[Talk-GB] Solar tagging app

[Gregory Williams]

On Sun, 2020-12-27 at 22:38 +0000, Gregory Williams wrote:
> On Sun, 2020-11-08 at 15:17 +0000, Jeremy Harris wrote:
> > On 19/10/2020 20:14, Gregory Williams wrote:
> > > I've now got round to updating the code on my solar comparison
> > > site
> > Suggestion for further improvement:  When a building is
> > identifiable
> > associated with a set of generators, only count one for comparison
> > against the FIT data.  Then (say) a house with panels on two
> > different
> > roof-faces will not over-count for the geographical region.
> > 
> > (I just found an LSOA with 4 of 13 mapped.  But 2 were on one home.
> > A terrace too, so now I must split it, so the 1 on another home in
> > the block is really distinct)
> 
> I've finally managed to find the time to update my comparison tool
> [1]
> to take this into account. Columns now show for both "Generator
> count"
> and "Filtered generator count". The latter only counts a generator
> once
> if there are multiple on the same building. That's the one that now
> gets used in Completeness calculation.
> 
> As a result the overall completeness has gone down from 35.76% to
> 32.00% for today's refreshed update.
> 
> So, now, excessively high levels of completeness may be down to
> modules
> having been mapped on different orientations of the same building,
> but
> the underlying building not mapped itself.
> 
> Now, low levels of completion may be because the building mapping
> isn't
> complete. For example modules mapped on several different addresses
> in
> a terrace, but the terrace building not split up into those separate
> building ways.
> 
> I hope all are enjoying the Christmas break,
> 
> Gregory
> 
> [1] https://osm.gregorywilliams.me.uk/solar/
> _______________________________________________
> Talk-GB mailing list
> Talk-GB at openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk-gb

Eagle-eyed Brian Prangle spotted that the filtering to avoid extra
counting of installations on the same building wasn't quite working in
some of the areas that he's been editing. Thanks for making me aware,
Brian.

I investigated this evening and found the cause of the issue and have
now resolved it. For cases where a solar installation was also mapped
as a building, and where that building shared common vertices with
other buildings (such as a terrace), the other buildings in that
terrace were also being mistakenly adding to the count.

Nationally, this has only made a small difference to the count (was
278138, now xx), but enough to be visible on the progress chart. And,
of course, enough to give a false impression of completeness on the
local level in affected areas.

Gregory