We are currently in the process of updating our short circuit / coordination / Arc Flash Risk Assement study. We are doing this due to modifications in our power system. How far down in the system do I have to go to satisfy NFPA 70E (what voltage level)? Our incoming voltage is 34.5kv.

NFPA70E applies all the down to 50V.
Arc flash calculations, using IEEE1584, are not practical below 208V, so other methods must be used at these levels.
Don't forget to include a risk assessment as well as the hazard assessment, when you update your Electrical Safe Work Practices procedures.

Effectively, if it is energized and you are going to touch it, you need to know what PPE is appropriate.

You will also find that paradoxically, the higher the voltage, generally the lower the incident energy.

There is some controversy about the lower voltage levels. In general, lower voltages (120V 240V) cannot sustain an arc. Further more, the tools for single phase calculations are not completely tested. In addition, single phase arcs have less energy.

We still maintain that below 240V/125KVA does not need to be modeled, until there is better data. We also do not model an single phase since the tools are not accurate. I do not find many large single phase panels anyway.

There are the practices that we use but you need to research and justify your study parameters. There is plenty of data/discussion on this forum.

"In general, lower voltages (120V 240V) cannot sustain an arc"

This sort of blanket statement has always bothered me.
Arc welders operate with open circuit voltages below this level,
and yet manage to sustain an arc for extended periods.
Clearly it is POSSIBLE, although unlikely

Need more clarification indeed. Lower voltages cannot sustain an arc over the arc gaps used for terminals with that type of equipment (1/4" or larger). Arc welders use much smaller arc gaps. Also and this matters, arc welders have some kind of insulating gas or liquid surrounding the melt zone (flux or gas). Without this protective envelope, the metals would oxidize rapidly and prevent further arcing. So the two are not directly comparable.

Also there is a difference there as far as "cannot be sustained". An arc CAN certainly be sustained, but may not be sustainable "indefinitely" where "indefinite" following current practice would mean for at least 2 seconds. So during those two seconds we need to know how much heat (incident energy) is released, and for how long (if it doesn't go for a full two seconds).

As to the comment that single phase arcs haven't been modelled, actually paradoxically this is not true. At least in open literature, 3 phase arcs are empirically modelled to a high degree of accuracy but not theoretically. In contrast single phase arcs have been modelled historically such as by the Duke Heat Flux software or ArcPro, and some empirical calculations are available from EPRI that do a pretty good job overall. These methods are used specifically for calculating incident energy above 10 kV, where the empirical model in IEEE 1584 at the current time does not work.

The inherent difficulty with single phase low voltage arcs (or even 3 phase arcs) is that they are not very stable. Predicting incident energy for these arcs will probably of necessity have to include a maximum arcing time value but so far the data is all over the board so it somewhat defies modelling attempts. So far most of the attempts try to "normalize" it to 2 seconds and then model it based on this but it is very obvious from looking at the data that this approach is not going to work.

Take your study down to 208 VAC utilization equipment connected to a 75k VA transformer; direction is from a recent (Feb 2015) IEEE 1584 class I attended in Anchorage Alaska.

If you see high calculated incident energies for this equipment, due to 2 sec trip times, look at reducing your "trip time" to one (1) second--see [/url]https://duckduckgo.com/l/?kh=-1&uddg=http%3A%2F%2Ffiles.bgbsupply.com%2Fbgbsupply-files%2FMersen-Effect-of-Insulating-Barriers-in-Arc-Flash-Testing-White-Paper.pdf.

David