Zero export is a design constraint, not a setting, and the control that makes it work sits at the point of supply rather than on the roof. A municipality can decline to buy your generation, but it has no reason to refuse a connection that never pushes energy outwards, so the real question stops being "will they allow it" and becomes "at the instant the house stops using what the panels are making, where does that energy go?" Answer that properly and you can run a substantial array behind a meter that has no export arrangement at all.
I keep a clamp meter in the bakkie for mornings like this one. We are at a house on the edge of a small town, the neighbour's pool pump humming over the wall, and the owner has already been told twice by installers that zero export "is not allowed around here". We open the meter box and stand in front of it together. The municipal meter is the old disc type. The inverter on the wall has been running since sunrise.
Watch the disc while I talk him through the morning. The kettle goes on, two thousand watts, and the disc slows to nothing. The inverter has already pulled its output down to match the house. The kettle clicks off, and the disc twitches. For a moment, less than a second, and how long depends on the hardware, the inverter is still delivering what it was delivering a breath earlier. Nothing in the house is absorbing it. So it either lands in the battery, or in the geyser element if that happens to be drawing, or it goes out through the meter.
That moment is the whole topic. Everything else about zero export is bookkeeping around it.
The decision is taken at the boundary
An inverter cannot know what a house is using by looking at its own DC side. It can estimate from its own output, and estimates drift. The only reliable signal is a measurement taken at or near the point of supply, compared against what the inverter is producing, with the difference used to pull production back. That is why a genuine export-limited arrangement is described by where it measures, not by what the datasheet says the inverter is capable of.
This matters when you read a quote. "Export limiting built in" can mean a real control loop fed by an external measurement at the meter, or it can mean a firmware mode that reacts to its own output and never actually looks at the boundary. Only one of those can be trusted where a municipality has refused export.
It also matters in a three-phase house. Most South African homes are single phase, but where the property is three phase and the loads are spread unevenly, limiting on one phase while the others import is not the same thing as limiting the property. I have seen a system that was beautifully limited and still pushing a trickle out on a phase nobody was watching.
Energy has to land somewhere in the same instant
Generation is not a tap you can close without cost. When the array is making more than the house is using, there are only three destinations: the battery, a controllable load, or curtailment.
Curtailment is the inverter deliberately producing less than the panels could. It is not a fault. It is the system doing exactly what it was told. But it is also money you spent on panels and did not get back, so a system that spends most of its clear days curtailed is a system that was sized wrong. I would rather see a smaller array with a battery that can absorb the surplus and a daytime load schedule that soaks up the rest.
That is the part most people miss when they first hear "zero export": the design is not really about the panels. It is about the battery's charge rate, the sizes of the loads you can move, and whether the house's base load during the day sits anywhere near the array's output. A battery that accepts two kilowatts while the roof is making six leaves a four-kilowatt hole, and something has to fill it.
Controllable loads are the third destination, and they are underrated. A geyser on a timer, a pool pump moved to the middle of the day, a heat pump, an EV charger set to run while the sun is high. These are loads you can summon when the surplus appears, which is exactly what the system needs at that instant.
What actually changes at application stage
Municipalities differ, and tariff categories within the same municipality differ. In some places a non-exporting installation follows a lighter route than an exporting one: a notification or registration with an inverter that appears on the approved list, rather than a full application involving a meter change and an export credit arrangement. In others the process looks much the same, because the municipality is concerned with the safety of the interconnection, not only with whether it is buying anything.
Two things do not change. The installation still has to comply with SANS 10142-1, and you still need a Certificate of Compliance. And the interconnection protection still has to meet NRS 097-2-1, which is about how the inverter behaves on the grid, not about whether it exports.
So when someone tells you the municipality will not allow solar, the useful response is to ask which document said so. A verbal no at a counter is not the same as the published application form. Read the form for your own municipality and work out which category your design falls into. Do not take my word, or an installer's, for it.
Load shedding does not change any of this. When the grid is down the inverter islands and export is impossible by definition. The question is what happens in the second the grid comes back and the inverter resynchronises. That is a design question about the control loop, and it deserves an answer before you sign anything.
Where a claimed limit quietly stops being one
The failures I have seen are almost never about the concept. They are about verification.
A limiter reading the wrong phase, or reading its measurement backwards, will hold the inverter back perfectly while the property still exports. A single-phase limiter on a three-phase supply will do the same on two of the phases. Firmware that corrects on an average, rather than quickly enough, will let short spikes through every time a big load switches. Induction hobs, inverter air conditioners and geyser elements switch in hard steps, and every step is a moment where the balance is wrong.
Then there is the load-drop test, which is the only test that tells you anything. You take the big loads off one at a time while watching the municipal meter and the measurement display at the same time. Geyser off. Pool pump off. Oven off. If the disc never runs backwards and the display never goes negative, the limit holds under the loads you actually have. If nobody has ever done that in front of you, the system is unverified, whatever the paperwork says.
I have also seen systems where the measurement device fails or its cable is damaged and the inverter simply carries on producing as though nothing were wrong. That failure mode is worth asking about directly. What does the system do when it loses its measurement? The answer you want is that it stops exporting, not that it guesses.
The trade-offs worth saying out loud
Zero export does not mean free electricity. It does not mean no application, and it does not mean you escape fixed municipal charges. You still pay for everything you import, you still pay the connection and service charges, and you still need a compliant installation with a CoC. What you get is permission to generate for yourself where an exporting connection was refused, and the ability to keep the grid as your backup supply.
What you give up is any possibility of earning from your generation. If your municipality does offer a feed-in credit and you are willing to do the paperwork and the meter change, exporting is a different design with different economics. That is a legitimate choice, not a lesser one, and it is a different conversation from this one.
The disc sitting still while the sun is high is the point. Getting it to sit still during the kettle, and during the second after the kettle, is the engineering.
