The first panel was commissioned on 03 May and the fifth on 20 June. Operational data was collected on a daily basis from the time the first panel was in place, and so at the time of writing we've exactly six months' worth of data.
This summer has been very much a mixed bag in terms of weather - May wasn't too bad, June was very poor and the beginning of July only a little better, but since then it's been more typical of what we can generally expect around here.
Graph 1 below shows the total cumulative energy from the array since the time of hooking up the first panel. The curve starts off rather shallowly and then gradually steepens as more panels become live, but the gradient is not as high as expected since the weather worsened through the commissioning period.
Note that all of these graphs shown here represent net AC energy provided to the house grid, i.e. all the losses and inefficiencies in the grid-tie inverter during rectification and transformation from DC to 240V AC are already accounted for. The light red line on the graph is a linear trend line.
(click on any of the figures for a larger image)
Graph 2 shows the same data on a daily basis. Where several adjacent bars have the same value, it's because we couldn't take readings every day for whatever reason and therefore the measured output has been averaged across the periods we weren't around. This is particularly pronounced in September when we were away on holiday for a few weeks. The horizontal red line is the mean value.
Graph 3 shows the energy produced on a per-panel-per-day basis. The horizontal red line is again the mean value.
And Graph 4 shows the energy produced on a per-installed-Watt-peak per day basis. This 'unitised' data produces the most useful Key Performance Indicator (KPI) for the installation, since it allows us to predict how much energy would result from any size of solar array at our particular location, simply by multiplying the values by any total number of installed Watt-peak capacity. The horizontal red line is again the mean value.
For reference, we can see from our monitoring equipment that the maximum instantaneous AC power produced to date by our 540 Wp system is 434W. Assuming that the rated output of the panels is actually achievable for very short periods at our latitude and roof angle, then we think that the difference between rated and observed power can mostly be attributable to the efficiency of our grid-tie inverter.
Additionally, the cell conversion of light to heat is very inefficient, and means that most of the solar radiation is actually absorbed as heat, and the cell efficiency falls as they become hotter. Our particular panel design does not allow this heat to be dissipated very well, and so on warm sunny days the maximum continuous power output we've observed is around the 300W mark... cold sunny days will allow the panels to operate more efficiently, but of course in winter the sun is much lower in the sky and also less intense.
We have two unit costs for our panels - those three encapsulated with Q-Sil cost £91 each to make, and those two with the cheaper Polastosil compound (from Poland) were £78 each. Therefore, the total cost of five panels was £429. Add the cost of the grid-tie inverter at £234, and another £150 for the roof frame materials, cabling and power cabinet components, then let's say our 540 Watt-peak installation cost a total of £813.
Note that we actually spent more than this - as I said in Part 1, we bought a 1 kW cell kit so we still have cells available to make some further panels. The costs of the panels therefore include a pro-rata cell cost based on the Watt-peak rating of each.
This £813 equates to £1.51 for each Watt-peak of installed capacity. If we compare this to an installed capacity of £1.70 per Wp for a commercial system (a fully-installed 2.5 kW system is currently on offer on eBay for £4,250), then we're around 10% cheaper but we've none of the economies of scale to be expected from constructing a larger system. For example, we could add another two or three panels and the system would still be within the capacity of our existing 1,000W grid-tie inverter, and this would bring the installed capacity per watt-peak significantly down.
To look at the economics of our particular situation, we first need to establish some benchmarks of our expectations. We have a few sources of relative data for comparison.
One, from a Dutch website, predicts that each Watt-peak of installed photovoltaic generating capacity will produce around 0.85 kWhr per year on average - although even the most northerly part of the Netherlands is still south of our location, it's not a million miles from our latitude and therefore considered to be a useful benchmark.
If we calculate the expected annual energy produced from our 540 Wp array using this value, then we get a figure of 459 kWhr per year. Note however that this is total energy from a photovoltaic array, whereas we are measuring AC energy after DC-AC conversion and transformation. Let's believe the manufacturer's claim that the peak efficiency of our grid-tie inverter is 92%, but of course this is peak, and so let's also assume that a more realistic average efficiency within our installation is say 80%. This correlates well with the 434W maximum output we've seen so far from the panels. This then produces a benchmark figure of 367 kWhr per year.
Another benchmark source is from a UK academic website, which gives the expected 'insolation' for both London and Edinburgh. Insolation is simply a hybrid word meaning 'incident solar radiation', i.e. the unit amount of sunlight energy available at a particular location, and is expressed in terms of kWhr per square metre. The data is given in the first few lines of the following spreadsheet.
Note two things here - firstly, this relates to sunlight falling on a horizontal surface, and secondly this is the total amount of radiated energy available from the sun before any energy conversion inefficiences are considered, i.e. when changing the form of energy from light to electricity.
For this data to be useful to us, and by using the latitudes quoted, we interpolated the values at our particular latitude which lies someway between London and Edinburgh. We also assumed a conversion efficiency of 12% which is considered typical of photovoltaic cells (this of course is a very low number, and we've carried out other experiments on thermal solar panels in which we've seen conversion efficiencies of more than 60% - see our earlier post on the subject ....)
By adding further rows and data to our spreadsheet, we've taken the raw information provided in the source to predict an annual electrical energy collection figure of 400 kWhr per year. Again using our inverter efficiency of 80%, we should therefore produce 320 kWhr per year of AC energy from our array.
We now have data from two independent sources that predict we should expect somewhere between 320-360 kWhr of AC energy per year from our array.
So, how are we doing ? Well, from 20 June when all five panels were first operational, to 20 September, we produced 101 kWhr of AC electrical energy. This is almost exactly three months, and so all things being equal if we multiplied this by four we should get a rough idea of our annual output. However, all things are not equal - this time represents almost exactly the period from the summer solstice to the autumn equinox. If we assume it's also representative of the period from the spring equinox to the summer solstice, we can double this value to predict the energy produced during the period when the days are longer than the nights. However, we're now well into a downward trend when the days are shorter than the nights, and on 21 December we can expect less than 8 hours of daylight, and with the sun much lower in the sky.
So, if we can say from our measured data that we'll produce 202 kWhr during April through September, we need to look at the above spreadsheet again to estimate the output for remainder for the year. From the row marked 'Output per Month', if we examine the values we can calculate that the predicted energy during October to March is only 26% of that for April through September. On this basis, the trend for our array is to produce only 254 kWhr per year, some way short of our benchmarks.
At this stage, without measured data for a full year, we cannot identify why we seem to be falling short, so we'll post an update in the middle of next year. In the meantime, we'll use our predicted figure of 254 kWhr in an economic analysis.
If we assume that the unit value of our AC energy is £0.14 per kWhr, a reasonable figure from our electricity bills and including VAT, then we are generating electricity from our array worth around £36 per year.
On the very simplest analysis of the current data, £36 / £813 is therefore a 3.2% annual return on the initial investment and would take 22 years to recover these costs. Working on the almost certain assumption that domestic electricity prices will increase faster than both inflation and interest rates over the coming years, this would improve the situation slightly but is still a disappointing return. Again on the downside, the output produced by the array will also reduce over time as the cells degrade - as they age, the power falls away.
Even if we were to achieve our highest benchmark output from the array, i.e. 367 kWhr per year, then this would only be £51 per year or a 6.3% annual return with a 16 year payback period.
So based on the available data at the present time, and looking at it purely from a financial point of view, our experimental solar array would appear to be a poor investment.
We appear to have merely confirmed what seems to be common knowledge, i.e. without the benefit of greatly enhanced and government-subsidised feed-in tariffs, solar microgeneration is not a financially viable investment....
However, as I said, we'll keep collecting operational data throughout the winter and spring and review the situation again in the middle of next year.