After the wetness of mid November, things have turned dry again, so I was in the situation of still having the largest nozzles in the turbine, delivering the design flow of 3 lps, with a supply of water which was slowly diminishing.
Inevitably, a time came when water inflow to the tank fell below outflow down the penstock to the turbine, and instead of overflowing, as is the normal operating condition, the tank gradually began to empty.
The dimensions of the tank are 2.74 x 1.52 x 1.37 metres but since the bottom 83mm cannot be accessed, the useable volume is 5.3 m³. With this generous volume of reserve, and only a small difference between inflow and outflow rates, it actually takes several hours, more than 12 on this occasion, for the tank to drain down. Somewhat inconveniently this meant all the action was set to happen after dark.
Here is an aerial picture (actually taken from up a tree) of the header tank and the arrangement for gathering water into it:
And here is the record of power output from the inverter into the grid over the period when the tank was emptying. The record was captured using Wattson power monitoring and the data downloaded to my computer using the Holmes software that Energeno provide free for this purpose:
If anything, the system 'fails safely' because the dc voltage falls to a lower operating level than normal (in this event it was 265 v dc at its lowest) because with the lower 'head' created in the penstock, turbine rpm will have decreased marginally. Against this effect, there will have been a counter tendency to raise voltage created by the inverter lowering the load it places on the pma. That the overall result was a drop was interesting. It was not obvious by observation and listening that the turbine rpm was very different from usual.
So, all in all, an exercise which was worth doing even if it did give me a sleepless night !