Capacitor Case Rupture and How Fusing Prevents It

What happens inside a failing capacitor unit

A capacitor unit is a sealed can of dielectric elements. In older units with kraft paper or PCB-impregnated dielectric, a failed element did not weld shut: the cellulose kept arcing, the paper charred, and gas built up inside the sealed can. A widely cited protection paper notes that in many cases that gas buildup ruptured the unit.

Units built with a few layers of impregnated polypropylene film behave differently. A failure tends to weld the foils together into a solid connection without arcing or charring, and a relay maker's paper notes that modern all-film units do not bulge from internal gassing when a failure occurs; in an externally fused bank the spring-loaded fuse pops out instead. The risk has not gone away. A welded element shorts its section, the remaining elements see more voltage, and if the unit then fails through and draws fault current, internal arcing and gas generation return.

Field experience reported in a utility co-authored paper is that electrical arcing and case ruptures are no longer the major causes of capacitor bank tripping; animals, contamination and human error now cause most trips. That is partly a result of fusing and protection doing their job, which is why the checks below are still made.

Where the energy comes from

A failed unit receives energy from two places, and which one dominates depends on how the bank is connected.

The system. One manufacturer's capacitor fuse selection guide explains that in single series group grounded wye or delta banks, faults are inductively limited: the fault current is limited only by the available system fault current. In ungrounded banks and banks with several series groups, faults are capacitively limited, and the available fault current is typically less than two or three times the bank load current.

The parallel group. When a capacitor fails, the energy stored in its series group is available to dump into the failed capacitor and its fuse. The guide calculates parallel energy with the capacitance charged to 1.1 times the crest of rated voltage, which it states as 3.19 joules per kvar, or 3.80 joules per kvar for a 20% overvoltage. Adjacent banks on a common bus add to it, a point PJM's shunt capacitor guideline makes explicitly.

The fuse also passes energy as it operates. The guide defines I2t let-through as the I2t the failed capacitor sees while the fuse is operating. For any fuse, I2t to clear is always greater than I2t to melt.

Worked example: energy into a failed unit

The 3.19 figure can be checked from first principles. Capacitance is C = Q / (2 pi f V^2) with Q in var and V in volts, and stored energy is 0.5 C V^2. Charged to 1.1 times crest, E = 0.5 x C x (1.1 x 1.414 x V)^2 = 1.21 x Q / (2 pi f). Per kvar at 60 Hz that is 1.21 x 1000 / 377 = 3.21 joules; the guide's 3.19 reflects a rounded constant. At 50 Hz the same arithmetic gives 1.21 x 1000 / 314.2 = 3.85 joules per kvar.

Now a bank with 12 units of 200 kvar in each parallel group. When one unit fails, the other 11 discharge into it: 11 x 200 = 2,200 kvar, and 2,200 x 3.19 = 7,018 joules. Double the group to 24 units and the failed unit sees 23 x 200 = 4,600 kvar, or 14,674 joules. A single unit holds far less: a 50 kvar, 7,200 V unit (2.56 uF) charged to its 10,182 V crest holds 0.5 x 2.56e-6 x 10,182^2 = 133 joules.

The table converts the published limits to kvar at 3.19 joules per kvar so they can be compared with a bank layout. Each limit comes from a different source and fuse type; attribute the one you use and do not merge them.

Reading a case rupture curve

Capacitor manufacturers publish case (tank) rupture curves for their units. A software vendor's application guide defines the curve as an internal gas pressure limit caused by an internal arcing fault, and notes that the engineer setting the protection cannot calculate rupture points; they come from the manufacturer. Faults cleared to the left of the curve indicate minimal likelihood of rupture.

The curves show a safe zone, where the result is usually no greater damage than slight swelling of the case, and an unsafe zone, where the case may rupture. The 2023-cycle draft of NEC Article 460 uses the same idea, defining the safe zone as low probability of damage other than a slight swelling of the capacitor case, as identified by the case rupture curve.

One capacitor maker states the coordination rule: the fuse maximum clear curve must fall below the tank rupture curve at and below the level of available fault current, and for high fault currents the rupture curve should be compensated for asymmetry. Its curves note that for times shorter than 0.1 second the current axis is in asymmetrical rms amperes, and they mark a minimum clearing time of 0.8 cycle for coordination with expulsion fuses.

• Use total (maximum) clearing time, not minimum melt, for the rupture check.
• Cover the whole fault range, from low capacitively limited currents up to the full available system fault current.
• At short times, compare in asymmetrical rms amperes, as the curves note.

Published tank rupture values

Some makers print the I2t that sits under their curves. The short-time part of a rupture curve is close to constant I2t, so time at a given current is roughly that I2t divided by current squared. On one maker's standard-duty curve, 2,000,000 A2s at 10,000 A works out to 2,000,000 / 10,000^2 = 0.020 s, about 1.2 cycles at 60 Hz. At 8,000 A it is 2,000,000 / 64,000,000 = 0.031 s. A fuse that clears an 8,000 A fault must therefore pass less than 2,000,000 A2s, with asymmetry accounted for. A software vendor's example curve for a 300 kvar, 13,800 V unit runs from 275 A at 300 s to 10,000 A at 0.0133 s, and its short-time points work out to about 1,300,000 A2s.

The fault current ceiling for expulsion links

Expulsion links reach a limit as available fault current rises. The capacitor maker cited above states that the largest links that coordinate with its units are 100K and 65T, and only up to the fault currents below, which fall as X/R rises because asymmetry adds energy. Above those levels it lists three options: current-limiting fuses, ungrounded wye operation, or moving the rack. A municipal utility's protection standard states the same boundary in its own words: fusing greater than 100K is not allowed because of the capacitor tank rupture curve.

How the fuse type changes the energy split

In an expulsion fuse, the link melts, the arc acts on the fiber tube lining to build pressure, and the fuse continues to conduct until a natural current zero. The selection guide notes that most of the parallel stored energy is absorbed in the failed capacitor, which is why expulsion fuses carry parallel energy limits.

A back-voltage current-limiting fuse absorbs a large portion of the energy itself, the guide says, while a current-chopping design leaves trapped voltage on the other units. Current-limiting fuses also reduce peak let-through current, often to a fraction of the available fault current, which lowers the energy reaching the failed can in inductively limited faults.

Breakers are not a substitute. A software vendor's capacitor protection guide notes that the relay-breaker combination is generally not fast enough to protect the capacitor from case rupture due to internal arcing faults, and that providing that protection is the fuse's purpose. DENCO's catalog describes its C-rated capacitor fuses as protecting capacitors and capacitor banks against case rupture and dielectric failure, as current-limiting fuses with silent operation and no discharge. For a rupture check, ask DENCO engineering for the total clearing curve of the specific rating.

Fusing lowers the risk; it does not remove it

One capacitor maker's instruction sheet is direct: it is possible to get explosive case rupture in any application, even with proper fusing. It also warns about partial fusing. For three-phase capacitors fused on only two terminals, or single-phase two-bushing capacitors fused on only one terminal, applied on delta or ungrounded wye systems, an internal fault from the unfused phase to case might result in case rupture. The same sheet notes that capacitors contain a Class IIIB combustible liquid that could ignite if the case punctures or ruptures in the presence of an arc.

Low-voltage banks have the same physics. A fuse maker's application note explains that when one capacitor in a parallel bank fails, the others discharge into it and violent case rupture can result, and that individual capacitor fusing addresses it. Internally fused units handle it inside the can: a relay maker's paper notes that instant disconnection of a failed element prevents sustained arcing and minimizes the risk of can rupture.

Data for a case rupture check

To run the check, you need data from both the capacitor side and the fuse side.

• Case rupture curve for the exact capacitor unit, and its duty class, from its manufacturer.
• Unit kvar and rated voltage, and the number of units in parallel in the group, plus any adjacent banks on the same bus.
• Bank connection, which decides whether faults are inductively or capacitively limited.
• Available system fault current at the bank, with X/R for asymmetrical values.
• Fuse total clearing curve and total clearing I2t for the candidate rating.

Common mistakes

These errors show up when rupture checks are done in a hurry.

• Checking minimum melt instead of total clearing against the rupture curve.
• Comparing symmetrical currents at times below 0.1 second, where the curves call for asymmetrical rms amperes.
• Using a typical curve from a different maker or duty class instead of the curve for the actual unit.
• Counting only the bank's own units for parallel energy and ignoring adjacent banks on the same bus.
• Mixing limits from different sources, for example applying one maker's 20 kJ expulsion figure to another maker's fuse.
• Raising a fuse rating to stop nuisance operations without rechecking the clearing curve against the rupture curve.

Frequently asked questions