In conceptual terms, this seems plausible. In terms of mandating this course of action, say in standards, what "rule of thumb" would (or could) be used for that?

Good question. I've been pondering how best to answer. I think it needs a bit of detail (so apologies in advance for the length of this...).

I think it's easiest to look at likely faults and what their characteristics would probably be:

L-N fault on the same line (phase) - point of the fault is pulled to around Uo/2 (say 115V for 230V) - if the EVSE Is directly downstream of the fault it'll see L-N voltage as zero; if the EVSE connection branches off somewhere upstream of the fault, there will still be some voltage difference L-N. So apparent L-N voltage somewhere between 0 and Uo - but should clear within 5s or so - either returning to normal or being disconnected.

L-N fault on a different line (phase) - N will be pulled toward the voltage at the point of the fault (Uo/2) but in phase with the faulty line. In the simpler case of a split-phase supply that would mean a difference from our line of 1.5xUo; for 3-phase I think it'll be in the region of 1.3xUo (sqrt( (Uo+Uo/2*cos(60))^2 + (Uo/2*sin(60))^2 if you want to check my trig). So EVSE should see L-N of between Uo and 1.3xUo (for 3-phase). Again should typically clear within around 5s.

L-L fault between this line and one other (3-phase) - N stays put but both Ls pulled towards a mid-point voltage - (please check my phasor thinking here) - which I think would put it a little below Uo/2. So again depending on the connections, EVSE would see something between a bit below Uo/2 and Uo. Likewise clearance in the region of 5s.

L-L-L fault - all lines pulled towards 0V - so EVSE sees something between 0 and Uo.

Broken PEN (or N) - L-N could be anywhere between 0V and U (U=√3 Uo) - may well fluctuate as loads connect/disconnect (or fry) - and won't clear for some considerable time (although may at times appear to be within normal range).

So observations so far:

If L-N remains within 0 to 1.3x Uo (or 1.5x Uo for a split phase system), and after a short period (a little over 5s say) it either becomes a steady 0V or steady Uo - then chances are it was a conventional fault rather than a broken PEN.

Conversely, if the voltage at any point in time exceeded 1.3x Uo (or 1.5x Uo for split-phase) or the disruption existed for longer than 5s, then chances are it is a broken PEN.

By and large I've presumed that L-PEN faults are similar to L-N ones. Although the DNOs often have longer disconnection times - so maybe extend 5s a bit - as long as the device opened at first hint of any trouble and we're only looking at decision to auto re-close or not, there's little harm in making the period longer (60s maybe? I'm not sure what DNO disconnection times are likely). Likewise L-PE faults, although reduced c.p.c.s and/or parallel paths can influence the numbers to some extent. We can probably debate what sort of number to assume for Uo - eg.. 230V, 253V or even 262.2V (maybe Uo isn't the best symbol).

Power cuts (or local disconnection) will of course appear as 0V L-N at the EVSE for a considerable period (as might a broken PEN on a single-phase system - but in that case it can't have PE-Earth > 70V, so less of an issue).

So my thinking so far beings me to:

1. As soon as the L-N voltage goes out of range, disconnect (within 0.4s say).

2. Continue to monitor the L-N voltage for a short period (>5s <60s say). If the voltage returns either to normal or 0V, and never exceeded 1.3x Uo then allow auto reset once the L-N voltage remains within range for another period (e.g. 1 min).

3. Otherwise presume an open-PEN fault, and either don't auto reclose, or demand an extended period of consistent normal L-N voltage (hours) before doing so.

There is a bit of a challenge in this - being able to monitor voltages (and duration) when the supply may be 0V. I think hat's probably doable though - either with a small battery backup (capacitor perhaps) or maybe more likely by using a battery backed-up real time clock (RTC) and logging (non-zero) voltage readings with a timestamp into non-volatile memory (pretty standard stuff these days) - once the power's back the log will reveal now long the power was off for, Likewise it might be necessary to treat all voltages below some safe threshold (50V say?) as 0V.

There might want to be some additional random delay in the re-close, to prevent all the EVSEs in a neighbourhood switching back on at the same moment when power returns after a power cut.

There are a few assumptions in all this - e.g. the voltage measurement isn't affected by surges (e.g. from switching off inductive loads) - so either such surges are filtered out or the voltage reading is averaged over a suitable length of time (a few cycles perhaps); Also that N/PEN and live conductors have roughly matching impedances.

I might have to think a bit more about brownouts (voltage reductions lasting some period) perhaps due to starting currents from nearby heavy motors.

   - Andy.

AJJewsbury  thank you for contemplating this. It's not an easy subject.

It might need to be the subject of some research to fully address this, as you say - especially voltage variations such as "brownbouts" and other network events.

AJJewsbury  thank you for contemplating this. It's not an easy subject.

It might need to be the subject of some research to fully address this, as you say - especially voltage variations such as "brownbouts" and other network events.

It might need to be the subject of some research to fully address this, as you say - especially voltage variations such as "brownbouts" and other network events.

Agreed! I wonder if anyone has any power analyser logs lying around that might help?

I wonder too how common "brownouts" are in the UK (unlike place like the US where they seem to endemic). The traditional light flickering when the factory next door starts up sort of thing perhaps needs only a modest voltage drop to be noticeable - I guess a drop from 240V to 220V would be very noticeable - so how often would it drop to below 207V? If there were deliberate 1976-style voltage reductions across the country in an attempt to reduce demand, load-shedding a few extra EVs might even been seen as beneficial rather than a problem.

   - Andy.