reliability. Tom has performed numerous measurements and studies. In addition, he has published technical papers and taught engineering workshops and training seminars on power quality issues.