STEG - solar thermo-electric generator

The neat thing about this is that it gets around one of the major limitations of solar photovoltaic panels, that they only work during daylight. If this thing is properly designed, it should generate electricity 24 hours of the day.

None of this is secret, in fact when I am finished building and testing it I'll publish all data on the net for anybody who wants to build one too. I don't think this is patentable -- I hope not, because I think patents now impede progress. The technology is quite old and dates back to the 1800s.

It uses the thermoelectric effect, but approaches it a slightly different way than usual.

It is easiest to understand by starting at the bottom of the diagram.

A solar heating panel is filled with water. Its bottom has a well-insulated pipe leading to the bottom of a well-insulated hot storage tank above and behind it. The top of the panel is connected by a similar insulated pipe to the top of the same tank. This is a closed-circuit water heater. Water warms up in the panel. Warm is less dense than cold water so it rises into the top of the hot storage tank and the colder water in the hot storage tank falls down into the panel where it is warmed. The water circulates like this all day, driven simply by convection. At night the water won't circulate because the convection is driven only if the panel is warmer than the tank, so no valves are needed.

Above the hot storage tank is another well-insulated tank, also containing water. This is the cold storage tank. It is connected to another panel above and behind it, with the top pipe leading to the top of the panel and the bottom pipe leading to the bottom of the panel. This panel is not a solar collector though. It is a radiator. It has no glass covering and it is aimed away from the sun, with the radiating coils and fins exposed to the air. This panel comes into play at mostly night. The water in the panel cools and the cold water falls down into the bottom of the cold storage tank, while the warmer water rises from the tank to the panel to be cooled. Once again the water circulates by convection. Again no valves are needed -- if the radiating panel gets warmer than the cold tank then circulation stops.

The temperature difference between these two tanks generates electricity using a thermoelectric device (also known as a thermopile). The top of the hot storage tank holds the hottest water and the bottom of the cold storage tank contains the coldest water. Between the tanks is either a zig-zag of paired metals that generate electricity when their ends are at different temperatures (cheap but fiddly and low efficiency), or else semiconductor peltier cooler modules can be used. They can be bought quite cheaply nowadays. Not many people realise that the coolers also work in reverse. Instead of running a current through them to generate cool on one side and heat on the other, if you heat one side and cool the other and the device generates electricity.

A single Thermoelectric (Peltier) module from Jaycar Electronics cost me just AU$23. I notice the device has "TEC1-7108" stamped on it (the maker's designation, not Jaycar's), however I believe it is actually a TEC1-12708 based on data from various places on the web. The device measures 40mm x 40mm x 3.5mm and is supposed to deliver 68.09 Watts cooling power (whatever that means) at 8 Amps and 15 Volts input with 27°C difference across it. (See more data at I'm not sure how that information will translate to using it in reverse. I'll post my results as I experiment with it.

Here is a more detailed svg diagram of the generator. If you use Microsoft's crappy (and insecure) InternetExplorer you might not be able to view svg images. My advice is to get a better web browser, like one of the Mozilla browsers (Firefox or Seamonkey) or Opera. I think Safari (on the Mac), and Google's Chrome display SVG too.

There already are some thermoelectric generators on the market, but the designers seem to get caught in the trap of thinking only about heating. The thing is, it is not the heat that does the generating, it is the temperature difference. Also I think people get too focussed on high output. They burn fuel to heat one side of the device and build complicated pumps with coolant and thermostats to prevent the device being damaged by the heat. Using a simpler design avoids a lot of the problems and reduces the possible points of failure of the system. Using convection does away with water pumps and using a radiator to capitalise on cold night air lets you generate electricity from the difference between the hot water tank and the cold water tank -- it is actually generating electricity from the difference between daytime and nighttime temperature. Because it already stores energy as heat (actually heat difference), this system should work without electric batteries too, eliminating another great expense and common point of failure in solar power systems.

The Earth swings around the sun in its yearly orbit. The Earth also spins on its axis each day. This axis is tilted by 23.422° so that at one side of the orbit the southern hemisphere is in summer and the northern hemisphere is in winter. Then at the other side of the orbit the north is in summer while the south is in winter. At the height of summer for each hemisphere you will notice that the sun is directly overhead of the Tropic of Capricorn or the Tropic of Cancer (the two dashed orange lines). Those two tropics are at 23.433° angle from the Equator. The green person and the blue person are at 33° north and south of the Equator.

If you lived on the Equator (that is, at 0° latitude) then, over the year, the sun would average out to being right overhead, so a solar collector placed flat on the ground would pick up maximum sunlight year round. We must try to point our solar collector in roughly that same direction wherever we are on Earth. You can see that if we are at some number of degrees away from the equator then we must tilt our panel back toward the equator by the same number of degrees to make it point in the same direction as one flat on the ground at the equator -- that is, we want it to be parallel to the collector at the equator.

Orienting the solar collector

During the seasons the sun appears to rise and set at different angles. This is because our planet is tilted on its axis by 23.433° with respect to its orbit around the sun. But this doesn't much affect how we point the panel because we want to average out seasonal changes (mostly). If you don't have a tracking solar collector then you will want to point the panel in the same direction as a horizonal panel on the Earth's equator. Your latitude on the Earth is given as an angle from the equator to the center of the Earth and to where you are. Conveniently, this also happens to be the angle of tilt from the horizonal that you should position your panel. Of course you should aim the panel northward if you're in the southern hemisphere, but southward if in the northern hemisphere.

Earlier I said that you want to mostly want to average out seasonal changes. I would suggest aiming the panel a little more steeply than your latitude because it is during the winter when heat is most difficult to collect from the sun. The further from the tropics you are, the more important this becomes. In cold latitudes it might be worth aiming it almost 23° more steeply to point more directly to the winter sun. In more temperate latitudes 10° or less might suffice.


There are probably more modifications I haven't thought of yet.


My main focus at the moment is on building a simple prototype and the electronics to measure and log the performance. To do this I'll need to log temperatures of the two storage tanks at the points most useful to the thermoelectric device. It might also be useful to keep a record of environmental temperatures too so that weather can be evaluated against its effect on the generator. I also need to record the electrical output of the thermopile.

I'd originally intended to use an ADC0809 analogue-to-digital converter with some thermistors to monitor temperature in the two storage containers and environmental temperature, however Alina suggested using DS1820 devices. I'd heard of these before but had thought them too difficult to use because they use a serial interface and I had no idea how to program Linux to read them. Alina pointed out a wonderful project at which not only gives the circuitry but also software to do the job.

I'm still thinking about the best way to record the electrical output of the device.

updated 2010-08-29