[Talk-GB] RFC: Solar panel mapping in the UK
sk53.osm at gmail.com
Thu Apr 11 09:59:03 UTC 2019
I'll quickly add my responses on the thread:
- *REPD issues*. All of Rob's points taken, but we mustn't forget that
OSM data have always been acquired and refined iteratively. Of course data
from REPD has to be taken with a pinch of salt, but at least for now it's
very useful for hunting for missing installations. In practice I've
found most REPD installations relatively easy to resolve (but see below for
an exception). Russ does compute a power output for those sites which
don't have the output explicitly tagged, so there is the potential to
compare the REPD output and a computed value based on area.
- *ML & Solar Farms.* Tyler Busby has been working to identify rooftop
solar using machine learning. He has a MapRoulette challenge running for
Austn Texas at the moment. I imagine it might be possible to reuse some of
his techniques to identify individual rows of panels within solar farms,
which could improve power estimation from OSM data.
- *Sections in Installations*. Exceptions, such as single installations
with multiple sites certainly exist too. I recently mapped panels on the
site <https://www.openstreetmap.org/#map=15/52.7803/-0.9372> of the
former Asfordby super pit. There are two groups of panels which a Geograph
photographer calls, on the basis of photos of ancillary electrical plant,
Asfordby A and Asfordby B <https://www.geograph.org.uk/photo/5925161>.
There are also photos of Asfordby C. As usual more can be learned from
on-the-ground visits, but as above this is for future refinement.
- *Rooftop angles*. I had a futile attempt to try & calculate roof
angles from Lidar data. The 1 m resolution doesn't seem to be adequate.
Maximum roof height is more reliable (available for instance via the
dataset). Estimating the height of eaves can be done from Lidar, but it's
fairly fuzzy. I think using rules of thumb for different periods of
construction may be just as fruitful (perhaps 9 foot ceilings for pre-WWII,
8 foot for interwar housing, and 7 foot 6 thereafter, with 1-1.5 feet
between floors). Counting courses of bricks would give a more precise
measure and only needs to be done for basic ranges of housing. Most local
archives are likely to have architects drawings for houses built as council
housing which is perhaps a third of the total stock. However a basic
estimation of eave level from 5-6 m will not be hugely out. See next bullet
for a suitable tag.
- *Other tags*. After much faffing about, and on Russ' advice, I have
now moved to using location=roof instead of generator:place or
generator:location. This doesn't work if the generator tags are placed on
the building as is the case for some places in the West Midlands, but as
these result in gross over-estimation of likely output I'd regard this as
an interim stage of mapping. I'm still using generator:orientation, but
this may also be more unwieldy than required, and obviously relates to
solar installations only. Modules are tagged generator:solar:modules which
at least unambiguously shows that it relates to the panels, so despite the
unwieldiness something similar for angle would be clear. (As an aside I
don't think we have any UK solar farms with panels mounted on heliostats,
but they certainly exist in Spain, for instance at Almaraz).
- *Power tagging*. One thing which has become clear is in mapping groups
of panels within a solar farm and retagging the outline as power=plant
isthat the use of generator: and plant: tags is unfortunate. Most of them
would work just fine as they were originally with power.
- *Solar arrays vs solar panels*. The current tagging largely seems to
fail to distinguish between a large array of solar panels and single panels
consisting of a few modules. I really don't think we want to end up having
to map each group of panels individually so it would be nice to have a
better way of distinguishing them other than location=roof and overall
area. Perhaps less than half the area of an array of panels will be the
actual footprint of panels. Also I'd be unsurprised if some don't map
solar-powered rubbish bins, parking meters, road signs with power=generator
Lastly big thanks to Jez, Dan, and especially Russ for his updates to
OpenInfraMap which really help with the mapping.
On Wed, 10 Apr 2019 at 23:01, Dan S <danstowell+osm at gmail.com> wrote:
> Hi all,
> Thanks for the comments on solar panel mapping. (Plenty of mapping
> happening already: thousands of UK solar panels added to the database
> in the past month.) A few small responses:
> SOLAR FARMS:
> I'll defer to Russ's tagging advice about solar farms: power=plant
> polygon (or sometimes multipolygon) as the outline of a solar farm,
> with power=generator areas contained within it for the blocks of
> panels. Previously, I was mapping solar farms as relations, but I'm
> easily persuaded!
> I don't have any advice about landuse/landcover other than that it's a
> fairly separate issue, since those tags are not essential to the solar
> power mapping.
> I've been adding some solar farms that are listed in the REPD list on
> the wiki. For those ones I've used a tag "repd:id=*" which I hope
> makes it easy to identify them using the ID number in that database.
> Some solar farms have more than one entry in the REPD (they submit a
> new application form when they have an expansion).
> ROOFTOP SOLAR:
> For various reasons, if we can get solar installations mapped as areas
> not just nodes, that'll be helpful. Areas will be more useful than
> module-counting. However, I've noted that the imagery doesn't always
> make this easy for rooftop solar: clarity is variable per region.
> Is there any good way to tag the vertical tilt of a panel? I know in
> many cases we won't be able to measure it well, but I thought I'd ask.
> For example, there's roof:angle=* for the slope of a roof, which is a
> mildly related concept.
> Talk-GB mailing list
> Talk-GB at openstreetmap.org
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Talk-GB