The error message was "Earth Fault". Since the supply from my turbine is not intentionally grounded this meant a possible fault in the turbine, or possibly on the cable coming from the turbine to the inverter; fortunately the inverter continued to operate so I felt in no hurry to get to the root of the problem.
In this situation, SMA's trouble-shooting guide instructs you to exclude a genuine earth fault by inspection and testing. If this first-off approach doesn't reveal what's wrong, the next step is to test the 2 varistors housed within the inverter because, the guide says, a failed varistor can cause the Earth Fault warning to be displayed.
In an SMA inverter, the varistors look like this:
The trouble-shooting guide says to test for continuity between B and C, ...if there is no continuity the assembly needs to be replaced; if continuity exists, ...look for a fault elsewhere in the inverter.
It was only sometime later, after I had removed the insulating shroud and seen there were two components beneath it, that I realised testing in this way only tests for continuity across the thermal fuse; it cannot test the functionality of the varistor itself, which is the blue disc-shaped component.
When I did the continuity test, both the varistor-cum-fuses showed continuity, and so I concluded, as the SMA literature had led me to believe, that they were OK. But a week later, having exhausted all other possible causes of the fault condition, I replaced both with new ones and, hey presto, no longer was Earth Fault displayed: I had got to the bottom of it 😊. (** - but see post script below)
So what's the 'science' behind all this ? The purpose of a varistor is to eliminate voltage surges which might damage the equipment the varistor is protecting; they are 'sacrificial' devices, meaning they can be destroyed by the excess energy they absorb, and they are also 'wear' components, meaning they gradually lose their function from the cumulative effect of absorbing energy from lesser voltage surges which are not great enough to destroy them.
In extreme situations, varistors can catch fire, either because of the magnitude of the energy passing through them or because of the duration the energy flow exists for. In this situation they pose a fire risk which might destroy the very device they are meant to be protecting and to mitigate this, a thermal fuse is sometimes included in series with a varistor, - a fuse which will 'blow' and terminate the supply through the varistor if its temperature exceeds a set point. This is the arrangement of the two components found in what SMA call 'their varistors'.
SMA state that 'their varistors' are specially manufactured and are not commercially available, - except, of course, from SMA. The cost (in 2017) is €15 plus shipping and VAT (total €29.75), for a pack of two (part code SB-TV3, with insertion tool).
The two components that make up 'their varistor' can however be found on the open market, and with a soldering iron to connect them in series, they can be made up more cheaply. The varistor is an Epcos S20K320 (Vrms 320v, VDC 420v, Imax 8000A, Wmax184 joules Pmax 1 W) and costs 0.55p from Farnell (order code 100-4305).
The thermal fuse is more difficult to find but I managed to track down 6 on eBay: it is a Tamura E3F 250v, 3A~, 115℃ and each one set me back €2.20, inclusive of p&p. But it can also be purchased direct from China for less than this.
Why should the varistors have failed in my inverter to cause the Earth Fault warning? I figure that with 12,492 hours of continuous operation, at ~300v DC, the natural 'wearing' process going on in a varistor is accelerated and probably accounts for their failure. After all, an inverter handling power from a water turbine operating all the time sees a lot more use than one connected to a wind turbine or PV array.
12,492 hours is about 17 months and so I'm thinking this is the interval at which I should expect to see the Earth Fault warning recurring, - and therefore that I should have enough stock of the varistors and thermal fuses to meet that sort of replacement frequency.
It was nice to have gained a little more understanding of how the technology was designed to work.
** note added 3rd September 2017: the fix turned out to be not long lasting. Within 2 months, the inverter was displaying Earth Fault again, initially intermittently and then permanently. Replacing the varistors yet again, with new SMA ones, failed to get the fault light to extinguish. I conclude that either there is truly an earth fault somewhere or the sensing circuitry within the inverter has gone wrong. I think the latter more likely. It being an issue which does not affect performance, I plan to do no more about it.