But now looking at when solar is generating power, the supply from the solar is just protected by an MCB. Therefore when surplus solar is feeding the garage and submain to the house, from what I can see the garage circuits and submain do not benefit from RCD protection.

Not quite. Points to note - 1) the PV "source" doesn't have a N-PE link and 2) the RCD should be double pole.

That means: a) the grid-side RCD will see any current escaping from the circuit regardless of which direction power is flowing in. b) When the grid-side RCD disconnects, it opens N as well as L, so disconnecting the circuit from earth - and so although the circuit still has power from the PV inverter, it's in effect a separated circuit, so safe from the point of view of electric shock (to Earth) even though a L-N voltage persists.

    - Andy.

AJJewsbury said:

2) the RCD should be double pole.

All RCDs providing additional protection connecting generators, so both ends of the cable if you fit one at the remote end.

Also, if appropriate, RCDs providing additional protection in the chain right back to the origin (although in this case we would not have selectivity because of the additional protection - residual current rating 30 mA or less, and non-delay).

Alan B said:

My initial thought is to add a bi directional RCBO where the solar feed is connected to the garage board.

Is this needed if the inverter is not used as a grid-forming inverter in island mode? If the RCD at the "DNO supply" end of the cable operates, a grid-connected inverter conforming to G98/G99 requirements will automatically stop generating, usually rather rapidly due to rate of change of frequency (ROCOF) trip requirements.

I agree it may be required if the inverter operates as a grid-forming inverter in island mode.

Alan B said:

EICR with a house connected to a detached garage via a submain, part of which is PVC T+E and due to the route it takes through the house requires RCD protection.

In any case, more importantly, you need to select the correct Residual Current Type of RCD ... BUT FIRST check whether the inverter manufacturer recommends that RCDs are NOT used for the connecting circuit. If you don't, you might find that the RCD experiences lots of unwanted operation.

Basically, if RCDs are recommended not to be used by the inverter manufacturer, it may require another solution, such as replacing the wiring with SWA, or use of earthed metal containment, which perhaps should have been done in the first place under those conditions.

gkenyon said:

Is this needed if the inverter is not used as a grid-forming inverter in island mode? If the RCD at the "DNO supply" end of the cable operates, a grid-connected inverter conforming to G98/G99 requirements will automatically stop generating, usually rather rapidly due to rate of change of frequency (ROCOF) trip requ

I was considering this as a thought process to follow but unsure if it's acceptable as I couldn't find anything in BS7671 discussing it. Sounds like it is acceptable.

gkenyon said:

In any case, more importantly, you need to select the correct Residual Current Type of RCD ... BUT FIRST check whether the inverter manufacturer recommends that RCDs are NOT used for the connecting circuit. If you don't, you might find that the RCD experiences lots of unwanted operation.

Agreed I should check for information. But it's been connected to a standard unidirectional 30mA RCBO for more than 10 years without any problem, suspect it will be fine. It has to be worth a go, £20 for an RCBO against at least £1000 for alternative approaches.

AJJewsbury said:

b) When the grid-side RCD disconnects, it opens N as well as L, so disconnecting the circuit from earth - and so although the circuit still has power from the PV inverter, it's in effect a separated circuit, so safe from the point of view of electric shock (to Earth) even though a L-N voltage persists.

AJJewsbury said:

b) When the grid-side RCD disconnects, it opens N as well as L, so disconnecting the circuit from earth - and so although the circuit still has power from the PV inverter, it's in effect a separated circuit, so safe from the point of view of electric shock (to Earth) even though a L-N voltage persists.

I understand the principle. Need to do some more reading in BS7671. I have a vague recollection that floating supplies could only be used when supplying a single point or single circuit.

Need to focus on preparing for my napit assessment at the moment, will look at this again at the end of the week.. 

The following applies to this discussion as far as I can see:

The Code of Practice for Grid-connected Solar PV Systems (second edition) includes a requirement (section 8.8) “Solar PV systems shall not be installed to the load side of any RCD that is shared with other circuits (for example, where the RCD is feeding a number of circuit breakers / circuits).” with an explanatory note “this is necessary to ensure the continued safe operation of the RCD, prevent the possibility of a fault on a circuit continuing to be fed by the inverter for up to five seconds after an RCD has tripped, and prevent nuisance tripping”.

So having an RCD only at the house side is problematic.

A more general requirement for generating sets not sharing RCD protection is also included in the Draft Amendment 4 to BS7671, 551.7.1 (d): “A source of supply shall not be connected to the load side of any RCD providing additional protection in accordance with Regulation 415.1 that is shared with other circuits...”

ENA EREC G99 as well as the preceding G59 stipulate a delay time of 0.5 s before the Loss-of-Mains protection is triggered. For small inverters up to 16A, G98 does not mention a delay time so response may be swift – but the preceding G83 could allow a trip time up to 1 s. It should be known / visible which of these protection settings apply to the existing inverter. I do not know why the CofP mentions a more pessimistic five-second timescale.

Altogether, if there is a way to bypass the problematic section of cable in the house (SWA around the house for instance), and thus avoid the need for an RCD at the house end, that would be ideal – but may be a lot of hassle. Failing that, I would vote for RCD protection at both the house and the garage end – with the RCD for the PV not shared with other circuits in the garage. This will not allow for selectivity between the upstream and downstream RCDs but that is probably the lesser evil.

I have a vague recollection that floating supplies could only be used when supplying a single point or single circuit.

Your vague recollection isn't far off. There are two versions of separated circuits - one for single items (413) and another for multiple items (418.3), the latter however is only meant to be used under the control or supervision of skilled/instructed persons (although in practice it's often used by all and sundry when it comes to plugging things into small portable generators, but that's a different story).

In this case things are still classed as ADS (under 411), but the principle of separation is employed only after the protective device has opened - as per 551.7.1 (ii) - unlike a truly separated circuit where there's no disconnection at all on 1st fault. It is slightly a case of juggling words (the physics don't care), but as we're not claiming that shock protection is by separation, the limitations of 413 and 418.3 don't apply (by my reckoning at least).

   - Andy.