Who from the IET is working with the government to allow <800W inverters to be 'plugged in'?

In respect of BSI committee work (BS 7671 is developed by a committee jointly managed by the IET and BSI, but that follows the BSI rules for developing standards) specific discussions in committees cannot be shared. Similarly, content of standards and draft standards can only be discussed when they are published, or made available as drafts for public comment.

Would those rules prevent someone confirming whether a committee (or indeed which sub committee) had been asked to consider a particular request or requirement?

The comments on RCD operation seem somewhat unclear to me. Mark started off taking about a certain type of RCD that the Germans had the foresight to drop many years ago (which I take to be AC types) - but the article says "But some older RCDs were never designed for electricity flowing back into the circuit" - but bi-directionality is a completely different feature - and not one universally found even in the latest RCDs. I suppose we should add in single pole switching RCDs (mostly RCBOs) where the local generation isn't separated from Earth when the RCD trips, so the shock hazard persists until inverter shutdown - again a problem that isn't mitigated by ensuring your installation is up to current standards.

   - Andy.

Chris Pearson said:

BS 4293? ;-)

Not specifically, Type AC, but the debate might have started with international standardization on RCDs prior to the introduction of BS EN 61008 series and BS EN 61009 series.

A key thing about BS 4293 series, and earlier versions of BS 7288, is that they never included requirements for EMC testing, yet at the same time never prohibited electronic circuits from the detection and actuation paths.

AJJewsbury said:

Why shouldn't it see the fault just the same? With a N-PE link only on the grid side a L-earth fault on the load/local-gen side there will be an imbalance just the same - only difference is that the N coil will carry the "extra" rather than the L.

gkenyon said:

The 'bidirectional vs unidirectional' is far more recent, and itself has further complexities.

I don't know, but Amendment 3 was made for a reason.

If an RCD is unidirectional, it may well trip in both directions, but it is not certified to do so.

In an all RCBO board, all the other circuits will be fine irrespective of whether the source is the mains or solar, but that does not apply to the circuit to which the solar is plugged in.

I wonder how HMG will deal with this issue. Perhaps they will simply say that balcony solar is out of scope of BS 7671 because it is not part of the fixed wiring?

Chris Pearson said:

If an RCD is unidirectional, it may well trip in both directions, but it is not certified to do so.

Agreed, and because protective devices have been marked for many years (started with older British Standards prior to the ENs we use today), isn't it the installer's 'fault' if the connection is incorrect, or an inappropriate device is selected for a particular application?

However, if there's confusion in the industry, we need to address that.

Chris Pearson said:

I wonder how HMG will deal with this issue. Perhaps they will simply say that balcony solar is out of scope of BS 7671 because it is not part of the fixed wiring?

Well, that's my personal perspective on this pretty much summed up. BS 7671 doesn't tell consumers what they can plug into a socket-outlet. That's what the Electrical Equipment (Safety) Regulations is for.

At present, there isn't a designated standard for the products, so we can't really develop a 'use case' in BS 7671 with any additional provisions that may be necessary.

And any designated standard would have to address the fact that the products must meet the essential requirements of the Electrical Equipment (Safety) Regulations, including items (b) and (d) on that list, i.e. not generate unsuitable temperatures and consider the prevailing conditions (i.e. existing wiring practices in the UK).

So, if we will not be able to rely upon the presence of bi-directional RCDs (or any at all) in the fixed wiring, the balcony system will need one in its output.

But it won't have one.

What it will do is shut off immediately if the mains supply is disconnected. Which means that a non-bidirectional RCD won't burn out.

The whole bidirectional RCD thing is trying to solve a problem that in practice doesn't actually exist.

the balcony system will need one in its output.

But without a N-PE link at the balcony system (which would be prohibited for parallel operation with the mains) - an RCD at that end would never "see" the imbalance - so never trip.

   - Andy.

Simon Barker said:

The whole bidirectional RCD thing is trying to solve a problem that in practice doesn't actually exist.

AJJewsbury said:

But without a N-PE link at the balcony system (which would be prohibited for parallel operation with the mains) - an RCD at that end would never "see" the imbalance - so never trip.

What about island mode arrangements, where the inverter is self-commutating and is re-energized once the grid has been disconnected by the island mode isolator (now termed switching device for islanding, SDFI) and a system referencing connection made by the system referencing switch (now called SRCSD or system referencing conductor switching device)? 

At present, there isn't a designated standard for the products, so we can't really develop a 'use case' in BS 7671 with any additional provisions that may be necessary.

I guess JPEL/64 might be in a good position to anticipate likely problems and so feed into a suitable BS for such systems? (i.e. look at the problem the other way around).

One thought - the issue of several inverters fooling each other into thinking the mains is still present might be addresses by the inverters monitoring they output current? In the case of 4-inverters plugged into a 4-way extension lead which is then unplugged from the wall the output currents must surely drop to practically zero - so the inverters could use that as an additional signal that something was amiss and shutdown (perhaps until voltage had disappeared and then reappeared). In other situation where some loads might still be connected on the inverter side of the break, the output currents wouldn't drop to zero but in almost all cases would alter from what would normally be expected - so possibly still detectable by the same means?

The case of RCDs being potentially compromised by new types of "appliances" isn't new of course - we've had that a few times over with AC vs A vs F vs B vs B-EV types - yet BS 7671 hasn't seeked to prohibit such devices - it's merely updated the regs for new work to better handle the situation. I guess we'll end up in a similar situation with this one.

   - Andy.

What about island mode arrangements

I don't think anyone's suggesting an 800W plug-in balcony system should be capable of island operation (at least not when feeding back into the main "plug" - it would need upstream isolation from the grid - which would defat the easy plug & play aspect).

   - Andy.

AJJewsbury said:

I don't think anyone's suggesting an 800W plug-in balcony system should be capable of island operation (at least not when feeding back into the main "plug" - it would need upstream isolation from the grid - which would defat the easy plug & play aspect).

If we are keeping this purely to even if there are no other inverters, and we are looking at the plug-in solar debate, the power to the final circuit might be disconnected by the operation of the RCD because of a faulty appliance on that circuit ... mains loss protection is not instantaneous, as in 'zero time'.

G98/G99 has time delays for under-voltage and under-frequency, that are at least 500 ms (longer ... nearly twice as long ... as the disconnection time of the RCD). RoCoF doesn't have a time delay, but would still need some time to detect the phenomenon (somewhere between 100 and 500 ms).


So, I believe it still leaves the unidirectional vs bidirectional to be considered; I'd be happy for testing to demonstrate to me it's not an issue.

I accept that, if the RCD disconnects the neutral, the shock risk isn't increased as the disconnection time of the RCD isn't increased.

BUT

Many RCBOs are single-pole with unswitched neutral, so the presence of plug-in PV has effectively increased the RCBO disconnection time from 40 ms to at least 140 ms.

So ... split-load boards may well be safer for plug-in PV than single-pole RCBO boards !

Below from G98 Issue 2 2025:

AJJewsbury said:

I don't think anyone's suggesting an 800W plug-in balcony system should be capable of island operation (at least not when feeding back into the main "plug" - it would need upstream isolation from the grid - which would defat the easy plug & play aspect).

If we are keeping this purely to even if there are no other inverters, and we are looking at the plug-in solar debate, the power to the final circuit might be disconnected by the operation of the RCD because of a faulty appliance on that circuit ... mains loss protection is not instantaneous, as in 'zero time'.

G98/G99 has time delays for under-voltage and under-frequency, that are at least 500 ms (longer ... nearly twice as long ... as the disconnection time of the RCD). RoCoF doesn't have a time delay, but would still need some time to detect the phenomenon (somewhere between 100 and 500 ms).


So, I believe it still leaves the unidirectional vs bidirectional to be considered; I'd be happy for testing to demonstrate to me it's not an issue.

I accept that, if the RCD disconnects the neutral, the shock risk isn't increased as the disconnection time of the RCD isn't increased.

BUT

Many RCBOs are single-pole with unswitched neutral, so the presence of plug-in PV has effectively increased the RCBO disconnection time from 40 ms to at least 140 ms.

So ... split-load boards may well be safer for plug-in PV than single-pole RCBO boards !

Below from G98 Issue 2 2025:

G98 is not an appropriate standard for plug in solar so quoting is doesn't really help (as identified at the start of this thread).  In particular, G98 is built around hard wired equipment with a degree of tolerance to voltage / frequency fluctuation, with a reasonably long delay in ceasing generation so as to avoid unfortunate events where the grid might suddenly lose a significant chunk of generation at a national level because all the solar systems shut down.

With plug in solar, the trade-offs have to be different.  Plug in solar, by its nature, can be unplugged and the risk presented by a plug remaining live (or a connected circuit remaining live after a RCBO has tripped) for more than a few milliseconds is intolerable.  Plug in solar will either need a separate standard akin to G98, or a significant update to it.  Of course, plug in solar will as a result be an inherently less stable source of generation (more accurately load reduction!) for the grid than "normal" solar and that is something that those who operate the grid will need to buy into if plug in solar is expected to achieve significant scale.

As I noted in one of the other threads on this subject, there is a class of devices (inverter / chargers often used on road vehicles / smaller boats, operating in parallel with a shore supply) which have much commonality in terms of shutdown requirements with plug in solar - I don't know what standards they adhere to (I suspect there may be none applicable with respect to loss of mains detection) but whatever approach is taken there seems to work and perhaps might work for plug in solar.  I suspect that "line interactive" computer UPSs also need to detect loss of mains in order to avoid their plugs being live on loss so there might be some standardisation or learning in that area too.

An equipment standard on the subject of fast loss of mains detection and disconnection of plug in equipment connected to a separate power source would be a sensible thing to exist if it doesn't already given the various classes of device which might benefit from such a standard.

the trade-offs have to be different.

I agree entirely. The Germans felt it necessary to have their own standards as installation conditions vary across nations. As I recall Germany tends to use TN throughout, whereas we have a mixture of TN and TT installations - so while compromise of the effectiveness of RCDs might "only" result in loss/reduction in additional protection there, here it might mean basic ADS is compromised. The Germans have decided on somewhat different overall parameters too (e.g. the overall rating) so there's no precedent to comply with a single common standard (at least not in the short term - longer term no doubt a common denominator will emerge).

I think we should keep our minds open to other means of achieving similar ends too - e.g. if it were to be decided that the last major stumbling block was the loss of additional protection to other outdoor equipment fed by the same final circuit (e.g. cut lawnmover lead) then consumer advice to revert to using plug-in RCDs for their garden equipment (as they would in days past if the sockets weren't 30mA RCD protected) might be an adequate compromise.

But if on the other hand we decided that the only practical way forward was for inverter shut-down time had to be so fast (on top of typical RCD actual opening times) that neither ADS nor additional protection was significantly compromised, then that's what we should stipulate. (Cue others to point out technical limitations!). We don't necessary have to be limited to what has already been produced.

From the German paper showing waveforms on unplugging:

it seems that the current drops to zero just about instantaneously (green line) - perhaps a signal such as that could be used to trigger a much more prompt shutdown (or at least disconnection from the mains side).

   - Andy.

(edited to fix diagram)

It would not be hard though to add one line about plug in units  to the annex of G98 that covers loss of mains,  and make it the same as the standard for domestic and commercial equipment with a plug on it (EN IEC 62368-1:2024)  where no more  specialized standard applies.

That would require that pins must not remain "hazardous live" above ES1 limits for longer than 2 seconds after the mains plug is withdrawn.

Something like

'units connected by a removable plug with accessible pins shall meet the residual voltage requirements of EN 6236801, and residual voltage be below ES1 limits within 2 seconds of removal'

Now I've typed than in a few seconds at a cost of pence, so realistically, a joint govt and experts committee should be able to manage something similar in a few months and at a cost of millions.;-) 

In this context, depending what voltages appeared, those ES1 limits could require the voltage to fall below 60V DC or 30V RMS AC, if the available short circuit current at that time was more than 2mA DC or 0.5mA RMS AC respectively.

This long standing requirement is one reason we have bleed resistors on filter capacitors and brakes on very large motors that otherwise generate while spinning down.

That 2 second figure compares to G98 table 2 - while a modest under voltage should  hold on for ~ 2,5 seconds note that the thing that will operate when unplugged is the 2nd level fast under frequency detection of 0.5 seconds.  I suspect that most units will be fine with little or no adjustment.

note that the reputable German balkony  units seem to do all this, and also claim to stop inversion within 200ms of mains loss, and to be below 60V residual voltage on the pins within 2 seconds.

Mike.

mapj1 said:

That 2 second figure compares to G98 table 2 - while a modest under voltage should  hold on for ~ 2,5 seconds note that the thing that will operate when unplugged is the 2nd level fast under frequency detection of 0.5 seconds.  I suspect that most units will be fine with little or no adjustment.

There is no time delay for RoCoF ... and this will operate if the power is completely removed.

However, there is a delay whilst it's detected, so when we are looking at single-pole RCBOs protecting the final circuit, the inverter can continue to generate after the 40 ms disconnection time ... perhaps up to 100 ms. So, instead of max 40 ms for RCBO to disconnect, it's now 140 ms. OK for a fault to Earth in TN and TT systems, but NOT where the RCD is required to provide additional protection because someone is in direct contact with live parts (broken case of equipment, cut lawnmower lead etc.).

I take the point, but the risk to the user is not much changed as the lawnmower itself  was always allowed to hold up the voltage by generating as it spins down - that's the less than 2 seconds to less than 30V RMS. The fixed wiring is held to  a very different standard to the appliances that connect to it, in many ways actually, not just this.

If someone is between live and true earth, having cut an extension lead plugged into the same circuit as their plug in solar, the RCD will still do its task in time. This is just yet another appliance that adds some pulse-stretching to the voltage on the disconnected side. Actually I can imagine having fully loaded washer or a tumble dryer on the same circuit as the victim may be  more problematic than a mower, as there is no deliberate braking action and stopping time maybe longer. 

In these example graphs, the power is cut at the first vertical line, after the first 2 mains cycles, and the Vertical grid-lines are 0.2 second intervals, so this whole plot shows a machine that stops generating altogether in about a second, as both voltage and frequency decay.
Scaling the starting voltage from 230V, this example passes  below ES1 limits by 0.6-0.8 seconds. This is from a paper about designing control circuits for traditional  induction motors.  Big motors on inverters and VSDs don't behave like this of course and in any case it is all horribly variable with what else is plugged in at the same time and the impedance  reactance that is presented to the line, these are very clean plots.



regards,

Mike.

gkenyon said:

However, there is a delay whilst it's detected, so when we are looking at single-pole RCBOs protecting the final circuit, the inverter can continue to generate after the 40 ms disconnection time ... perhaps up to 100 ms. So, instead of max 40 ms for RCBO to disconnect, it's now 140 ms. OK for a fault to Earth in TN and TT systems, but NOT where the RCD is required to provide additional protection because someone is in direct contact with live parts (broken case of equipment, cut lawnmower lead etc.).

Once the RCBO protecting the final circuit opens, on a plug-in / balcony solar inverter with no solid neutral-pe link at the inverter, what is the current path for the person receiving an electric shock?

[Edit] Yes I was thinking of single module, SP+N RCBOs, where the neutral is switched. Agree, if it is purely SP with solid neutral, then the fault current is back via the neutral-pe link at the cut-out.

many older rcbo are single pole only. some are labelled SP+N which means anything the manufacturer decided on that day . in a proper double pole or single pole with switched neutral which is also bi directional i cant see the fault current path.

That depends if its an early single pole breaking RCBO, or a modern 2 pole one.

Edit, I was beaten to it !!

mike.

mapj1 said:

I take the point, but the risk to the user is not much changed as the lawnmower itself  was always allowed to hold up the voltage by generating as it spins down - that's the less than 2 seconds to less than 30V RMS. The fixed wiring is held to  a very different standard to the appliances that connect to it, in many ways actually, not just this.

If someone is between live and true earth, having cut an extension lead plugged into the same circuit as their plug in solar, the RCD will still do its task in time. This is just yet another appliance that adds some pulse-stretching to the voltage on the disconnected side. Actually I can imagine having fully loaded washer or a tumble dryer on the same circuit as the victim may be  more problematic than a mower, as there is no deliberate braking action and stopping time maybe longer. 

In these example graphs, the power is cut at the first vertical line, after the first 2 mains cycles, and the Vertical grid-lines are 0.2 second intervals, so this whole plot shows a machine that stops generating altogether in about a second, as both voltage and frequency decay.
Scaling the starting voltage from 230V, this example passes  below ES1 limits by 0.6-0.8 seconds. This is from a paper about designing control circuits for traditional  induction motors.  Big motors on inverters and VSDs don't behave like this of course and in any case it is all horribly variable with what else is plugged in at the same time and the impedance  reactance that is presented to the line, these are very clean plots.

I'm not in agreement with this 100 %.

Yes, there's the "wind down" time of certain appliances (not all); BUT:

1. With the lawnmower lead, the lawnmower has stopped, because I cut the lead ... that's why I picked up the now accessible energized part - I'm not picking up the lawnmower end !

2. Even taking into account a "wind down", that time is on top of the RCD disconnection time. When adding plug-in PV, the wind-down time is added to the disconnection time of the RCD plus the RoCoF detection time of the inverter.

We know that, if you cut one corner, accidents might happen, but they'd be unlikely. However, the more corners you cut, the more risky things become.

Whichever way you look at it, we are degrading the protection currently offered by BS 7671, in the full knowledge that that doesn't protect all of the people all of the time ...

Whichever way you look at it, we are degrading the protection currently offered by BS 7671, in the full knowledge that that doesn't protect all of the people all of the time ...

I agree, but my point is more that  we already do that every time we plug any appliance in that holds up the voltage for some short time after the breaker breaks - and there are more of those than just inverters, plenty in common use. As another example we also weaken the protection against direct contact when we plug in a bedside lamp and the protection damage with every flex. 

Perhaps the lawnmower lead is not the best example for what I was trying to say.

The corollary of course is that a perfectly safe  fixed wiring system that we dare not plug anything into is not a very useful thing, 

We only need to protect the people , to about the same level that we already have, when real products are plugged in,  To modify a common safety strap-line, just a few deaths is not really enough. 

Mike