I'm doing an arc flash study for a data center. This is the first time I've modeled UPSs in a system and am trying to determine the important and superfluous information. The component model in SKM is pretty simple and it appears I only need kVA, voltage, and pf to include it in my study. Should I ignore the static by-pass and internal switches in my model. Do the characteristics of the UPS prohibit a fault from passing thru it?

Eric-

Check your UPS specs. Most UPS systems have very limited fault clearing capability, 300-500% of UPS output, for just a few cycles. If you have multiple modules, this may become significant as a low level type arcing fault, but only if you assume it continues for the entire 2 seconds (or whatever duration), even though it isn't capable of producing that much current, for that long. In reality, it may only sustain an overload ~150% for that long.

Since the UPS has solid state components, it will not directly couple a fault from the utility or generator through to the load side.

For any decent fault, the UPS will go to bypass in less than a cycle, but I'm not sure how to model this in SKM. I don't think there is a reasonable way to approximate a static switch firing either.

I believe at a minimum, you should model the UPS for low level faults as a stand alone source. This may require multiple scenarios if there are multiple modules.

Scenarios should also be run in bypass mode. Again, this may require multiple scenarios based on utility variables, and additional scenarios for generator(s) in operation.

Good luck

Are you modeling the battery side (DC) as well? Those need working on also (and live, since there's no disconnect on a cell itself).

We aren't modelling the DC side until IEEE-1584 provides a method. This is the subject of some research that is in progress.

I agree that UPS's are inherently current limiting. 300 to 500% would be an upper limit and I would model closer to the 300% level.

If you have a static (manual) bypass switch, then I would recommend that you base you arc flash calcs with the UPS bypassed. This would basically be the line fault current and the highest value. Granted its not in bypass often, but thats the time most likely for live work.

As an alternative, the Electrical Safety Program could prohibit live work while the UPS is bypassed.

I'm interested to see if anyone has found the best way to model a static bypass switch. I too am doing an AFH study on a data center. I agree that bypassing the UPS's would make sense to find the AFH at other buses connected, but I don't know how to model the static bypass switch itself. If I have 4 500kVA UPS's connected to a bus I've got to assume that these units will contribute to fault current available on the bus.

If the assumption is that the UPS's themselves are significantly less than the line side available fault current and you model the condition with the bypass switch in bypass, then you ignore the UPS's entirely.

If however you are modelling the condition where the UPS's are in the circuit, then it becomes important to consider the TYPE of UPS. There are basically 3 types. In one type, the front end switches between the UPS and the line and works the same as looking at the bypass switch. The second type (line interactive) is slightly different but works about the same as far as faults are concerned. The third type, double conversion, is a bit different to understand. The UPS itself will have a maximum current listed as it's specifications. It is possible to pull about 2X this in the event that two IGBT's remain dead shorted such as if the firing board is messed up. Generally when looking at the DC side you may want to look at the series resistance of the batteries and use Annex D.8 to calculate the DC arc flash hazard but it usually ends up a non issue. You may also want to look at the size of the capacitor when the amount of fault energy released if the capacitor is dead shorted internally to the UPS or somehow during maintenance externally using the DC methodology (although the result is energy directly rather than calculating volts x amps) but again, there's not a lot of energy there.

This will satisfy your curiosity but also explain why the earlier suggestions to model it "with the bypass switch closed" (e.g. ignoring the UPS) was made.

One scenario - straight UPS, as Paul said about <2x FLA
Second scenario - static bypass closed, again as mentioned, it is common for UPS's to do this in order to provide enough current to clear faults.
Third scenario - maintenance bypass, it is not uncommon for the maintenance circuit to be entirely separate from the 'normal UPS' one, including the source.

It seems that often people performing studies are reluctant to investigate all of the potential operating and source scenarios in an order to find the worst case. For some reason they appear to just want to 'push a button' and get on to more important tasks.