Solar PV into load side of RCD

I think I share lyle's thoughts. Presuming the RCD (as required by 551.7.1(ii)) opens all live conductors, including N, then once the RCD has tripped, all the downstream circuits loose their Earth reference - in effect becoming one big separated system. It may continue to have 230V between L & N (or even 400V between L1/L2/L3) but it should remain safe both during 1st L-PE faults and direct contact even without any ADS as there's no path to Earth (other than at the fault itself) - all recognised in 413 or 418.3 (in principle of physics at least).

I can see that in the case of multiple faults (e.g. a 2nd L-PE fault or direct contact where a latent L-PE fault exists) danger may occur, but that's calling up probabilities that are beyond BS 7671's usual requirements (i.e. to provide protection in the case of 1st faults only).

I'm wondering about the cases which are less clear cut though - i.e. multiple class I appliances having filters or less than perfect insulation that might mean that the circuit isn't truly separated from Earth once the RCD is open ... whether such paths could have a sufficiently low impedance to allow shocks with the RCD open, but not 'nuisance trip' the RCD in normal conditions, seems less clear.

   - Andy.

That is a good point and something I had not picked up on in the original post.

Inverters do tend to give rise to some earth leakage and the Code of Practice highlights nuisance tripping of RCDs as another reason for not sharing the RCD with other circuits. 

For systems in island mode, a system reference relay (earth-neutral bond) is required. I did come across an inverter that lacked such a relay so it would allow island mode operation with a "floating" earth. Under no-load conditions I measured a L-E potential > 160 V in island mode. Quite how quickly such a potential would collapse where a moderate impedance (i.e. a person) is put across the terminals is not clear to me either.  

That is a good point and something I had not picked up on in the original post.

Inverters do tend to give rise to some earth leakage and the Code of Practice highlights nuisance tripping of RCDs as another reason for not sharing the RCD with other circuits. 

For systems in island mode, a system reference relay (earth-neutral bond) is required. I did come across an inverter that lacked such a relay so it would allow island mode operation with a "floating" earth. Under no-load conditions I measured a L-E potential > 160 V in island mode. Quite how quickly such a potential would collapse where a moderate impedance (i.e. a person) is put across the terminals is not clear to me either.