In my previous Photovoltaics for Domestic Hot Water post I said I'd write more on the use, caveats and interpretation of that calculator later. A recent thread on the Navitron forum triggers me to write about some of this.
That calculator just takes into account the raw cost of the panels. The default prices given don't even include VAT (as that can be reclaimed for a new build). More significantly, they don't include the cost of mounting hardware, controllers, pumps, and so on. These are not necessarily proportional to the raw costs of the panels for the different technologies and are not proportional to the number of panels actually employed.
The Navitron thread suggests connecting the panels directly to the immersion heater with just a blocking diode. In reality, some sort of controller would be needed for a number of reasons which are discussed there but not explained very clearly.
Biff suggests the use of blocking diodes. Jonesy says they're not needed. I, too, think they're not needed unless you have multiple strings of panels in which case they're a good idea to prevent one string in sunlight backfeeding into another shaded string.
Some caution is needed if you're using a thermostat not specifically designed for this sort of application. Most thermostats, and other switch gear, is designed for use with AC and relies on the current disappearing every 10 milliseconds or so, as the 50 or 60 Hz AC current changes direction, to stop an arc being drawn. If used with a high current DC source (even one which is well within the current rating of the switch) the arc will tend to persist burning out (perhaps welding shut) the contacts pretty quickly which would be a bad thing.
Some sort of DC rated thermostat is needed. Perhaps a suitable DC rated relay or semiconductor switch driven by a low current signal through the normal AC immersion would be good.
Matching the Load
However, the big problems arise from matching the consumption of the immersion heater to the generation of the PV panels.
A normal 1 or 3 kW AC mains immersion is not likely to be be suitable. Typically, PV is run at significantly lower voltages. However, that's not a disaster. While not as cheap as standard AC immersions it's quite easy to get custom ones made for any required voltage and current combination.
The bigger problem is matching the operation to the output of the PV panel at a specific time. Suppose, to make the arithmetic simple, you have four 250 W panels in series, each with a maximum power point voltage in bright sunshine of 25 volts. The Vmpp of the array is therefore 100 V and Impp is 10 amps giving 1 kW of output.
To match that you'd need to get an immersion with 10 ohms of resistance and capable of dissipating the 1 kW (plus a bit more for when the sunlight falling on the panels is a bit over the nominal 1000 W/m² or the temperature is a bit lower than the nominal.)
Suppose, though, that the sunshine is now about half as bright as in the test conditions. The Vmpp of the panels will fall a bit but the main effect will be that the output current capability will drop roughly proportionally to the sunlight. Say the current available drops to 5 A. With that going into the 10 Ω immersion the voltage will now be 50 V and the power being transferred into the water just 250 W.
In other words, when the sunlight drops to half its brightest the heat getting into the water drops to a quarter. Though we've got nearly 500 W of power available from the panels (5 A times nearly 100 V) we'll actually only be getting just over half of that. The other half (another nearly 250 W) will be lost as heat in the panels.
I suspect that the best solution is to have multiple heating elements and switch on just a few of them to match the available power. E.g., instead of one 1 kW, 10 Ω immersion it would be possible to have four 250 W, 40 Ω elements. In bright sunshine all four would be turned on. In our approximately half-power conditions two would be in use. With a parallel resistance of 20 Ω the 5 A current would allow the panels to still operate at about 100 V giving the full nearly 500 W into the water.
Though the controller to achieve this could be pretty simple it, combined with the extra cost of custom immersion elements, would need to be taken into account if comparing systems, as opposed to just doing a rough comparison of the basic panels as my calculator page does.
Update 2013-08-08: it appears that commercial controllers to do the MPPT tracking between a small PV array and an immersion heater are now available in the US. I've come across two, here's the second: TechLuck. Seems reasonable to me though it only appears to do 500 W or so.