Recently several companies have begun to produce peak power tracking (PPT) PV controllers for terrestrial systems. These companies include Fire, Wind and Rain and RV Power Products.
The basic concept has been around for a long time. I first heard of it when AMSAT put it in the ill-fated Phase III-A satellite launched in May of 1980. It's based on the fact that the current you can draw from a solar panel depends on operating voltage, and that by varying the operating voltage you can maximize the power you extract from the panel. You can see this property in the "current-voltage curves" included in nearly every spec sheet for a solar panel. Here is the spec sheet for the Astropower AP1206 panels I'm using.
As you can see, at a low operating voltage the panel output current will be high but the output power will be low, since power is the product of voltage times current. Similarly, at a high operating voltage the current will be low, so again the output power will be low. Somewhere in the middle is the panel's peak power point, where the product of voltage times current is maximized.
The peak power voltage isn't constant. It depends primarily on one thing: cell temperature. This in turn depends on both ambient temperature and solar insolation, because the sun will always heat the panel above ambient temperature. Lower cell temperatures mean higher peak power operating voltages and more power available at the peak power point.
As an aside, most PV panel power ratings are optimistic for a very simple reason: they rate their panels at 1 kW/m^2 solar isolation and at 25C cell temperature. Only on a very cold day will a PV cell remain at or below 25C when exposed to 1 kW/m^2 (which corresponds to a sunny cloudless day). In a place like coastal San Diego with its near-paradise climate, the cell temperature under 1 kW/m^2 insolation will invariably be much higher than 25C, so the actual power available from the panel will be less than the rated value. Things can obviously get much worse in the inland desert areas in the summer.
In a PV system without a peak power tracker, the solar panels are usually forced to operate at the battery voltage. This is almost always below the peak power point of the PV panels, so some of the power-generating capability of the PV panels is being lost. This difference is the potential gain with a PPT.
The question is, exactly how much can be gained? To figure this out, you need to look at how your panels perform under actual operating conditions, i.e., you want to see real voltage/current curves under actual solar insolation and cell temperatures.
There are several ways to measure panel performance. One way is to hook up a big rheostat (big enough to dissipate the full panel output) as a load and measure the voltage across it and the current through it as you vary the rheostat. This is the method Home Power Magazine uses in their PV evaluations, and it works well -- if you have the big rheostat.
I use two different methods that work with the equipment I already have -- my batteries and my inverter. In the discussion that follows, the voltages are for my 48V system; be sure to scale these appropriately if you have another system voltage.
In each case, I start by setting the high voltage cutout relay on the panel so that it will remain closed throughout the voltage range of interest, instead of interrupting the array output at its usual setpoint of 58V. Second, I hook up my multimeters to read panel voltage and current. Then I vary the panel operating point over the range of interest in one of two ways: by varying the SELL VOLTS DC setting on my Trace SW4048 inverter as it sells the PV power into the grid, or by shutting the inverter off and watching as the panels drive up the battery voltage. In each case, I record readings for each voltage and current pair.
The first approach has the advantage of being able to hold the operating point at a specified voltage indefinitely so you can make your reading. The problem, though, is that the Trace SW inverter can only maintain one of a small set of DC/AC voltage ratios, so it will often hunt every second or two +/-500mV or more around the specified SELL voltage. Constantly changing numbers on a digital voltmeter can be hard to read.
The other approach has the advantage of providing smoothly varying numbers, but they can change pretty fast in certain regions. I normally float my 220Ah battery at 54V during the daytime, so if I start by discharging it at 25A or so for a few minutes, e.g. by reducing the inverter SELL voltage so it begins to discharge at the desired rate, then when I turn off the inverter to begin the test the battery voltage will rise only slowly as the panels bring the battery back up to full charge. Once this happens, the voltage will start to rise somewhat more rapidly until it hits the gassing voltage of about 62V, when the increase will slow again. (I don't leave the system in this state any longer than I have to to get my readings).
Here are some typical readings I took in mid-March 2000 on a sunny but slightly hazy day around local noon. At the time, my array consisted of 12 Astropower AP1206 panels wired as three strings of four panels each. I used the "turn off the inverter and watch the voltage rise" method:
Two things to note: the peak power point occurs around 58V, and the peak is very broad, i.e., the power doesn't change very much for several volts around the peak. The numbers were not totally consistent because the solar insolation was varying slightly due to haze, so I ran the experiment several times. Although the specific numbers would change a little each time, each time the conclusions were the same: under these specific conditions, my panels were putting out about 30 watts less by operating at my battery float voltage of 54V rather than at their peak power point.
So 28W, or about 3.3%, is the most I could possibly gain under these conditions by installing a peak power tracker. I'd gain even less than this because no PPT is 100% efficient. Not much of an improvement, huh?
The Fire, Wind & Rain model 10302 PPT costs $450, which would be $16.07 per extra watt if it were 100% efficient. FW&R rates their unit at 94-98% efficient, so this would be more like $16.40-$17.10/W. But if I just bought another Astropower AP1206 panel at $549.75, I'd get another 859/12 = 71.6 watts, at $7.68/watt. (I pay $2199 for a box of four AP1206s, and I currently have 12 of them in my system). (Yes, I know I'd have to buy more panels in increments of four. But the idea here is to find the incremental cost per watt, so the numbers are valid).
On August 10, 2000 at around 1220-1230 PDT, I repeated the peak-power experiment (having upgraded my array from three to four strings in July). It was a clear, sunny day and the ambient temperature was about 80F. Here is the data in graphical form. I began by increasing the operating point from about 48V to about 62, then I decreased it again to about 54-55V. I attribute the small dispersion of data points around 55V to the small increase in solar insolation that occured during the experiment.
As you can see, peak power occurred at an even lower voltage than in March because of the warmer ambient temperature.
Battery selling requires that you lower the DC operating point of your inverter so power will flow out of the battery. For my system that's a decrease from 54V (normal float voltage) to about 48V. Since peak periods invariably occur during the day when your PV panels are producing power, this reduction in array operating point will decrease the array output power. A PPT could recapture this loss by allowing the array to again operate at its peak power point.
15 March 2000, Phil Karn
Updated 11 Aug 2000