I'm not really an expert on this topic, but I'll share what I know.
AC-connected storage is treated as a generator 'connected in parallel with a distributor’s network', just like a solar inverter, since that is how it is connected and behaves - you may not be intending to use it for export, but the device itself is connected in parallel with the network and you are just controlling the output so that it approximately balances the consumption elsewhere in the house. So with solar you have two generating devices connected.
DC-connected storage (ie. connected behind a common inverter used for the solar or the battery depending on the situation) would be different: the network can't see the storage in that case and it's just acting like a solar installation that works even though the sun isn't shining. DC-connected storage used to be at an artificial disadvantage in the UK due to FIT payments: with the storage connected on the DC side and the generation meter connected on the AC side, all the losses through the storage process subtract from your FIT payments. It also doesn't straightforwardly allow import from the grid, so your storage becomes strictly solar timeshifting rather than also being able to do tariff arbitrage with it. Nonetheless, some DC-connected systems are available, and indeed Powerwall 1 was advertised as available in AC- or DC-connect versions. Powerwall 2 seems to be mainly advertised as AC-connect, though it is possible a DC-connect version exists as well.
Anyhow, assuming AC then it's another generator, and the law (in the form of the
ESQCR) makes a hard distinction between systems below 16A/phase which merely have to meet technical requirements and then inform the DNO when you connect them, or systems above 16A/phase where you need to agree "specific requirements" with the DNO. These regulations date from 2002, so probably didn't have domestic storage particularly in mind, but are written in terms of "a source of energy" so that they cover everything. One implication of this (I believe - I haven't looked up chapter-and-verse) is that any reinforcement work required on the network has to be paid for by the DNO for the <16A systems, while for >16A systems they can make you pay for a share of it.
Then you have the ENA documents that codify best practice and are used to show how you have complied with the rather open-ended requirements in the law.
G98 (previously G83/2) is the technical specification for the <16A devices that are permitted to be connected with just post-notification. In particular, it defines the type-test procedures that the manufacturers have to perform and issue test certificates so that installers know the equipment is permitted to be connected. They must now also record their test results in a register of tested devices maintained by the ENA. Note that the rules allow more than one item to be connected if the total output is less than 16A (as might typically be the case with solar PV using microinverters). It is permissible to have the equipment software-locked to a lower output than its physical capability,
'provided these settings are not accessible to the Customer'.
For anything that doesn't fit into that there is
G99 (used to be G53) - which applies to anything from your little bit of solar PV up to huge power stations! However, there is a special simplified procedure for '
Integrated Micro Generation and Storage' where the individual units meet G98, the total doesn't exceed 32A, and there's an export limiting device to limit export to 16A.
G100 is the spec for export limiting devices, including both the
Integrated Micro Generation and Storage case and more general cases of larger systems, for which it provides formulae to calculate how much generating (and consumption) can be installed. This is based on the export control taking up to 5 seconds to respond, and ensuring that bad things don't happen during those 5 seconds. So you can't have an arbitrarily large installation even if you have the load control.
Most of the 'ordinary' requirements in these documents are simply references to standards (eg. EN61000 for harmonics) that any CE-marked equipment would be expected to conform to: the stuff actually defined in G98 itself is mostly about when the unit must (or must not) shut down for safety reasons in response to voltage or frequency deviations of different types. The key safety feature is 'anti-islanding' - ensuring that if the DNO's supply is switched off (either by a fault or because they deliberately want to work on the line) your generator doesn't continue energizing the network and electrocuting the DNO staff trying to repair the fault. This is not as easy as it might first appear, since if you have generation and load that happen to be well balanced (or you are deliberately export limiting to keep them in balance) then there might not be any current flowing in the DNO connection and so no immediate effect if it is cut off.
That much is the facts about what is allowed, as to why things are like this I am now speculating without solid sources or personal knowledge, so the following may not be entirely accurate.
The main reason why you have to count PV inverter + powerwall against the 16A limit (rather than just considering the export limit setting) seems to be this assumption that the export control will have a time lag - up to the permitted 5 seconds. So you then ask whether that's a reasonable assumption - could you make an export control that responds in milliseconds rather than seconds? At first glance that seems quite easy - many inverters nowadays are software driven (ie. the 50Hz sinewave is generated by software almost like a big DAC), so you can generate any output waveform you like and the software is recalculating it at sub-1ms intervals already. The snag is knowing what you want to do - if the sensor tells you (moment by moment) that the export current is increasing, you don't know whether that's because the PV inverter has increased its output, or the supply voltage has dropped due to a neighbour switching loads, or the grid frequency is shifting etc. I suspect the end result is that you can't do it in your primary (fast) control loop, you have to have a primary control loop that does the ordinary 50Hz behaviour and then a secondary control loop operating over multiple cycles to implement the export control.
Of course if the powerwall and the PV inverter were directly connected and knew what each other were doing at a very low level then you could do it - but now you've made them effectively into a single piece of kit that has a 16A output, and that's already allowed. It's only when you split them into two boxes working indirectly via an export sensor that the difficulty arises.
As to why there's the 16A limit in the first place and why generally the DNOs are more restrictive on generation than they are on extra demand, there's a number of factors at play:
- Obviously the 16A limit itself is an arbitrary round number for regulatory purposes, picked such that it's big enough to be actually useful but small enough that one or two systems suddenly installed on an individual supply main are unlikely to cause immediate trouble. Obviously we as consumers would prefer it larger, but indications are that even the 16A level is causing trouble to some DNOs.
- One of the major sources of trouble is overvoltage at times of light load in the daytime. The DNO is required to deliver to every customer 230V+10%-6%, so between 216V and 253V at all times. The way this is usually managed is that the transformer serving a group of customers is set to deliver the full 253V under no-load conditions, with the voltage then drooping (mainly due to the cable resistance/impedance) as you get further away from the transformer and as the load increases. So if the network has been built like that, as soon as the amount of PV generation exceeds the amount of actual load, then the 'droop' is in the opposite direction and the voltage at the customers will be greater than the legal limit. Since PV is generating during the middle of the day, and residential demand is often very low at that time, it doesn't take very much PV for the problem to occur. If you bear in mind that networks are planned on the basis of average demand per house of only about 2kW, PV at the rate of 3kW per house almost guarantees trouble if everybody installs PV. And they can't just tap down the transformer to less than 253V to give a bit more headroom, as it's still got to work on winter evenings when there's no PV output and load is maximum: unless there was already spare capacity in the network (ie. the worst case droop didn't go down as far as 216V). So in effect, adding PV to a residential network eats up extra capacity - either the network was already at maximum load and adding the PV broke it, or else the network had spare capacity and adding the PV caused you to tap down the transformer and now you haven't got the spare capacity you used to have for adding more houses, supporting EV charging or whatever. So a customer adding generation capacity costs the DNO money even if the generation power is lower than the notional rating of the supply . There has been discussion about relaxing the legal limit on voltage to help with this. It was originally intended to go to 230+/-10% for European harmonisation, but the UK stuck with +10%/-6% out of concern for existing appliances not tolerating the lower voltage; with so many modern appliances being electronically controlled and/or manufactured for EU-wide use it is arguable that it would now be OK to go to +/-10% and so allow a bit more headroom for PV (we'd see the 'typical' voltage drop a bit to allow room for the generation to push it back up again).
- DNOs are concerned with fault current, both for their own equipment and to tell consumers what grade of switchgear they need to install to withstand the worst-case fault current under a heavy short circuit. As an aside, domestic consumer units are typically built up out of MCBs etc with a 6000A maximum fault rating, but the overall CU has a 16,000A fault rating - essentially, the box guarantees to catch the pieces of the exploding MCB once you go above 6000A. DNOs have to publish the maximum fault current their network can deliver, so that you know it's safe to use a standard CU and you don't need to go for something more industrial. Typically, for domestic connections DNOs just make a blanket statement that it's "less than 16kA", and they can do this because they can calculate the maximum fault current a given transformer design can deliver, allow for the effects of a few metres of cable, and then know that the worst case at any customer can't be worse than that. However, once you've got extra generation on the network, those generators can supply extra current in parallel with the transformer. The short-circuit current of an individual powerwall is probably quite small, but put a hundred of them in the area supplied by a single transformer (think one housing estate) and it can add up. If you've got powerwall + PV inverter, then that's (roughly) double the extra fault current compared to one or the other.
- System stability is a concern. This embedded generating equipment (inverters) has to shut down if the voltage or frequency goes out of spec, in order to implement the anti-islanding protection (among other reasons). So if there is an overload situation causing the voltage to droop, all the embedded generators will shut down - causing the voltage to droop even further and lead to a cascading failure. A taste of this was given in the big national power failure on 9th August: as a result of two generators failing simultaneously, the grid frequency drooped and as a result a total of 500MW of embedded generation across the country (so that's everybody's domestic PV nationally and some larger scale systems). The total of the big generators that actually failed was 981MW; the total reserve capacity was approximately 1000MW, so the loss of embedded generation might have been critical.
So all in all, if you want a really big domestic PV+storage system it's probably best to plan it with at least part of the storage DC connected, or perhaps make part of the system 'off grid'. Planning for more than 16A (3.3kW) of PV to 'float' across the grid into the storage is tricky (though you can treble that if you have three-phase).