When I have a bank of many capacitors in series and parallel, what happens if one of the electrolytic capacitors fails and gets short circuited which is very probable with high voltage electrolytic capacitors?

Mahmood,

I disagree. For PEH 200 shell life at 40°C is ten years! Failure rate is 10**-7 per hour at 60°C! These capacitors are state of the art. You are not the first, who needs them for sensible high voltage tasks...


If an electrolytic capacitor fails, sometimes electrolytic material becomes so hot, that it goes into damp state. Then, of course, pressure inside becomes higher and higher and finally overpressure opening in capacitor housing opens. This is not an explosion, it's a fast opening BEFORE pressure becomes so high, that an explosion will occur!

If protection against damage is needed, look, that capacitors are not sitting in a totally closed enclosure. If the overpressure opening of capacitor opens, lot of damp is leaving the capacitor. This damp must find a way out of enclosure, otherwise pressure inside the enclosure could rise so much, that enclosure could explode! Put a protection wall between capacitor array and operator, so that nothing can hurt the operator.

These recommendations are valid, equally how many capacitors are failing or how much energy is involved.

A way to prevent that the whole battery is discharging through only one failing capacitor can be realized by the inserting of diodes, which prevent discharge in each unit of paralleled capacitors:



(In this schematic only three transformer windings are shown. This is only for simplicity purpose. If the transformer is powered by mains voltage, four windings are better.)
Another advantage of such solution is, that the current per charging diode is heavily decreased, in opposition to the 'one diode per transformer winding solution'.
Along each charging diode you will find a current limiting resistor Rlimit. This resistor prevents the transformer from being shortcircuited, when high current pulse is drawn from capacitor battery. It also makes it possible to use rather 'tiny' transformer for this task.
If a capacitor fails by showing a short circuit, others arround being connected in parallel, cannot be discharged through him. But still, according transformer winding will be overrated. Here, Rlimit can also help: If for this resistor a type is used, which provides also certain fuse characteristic, then there will be a full protection against damage. After a failing, damaged capacitor is disconnected by fused Rlimit.

In the schematic also for the discharging path (to external load) diodes are inserted. So, when a capacitor fails there's no way for eroneously discharging, either.

The many switches that can be seen represent your IGBTs, or whatever you use for switching. You could distribute the switching measure over the whole array, like shown in the schematic. By this, current rating per IGBT can be highly decreased. Please note, that also here fabrication of galvanically isolated voltages can ease the design of IGBT switching stages: If you use optically isolated drivers, you can reference the individual IGBT to the negative terminal of individual capacitor!
But if you insist on the central IGBT you can omit the many switches, of course.

Another possibility is to use relais contacts for the switches. Then, only when wanting to fire the current pulse with the help of central IGBT, switches are closed a little bit earlier. By this, time of existance of dangerous high voltage at output of capacitor array can be reduced to the moments, when your experiment is actually running. This is also valid, when instead using the many IGBTs...

Keep in mind, that the discharging diodes must be from the 'heavy ones'! Fortunately, due to the use of many diodes in 'parallel' current rating can be heavily decreased. Also, spike current through the external load is only lasting some milliseconds.
Sometimes, a low ohmic resistor is connected in series to each diode, like you can see with the charging diodes. The intention of this measure is, to guarantee that currents are distributed equally among all diodes, so that not only one diode must deliver the whole current. A standard diode makes this by itself, to some amount, but if this is not enough this low ohmic resistor is inserted.

Kai