I've only ever dealt with smaller UPS units but yes, ADS can be a challenge especially for submains. I would go to the manufacturer for their data. Note that unless they're specifically designed for it, you'll only get perhaps something of the order 10-50% over full load current as an overload with a specified time limit before the units shut down, but not in a controlled way.

Earth faults can be detected with RCDs etc set sensitively, but another gotcha to be wary of will be how the earth is derived: On principle you might not be able to rely on any upstream N and E link so do check what the UPS does.

Manufacturer's info, but a rough guide for inverter output is they typically current limit at around 20 % above their nominal output rating, or:

I_g(max) = 1.2 × I_g

Where

I_g(max) is the prospective short-circuit current of the inverter at the output terminals; and

I_g is the maximum nominal output rating of the inverter.

(Section 14.2 of the IET Electrical Installation Design Guide.)

Up to this maximum current limit, the inverter typically acts as a constant current source and internal impedances can often be ignored, although again the manufacturer may be able to provide more information on their product to permit you to take something into account so that the loop impedances for RCD (and if they meet the criteria OCPD) can be verified to comply with BS 7671.

The bad news is that the 60909 method, which is usually gives results that are close enough for the public supply, does not  work at all for private generators or  especially for UPS/ inverter derived supplies.
The reason is that that the it is not sensible to build a UPS or inverter capable of delivering many times its normal maximum load just so it can blow a fuse once or twice during its design life.. - I'd suggest that one that can deliver twice the nominal rating for more than a cycle is probably over-sized.

The solution is either earth fault trips, and not worrying too much about selective tripping on  overloads, or to immediately divide the output of the UPS into several breakers, any one of which will happily instant trip at a current that does not overload the inverter.
Either approach usually looks a bit odd compared to normal wiring.

And yes, watch the neutral-earth bond thing. There may not be one, and any that is added externally may need to be switched in on UPS operation in the UK to meet ESQCR . Other places may permit multiple NE links on one system, in the UK we by and large do not, but some big American UPS makers do not realise this.
Mike,

I think you meant to say constant voltage up to the 120% of I max, and then constant current and falling voltage, or indeed roll-back (both falling voltage and falling current) as the load increases beyond that point.

Mike

mapj1 said:

voltage

Sorry yes, in self-commutating mode that;s the case.

Effective constant current-source if inverter is grid-connected (for a given output power) ... which a UPS inverter would never be.

I raised a similar question last week, which has a similar issue for stand-by generators, I still haven’t got to the bottom of this, particular when a power distribution has been initially design around the supply transformer, what parameters would you use to disconnect ADS, clearly when a UPS or Stand-by generators are in circuit parameters are totally different and react so under fault conditions.

Thanks for the great replies. 

I never considered looking at the UPS as just an inverter output and as a constant voltage source. I was thinking more about the energy limiting effect of draining the capacitors in the DC stage and then the inverter adding some unknown impedance to that and generally overcomplicating things. 

So it seems then by considering the UPS as a constant voltage source we have two scenarios to consider. 

1) small / downstream final circuits where the fault current is going to be limited by the circuit impedance to well below the UPS max output

2) large circuits / submains where the fault current required by the OPD is greater than the UPS output  

In the first scenario we would calculate the fault loop impedance going back to the UPS but would consider the UPS itself to have zero impedance. The worst case lowest fault current will actually be in bypass because of the contribution of the upstream impedances (although made more complicated because the source voltage might be higher than the UPS output voltage setting!) 

In the second scenario the fault will draw more current than the UPS can deliver and it will switch to bypass and we are back to just calculating the loop impedance like usual back to the transformer.

We therefore really only need to consider the protection settings for when the system is on bypass.

Does this make sense or have I missed something?

What if it cant switch to by-pass ... ie. the mains has actually failed & its running as a 'true' UPS powered by its own batteries?

In the first scenario we would calculate the fault loop impedance going back to the UPS but would consider the UPS itself to have zero impedance.

That would imply that the UPS is capable of delivering infinite current (limited only by the downstream wiring) - I think you're going to be a bit disappointed on that score. Constant voltage perhaps, but with the equivalent of a large internal resistance I would have thought (or more likely the UPS's own protection systems would step in).

Where the UPS can reliably switch to bypass on overload - then it might be a little simpler - mains like when mains is present, but if mains is absent then switching to by-pass effectively provides automatic disconnection. Whether all that happens fast enough to satisfy the requirements for ADS (let alone any hopes of discrimination) would still need to be established though. RCDs (or equivalents) can make the problem a lot easier though.

   - Andy.

I split the two scenarios up to try to make it clear that I wasn't claiming the UPS can't deliver infinite current but is limited to it's rated current.

Up to that limit, if it acts as a constant voltage source then that would imply zero impedance contribution to the circuit