Previously I wrote about the problems with connecting PV panels directly to an immersion heater in which I suggested using multiple immersion elements switched by relays to achieve an approximate match to the PV output as the sunlight varies.
A year ago I had some email correspondence with Mike Phillips (firstname.lastname@example.org) about designing such a system for use in Spain. Pleasingly, he's just reported that he's only gone and done it and has had a system working for a while.
He sent me a word processor document describing his system with encouragement to publish it here. Not being overly happy with the security implications of flinging WP documents around I've converted it to PDF as:
His approach of trimming a standard immersion down to length to get the desired resistance is intriguing. I don't think I'd have thought of that. I'm also not sure that it's that worthwhile for the small difference in resistance between 19 and 15 Ω but, on the other hand, since power at constant voltage is proportional to I² the effective difference (about 25%) might well be enough to be significant so, as elements aren't expensive, it's probably worth a try.
He does say, though, that he had to make mechanical changes anyway to get multiple UK 3 kW elements into a Spanish tank which normally only has a 1500 W element.
One thing I like about this design is that there's plenty of safety margin in the element's power capabilities. Other series/parallel switching configurations I'd sketched with lower powered elements (e.g., using ones intended for towel rails or just high-power resistors) could over power a single element if it was accidentally switched alone across the PV in bright sunlight due to a wiring or software fault. These elements have a huge margin even if exposed to the highest open-circuit output voltage of the panels.
Using high-powered relays has a certain clunky air of reliability. On the other hand, this switching seems well within the capabilities of power FETs. The series/parallel switching might seem a little more awkward but need not be too complicated:
|Left FET||Right FET||Elements|
|Off||Off||Both in series (through the diode)|
|Off||On||Right element only|
|On||Off||Left element only|
|On||On||Both in parallel|
There'd probably be the need for a couple more transistors between the microcontroller and the FETs - perhaps for the low-side FET to get sufficient gate voltage and, more likely, for the high-side FET to be able to withstand the high voltage on the gate. You could even opto-isolate them if you felt like it.
Depending on the temperatures involved it might even make sense to goop the semiconductors to the tank to a) heatsink them and b) make good use of the small amount of waste heat they'll produce.