High Voltage Capacitors Designed To Avoid Catastrophic Failure Modes Capacitorss

User Manual: capacitorss-avoid-catastrophic-failure Maxwell

Open the PDF directly: View PDF .
Page Count: 6

GENERAL ATOMICS ENERGY PRODUCTS

Engineering Bulletin

High Voltage Capacitors

Designed to Avoid

Catastrophic Failure Modes

F. W. MacDougall G. L. McKee, J.B. Ennis, R.A. Cooper

Maxwell Energy Products, Inc. San Diego, CA.

R. M. Ness, - Cymer, Inc. San Diego CA.

W. A. Reass – Los Alamos National Laboratories

Presented at:

IEEE Pulsed Power Conference

June 1999

© 1999 IEEE

www.gaep.com

High Voltage Capacitors Designed to

Avoid Catastrophic Failure Modes

F. W. MacDougall G. L. McKee, J.B. Ennis, R.A. Cooper

Maxwell Energy Products, Inc. San Diego, CA.

R. M. Ness, - Cymer, Inc. San Diego CA.

W. A. Reass – Los Alamos National Laboratories

Abstract:

A major concern in the operation of high voltage

capacitors is the failure mode at end of life. While

progress has been made in this area at lower voltages,

little has changed for high current capacitors operating

above 30 kV in steady state and pulsed power

applications. Based on a new development at Maxwell

Energy Products, Inc., it is now possible to build

capacitors that can operate with a high degree of safety at

high voltages. The capacitors reach the normal end of life,

while avoiding the catastrophic failures, collateral

damage, and unscheduled maintenance, associated with

the normal failure mode of today’s high voltage

capacitors. This paper discusses the problem, capacitor

failure modes, capacitor designs and describes a solution

to the problem of catastrophic high voltage capacitor

failures.

I. HISTORY

Over the years, the emphasis on pulsed power

capacitor development has been on cost. The minimum

requirements are normally expressed in terms of life

expectancy, inductance, and in some cases volume.

Some specifications go as far as to require that testing be

done to demonstrate the quoted capability while other

specifications have relied on extrapolated data from the

manufacturer as an evaluation method.

Few specifications have reference to the failure mode

associated with at capacitor. Probably the primary reason

this occurs is that no one wants to dwell on the inevitable

demise of the equipment. Until recently, it has been

difficult to get a manufacturer to accept a contract where

the capacitor failure mode is specified.

II. POWER FACTOR CAPACITORS

In the 1970s there was a major change in high voltage

power factor correction (PFC) capacitors that resulted in

an intensive investigation of the capacitor time current

characteristics [1,2]. The dielectric fluid in the capacitors

was changed from non-flammable PCBs to flammable

fluids like MIPB.

The time current characteristics for a high voltage

PFC capacitors as well as typical expulsion and current-

limiting fuses [3] is shown in Figure 1. There are three

general areas of the curves shown. If the curves are

extended out long enough in time, the devices will reach

“thermal stability” and the curves become constant

current curves. At the fast end of the curve, the curves

approach a constant I2t. The I2t value is often referred to

as the “Action”. The constant Action section of the curve

results from the inability of the devices to dissipate energy

from the faulted area in the allotted time frame.

Connecting the constant current and constant Action

sections is the middle of the time current characteristic

(TCC) curve.

0.01

0.1

1

10

100

1000

10000

10 100 1000 10000

Log Time in Seconds

High Voltage (15kV)

Power Factor Correction Capacitor & Fuse

Time Current Characteristics (TCC)

Log Current in Amperes

Safe Unsafe

10Amp K Fuse

Min. Melt

Max Clear

35Amp NXC Fuse

Min. Melt

Max Clear

1,000,000

Amp2 Sec

Capacitor

TCC

Current

Limiting

Fuse

Expulsion

Fuse

Figure 1. TCC CURVES

Early on, it was realized that fuses could be used to

minimize the collateral damage done when a capacitor

fails. For fused in general, there is a point where they

will start to melt and a point at which they will always

clear. If the fuse will clear before the case will rupture,

the capacitor is properly protected. Note that the

maximum clear curve for the expulsion fuse (10 amp K

fuse) shown in Figure 1 approaches a minimum time to

clear. This type of fuse will wait for a current zero to

clear the circuit which can be as long as 0.8 cycles or

0.013 seconds at 60 Hz.

There is second limit that must be considered when

fusing a PFC capacitors. If there is more than 9300

kVAR or 25 kJ in parallel, current limiting fuses must be

used since the expulsion fuse will not interrupt the parallel

discharge of energy into the capacitor that has failed and

the energy level will be high enough to rupture the

capacitor case due to stored energy in the bank.

III. PULSE POWER CAPACITORS

Unlike PFC capacitors, pulse power capacitors come

in a wide variety of designs for very specific applications.

No group has been studied as extensively as PFC

capacitors, with the possible exception of the capacitors

developed for NIF, but the principles developed during

the PFC capacitor study are applicable to energy storage

capacitors.

Fuses are available for pulsed power capacitors [4].

The time frame of interest for single shot pulse power

fuse is short and falls into the constant Action area.

Typical values are shown in Figure 2.

Maxwell

Value Range for Standard

22kV Capacitor Fuses

Item Small Large Units

Rated Current 25 59 Amps

Inductance 250 150

nH

20°C Res. 74 13 mOhm

250°C Res. 145 25 mOhm

Action (Min) 1.5k 54k Amp2 Sec

Action to Clear 6.3k 222k Amp2 Sec

2 3/8”D 12”

Figure 2. PULSE POWER CURRENT LIMITING

FUSES

Considering the Action integral in terms of Amp2 Sec

as it applies to energy storage capacitors is a little

difficult, but the units of measure can be changed. For the

evaluation of the amount of energy that can be absorbed

by a capacitor the values could be state as:

Amp2 Seconds = Joules/Ω (1)

Where:

Joules ≡ Energy Stored

Ω ≡ The failure resistance

The energy stored in a capacitor or capacitor bank is

relatively easy to calculate. Determining the fault

resistance is not. The resistance of interest is not the fault

circuit resistance but the resistance at the fault point inside

the capacitor that will generate the arc plasma and from

which no heat will escape. While this value is difficult to

measure, it can be said that capacitors with low resistance

faults can absorb more energy than those with high

resistance faults before the case ruptures, because the

energy is dissipated away from the fault point.

Figure 3 shows a typical foil capacitor that has failed

and has been dissected. A typical failure spreads

vertically because that is the way the layers are

mechanically set up and it penetrates horizontally through

layer after layer of dielectric and electrode picking up

energy as it penetrates each layer. As the layers of

electrodes are penetrated, they are also mechanically and

electrically interconnected. The resistance between the

foils after the fault is the resistance of in Formula 1. The

shorting of foils during the failure process is an

inconsistent process with a wide variation in resistance

that changes in response to current surges. However,

experienced capacitor designers can design capacitors that

will fair better than others in the dynamic environment of

an internal fault.

13”

Cut Here

View

Here

Failure Point

Failure

Point

Figure 3. FOIL ELECTRODE CAPACITOR

DISSECTION

Foil electrode pulsed power capacitors are available

today operating between 10 kV and 100 kV, with peak

currents in the ½ M Amp range, and storing 60 kJ that do

not tend to rupture when they fail as terminal to terminal

shorts. This can be attributed to the internal design and

construction of the capacitors.

Self-clearing capacitors utilize metallized electrodes

that are only a few hundred angstroms thick. Should the

dielectric fail, the capacitor is designed so that the fault

current will clear away the electrode before any

significant current can flow through the fault. Metallized

electrode capacitors are not self-clearing under all

conditions of operation. One way to defeat the self-

clearing mechanism is to increasing the voltage across the

capacitor until it shorts.

If the self-clearing mechanism is defeated, the fault

area looks similar to that shown in Figure 4. The lower

part of Figure 4 shows the layers inside a typical

metallized electrode capacitor. The black area going

from the endspray at one end of the winding to the other

is typical of what happens when the self-clearing

mechanism is defeated. Since the fault path is several

inches long, the relative resistance of the fault is high.

The amount of energy that the capacitor can absorb with

this type of fault without rupturing is very limited. The

larger metallized electrode energy storage capacitors store

well over 25 kJ and tend to rupture the case if a fault like

that shown in Figure 4 occurs inside the capacitor.

Dielectric ~8µm Electrode ~300Å

~5”

Cut Here

View Here

End

Spray

Figure 4. METALLIZED ELECTRODE

CAPACITOR DISSECTION

Again, experienced capacitor designers have

developed commercially available capacitors for pulsed

power application that do not suffer from this failure

mode. The capacitors are available at voltages of up to

40kV.

IV. HIGH VOLTAGE CAPACITORS FOR

CONTINUOUS PULSED POWER

OPERATION

For some applications, capacitor banks are designed

with high voltage capacitors that must operate under

continuous pulse power duty. Often, the inductance and

resistance of the fuse as listed in Figure 2 will be more

than can be tolerated by the machine design parameters.

Whether or not fuses are used, the failure of a capacitor

comes as a surprise and is often accompanied by an

unscheduled equipment shutdown.

The capacitors used are typically designed for long

life and, as a result operate at fairly low stresses compared

to their high energy density single pulse counterparts. As

shown in Figure 5, the failure normally progresses from a

single shorted internal series section (5b) to the point

where about half the windings are shorted (5c) and at that

time, the remaining windings tend to short

simultaneously. Generally, because of the high peak

currents associated with pulse power applications, the

failed capacitor will not reach thermal equilibrium and the

failure will progress. The time it takes to for the capacitor

to progress from the initial fault (5b) to the point where

the capacitor is a dead short (5d) is significant. During

this time, there may be arcing at the failure point and gas

generated inside the capacitor. If this situation is allowed

too continue for any length of time, it is likely that the

capacitor case will rupture resulting in the unscheduled

shut down of the equipment.

abcd

aNormal Capacitor

bFirst Section Failure

c< Half the sections failed (Not a usual

condition for high energy density

pulsed power capacitors)

dCapacitor shorted terminal to terminal.

Figure 5. TYPICAL FAILURE PROGRESSION IN

A HIGH VOLTAGE FOIL ELECTRODE

PULSED POWER CAPACITOR

C1

C2C4

Unbalance

Threshold

Circuit

Unbalance Detection Output

-

+

Terminal

Case

Terminal

C3

Figure 6. HIGH VOLTAGE CAPACITOR WITH

AN INTERNAL UNBALANCE DETECION CIRCUIT

Figure 6 is a schematic of high voltage capacitor with

an internal unbalance detection scheme. As long as the

impedance of C1/C2 = C3/C4, there is no voltage across

the unbalance threshold circuit. If one of the internal

series sections shorts, the balanced circuit will become

unbalanced and there will be a signal. A typical threshold

detector is shown in Figure 7. Here the unbalance signal

is rectified and accumulated on a capacitor until the

voltage is high enough to cause the neon lamp to flash.

The output of this circuit is a series of light pulses with

the flash rate increasing as the failure in the capacitor

progresses.

Over Voltage Protection

Rectifier Circuit

Signal Accumulator

(Capacitor)

Threshold Detector

(neon lamp)

Unbalance Signal Input

Light Pulse Output

If one of the groups of internal capacitors sections in

Figure 6 starts to go through the failure sequence of

Figure 5, the flash rate can be used to identify the initial

internal capacitor section failure. It can also be used to

determine if the internal failure has progressed beyond the

initial section failure.

For many applications, capacitors can be designed so

that the capacitors will remain stable for some period of

time after the initial fault has taken place. That being the

case, an ALARM function associated with the initial

internal capacitor section failure, can be used to indicate

that the capacitor needs to be change out at the next

scheduled maintenance. Also, if the failure progresses

from the initial internal section failure to the point where

a major fault is of concern, a second signal level can be

used to shut down or STOP the equipment before a

catastrophic failure takes place.

Figure 7. TYPICAL THRESHOLD DETECTOR

(Relaxation Oscillator)

Pulse Count

Comparison Circuit

TD

TD

TD

Normal

Alarm

Stop

Unbalance Control

Capacitor

Capacitor

Voltage Divider

High Voltage Bus

Ground Potential Bus

Pulse

Counter

Optical

Output

Pulse

Counter

Pulse

Counter

Pulse

Counter

Pulse

Counter

Pulse

Counter

Figure 8. SCHEMATIC DIAGRAM OF AN UNBALANCE DETECTION CIRCUIT

In order to have an ALARM and STOP function, the

capacitor must be designed to carry the available current

after the initial internal failure has occurred. Also, the

circuit must be sensitive to the fault level. The schematic

of Figure 8 shows a typical schematic for a high voltage

capacitor bank using an unbalance detection scheme to

monitor the condition of the capacitors. The capacitors in

the bank each have two internal threshold detectors (TD)

feeding into a single optical output. The light output is

fed into a pulse counter that keeps track of the condition

of the capacitors.

The unbalance signal associated with the failure of an

internal capacitor section can be calculated in terms of the

bus voltage. In pulse power applications, the bus voltage

can vary significantly, operating at different levels,

charging and discharging the capacitor. In the circuit of

Figure 8, a separate capacitive voltage divider is used to

determine what voltage signal is on the bus.

A separate circuit compares the output from the bus

to the output from the capacitors to determine the relative

unbalance each of the capacitors in the bank. The output

of the comparison circuit is a set of contacts that indicate

that the situation is NORMAL, or an ALARM for the

initial failure of an internal capacitor section, or a STOP

signal that will be generated if the internal failure

progresses to the point of concern.

Figure 9 is an example of a high voltage capacitor

might looks like with an internal unbalance detection

scheme and fiber optic output. At the time of this

writing, the unbalance detection scheme with its controls

is still under development. Patent applications have been

filed. The schemes that are being evaluated at this time

are application specific. It is believed that this effort will

result in the development of equipment that can be used in

general applications

V. CONCLUSIONS

An understanding the failure modes of pulse

discharge capacitors can allow users to avoid the

problems associated with capacitor failures. There is

equipment available to day that will allow the user to

avoid most catastrophic failures in a wide variety of

applications. These include the use of fuses, as well as

properly applied foil electrode and metallized electrode

capacitors. One area where there is a lack of good

solutions is the high voltage repetition rate capacitor

banks discharging in microseconds. Here, the unbalance

detection scheme described above is designed to fill the

void and allow continuous operation with minimal

disruptions due to failing capacitors.

Capacitor on

Mounting Stand

Fiber

Optic

Cable

Plastic

Case

Terminal

Figure 9. HIGH VOLTAGE PLASTIC CASE

CAPACITOR WITH AN INTERNAL UNBALANCE

DETECTION SCHEME AND A FIBER OPTIC

OUTPUT

VI. REFERENCES

[1] Lawrence M. Burrage, Shunt Capacitor Rupture

Prevention – Large Bank Applications, IEEE

A76 366-5

[2] Lawrence M. Burrage, Capacitor Tank-Rupture

Prevention – I2t Considerations, IEEE Transactions

on Power Apparatus and Systems, Vol. PAS-97, no 2,

Mar/Apr 1978, IEEE No. F 77 567-1

[3] Cooper Power Systems Reference Data: R230-91-1,

R240-91-2, R240-91-38.

[4] Maxwell Energy Products – High Energy Capacitor

Protection Fuse Selection Guide.