PVGIS predictions over the year, on BDPV do they seem right?

[12TA]

When I took a punt and installed a new solar PV system this year I was keen to check the performance. Actually I found this quite difficult. Sure it works, but did I get as good a system as I thought I was? My system did not come with any telemetry. ! had to take measurements, but how to compare them? PVGIS is an obvious source and I found the DBPV http://bdpv.com/index_en.php site through this forum.
This site lets you upload monthly data and compare with other users, they also use PVGIS as a reference. Therefore there is a standard curve for expected insolation over the year. My question:- Is this curve the correct shape? I notice a tendency for people to exceed the predicted in summer but fall short in the winter. November has been a great month for sunshine but it has been disappointing versus PVGIS is this just down to the pitch of roofs or a bug or some other factor?
Paul

[nowty]

I have noticed this undershoot in winter too, I checked several peoples stats on pvoutput and found their winter stats to be well below the theoretical PVGIS estimates. I personally think it’s a shading issue. As the sun gets very low in winter more and more obstacles get in the way and result in an increased cut off or interference both early and late in the day. Whilst the theoretical estimates assume a perfect flat horizon. As the yields are low in winter anyway, any increased shading gives a disproportionately higher reduction.

[JohnS]

Certainly in my case I have a shortfall compared to PVGIS in winter due to shading.  However, due to the low winter output, this has little effect on the production for the whole year.  Last winter my March production equalled the sum of the nov, dec, jan and feb production.

Another factor which was relevant last winter was that a lot of us lost output when snow covered our panels. Some of the days lost were quite sunny.

John

[dimengineer]

I have noticed this undershoot in winter too, I checked several peoples stats on pvoutput and found their winter stats to be well below the theoretical PVGIS estimates. I personally think it’s a shading issue. As the sun gets very low in winter more and more obstacles get in the way and result in an increased cut off or interference both early and late in the day. Whilst the theoretical estimates assume a perfect flat horizon. As the yields are low in winter anyway, any increased shading gives a disproportionately higher reduction.

I'd agree with that. My panels point towards some pretty tall trees. Up until late October the Sun clears the trees, so the panels are unshaded. Now the (mostly) leafless trees do get in the way - the sun drops behind them for all but a couple of hours in the middle of the day. So even in bright sunshine I was only getting 100W or so yesterday (sunny in West London) 

[nowty]

And I also had a think about why people exceed the summer estimates. I have another theory about this and that is that the assumed losses (inverter and cables) in the calculation of 14% are a bit pessimistic. Most transformaless inverters are 95% ish efficient and voltage drop through cables are only a few percent so I think 10% is more representative for a modern system. When I ran PVGIS Climate against several peoples stats on the south coast this matched better with the assumed losses set to 10% rather than the default of 14%.

[Mike777]

Up until late October the Sun clears the trees, so the panels are unshaded. Now the (mostly) leafless trees do get in the way - the sun drops behind them for all but a couple of hours in the middle of the day.

I am in a simlar postion so guess between Nov - Feb my generations totals will be down..............wondered what sort of reductions people are referring to in winter months.

I think I will be down around 30 % in November.  Expected 134 estimated production 102 !  Does that match other peoples figures ?

Regards

Mike

[langstroth3]

Would different array angles between systems have a greater effect in the winter as well perhaps, with the low insolation and low sun angles?

[supremetwo]

November has been a great month for sunshine but it has been disappointing versus PVGIS is this just down to the pitch of roofs or a bug or some other factor? Paul

Sunny perhaps, but many hazy days.

It takes little attenuation from the atmosphere to reduce the insolation.

[wattever]

I have also noticed a tendency for the PVGIS results to be a bit high in winter. The cloud data in the database is derived from a satellite with a scanning radiometer which looks vertically downwards near enough. At any location on earth, this data must be converted to irradiance levels using an algorithm which depends not only on sun elevation, azimuth etc but also cloud cover and vertical cloud extent. This vertical data is
not available directly but must be derived empirically. It appears that they assume more sunlight reaches the surface at low sun angles than is actually the case. The database and ouputs are relatively new, so I expect they will adjust the algorithms after a little experience and some statistical analysis.

wattever

[supremetwo]

The cloud data in the database is derived from a satellite with a scanning radiometer which looks vertically downwards near enough.

Depends which one you use.

PVGIS via BDPV uses app3, which is derived from ground-based readings during the period 1998-2005.

App4 is derived from the satellite readings during the period 2006-2010.

Either app is suitable to estimate an overall yearly energy production but cannot be relied upon to accurately assess the energy production of any PV system for one particular month of one particular year.

[skyewright]

November has been a great month for sunshine but it has been disappointing versus PVGIS is this just down to the pitch of roofs or a bug or some other factor?

Pitch & orientation are part of what you tell PVGIS, so they should not be a problem.

Have you considered setting up a custom horizon file for PVGIS?

If you don't provide one then PVGIS will just use topographical data to create one. Experience here is that their estimate is surprisingly good, but then there are very few houses or trees here to disturb the skyline! Creating a custom horizon is not hard - needs little more than a compass, a protractor, a weighted string and some paper. You list the results in a text file (there are instructions about what's needed on the PVGIS site), then feed that to PVGIS. The main problem is that ideally you should be sat in the middle of your array to take the measurements. 

[12TA]

There are different data sources for insolation. I have looked at some data from NASA also localised for location and adjusted to the same annual total output. That data looks steeper with less output in winter and more in summer. I think the weather this November has been above averagely bright (just my impression) and yet I am well below the PVGIS estimate for this month as others seem to be. I have 118 projected and 162 predicted by PVGIS.
The PVGIS curve just seems to be flatter than experience.

[skyewright]

yet I am well below the PVGIS estimate for this month as others seem to be.

The weather in the 1st 2 weeks were great here. Poor since (last Thursday and yesterday especially so producing 0Wh each, something I didn't see at all last winter), however overall production for the month is already a bit over 100% of our PVGIS November estimate.

Weather varies. In the South, Spring this year was exceptionally good, but it's only a few years since at around the same time of year there was prolonged exceptionally heavy rain in the same sort of areas.

Even if perfect, PVGIS is just an average so it will be several years at least before any patterns emerge...

[StationHouse]

PVGIS daily average for me is 3.75 units  in Nov, managed 3.45 units which seems fine.

Tis a new system that has only been going for 19 days 

[book_woorm]

I've always been skeptical about some of the prediction tools, especially the ones that use data from satalite monitoring. A while ago I built myself a spreadsheet predictor that used 10 years worth of actual BADC hour by hour, day by day (Met Office) data for both Global and Incident Iradiance. Using the average figures over the 10 years as a basis I got predictions within 5% of the actual power generated off of the panels.  It will be intereating to see what next years figures are like now the neighbours tree is shading the array part of the time.