This is the follow-up post to
After failure of dynamic shading with a second shading group I tried to approach the concept of having split blinds with high reflectance in the upper part by using the upmost 9 slats horizontally as a fixed shading application. The lower 19 slats I simulated dynamically (shading group 1: open / 30° / 50° / 70° / 86°). Again from row1_west of the reference office. The dgp outcome:
Obviously the 30° tilt of the lower blinds is almost good enough for killing most of the glare before 2pm.
The remaining glare situations must be resulting from the upper, horizontal slats.
After doubting these results I checked the critical hours (dgp> 0.45 in …..effective.dgp) by batch running point in time simulations in gh using the according shading state from ….intgain.csv. The point in time results seem to be quite reasonable but differ completely from the annual run results.
63 hours showed dgp values higher than 0.45. as a representative example here the result for November, 6th 12 o’clock (dgp annual: 0.78 vs. p-i-t: 0.19) :
Single point in time simulations using the toolbar application support the results of the grasshopper results:
The shading state of all 63 hours is 1.0 (meaning maximum shade, upper 0° and lower 86°):
And finally the dgp comparison of point-in-time results vs. annual results:
Which results do you judge as correct?
I completely depend on correct annual values for lighting energy investigations in my thesis…
Any ideas (maybe another bad duplicate binary file ;-) – which would make it quite easy hehe ) ?
P.S. 3dm attached, epw of Stuttgart, Germany used , but principally the same issue with Boston weather file
Some new findings:
I compared the annual run dgp file for the maximum shading state (upper slats statically included in building geometry, lower slats at an angle of 86 degrees = closed) of the dynamic shading simulation mentioned above with a dgp file produced with the same shading geometry but fixed at 86° (no dynamic shading). I assume these two dgp files should look almost identical because the geometries are identical).
But instead of 63 hours with dgp>0.45 there are only 7 hours left (toolbar annual run, gh just used for visualization of gdp results):
And point-in-time control simulations for these hours still show giant discrepancies.
(upper right corner: corresponding dgp value of annual run)
The exist some few hours when the annual results exceed the p-i-t results. E.g.:
Main problem: without reliable annual dgp results there will be no shading schedule and hence no lighting energy calculation…
What’s the difference in dgp production? Annual run uses DaysimBinaries’ gen_dgp_profile as far as I can see but I don’t know how the p-i-t – files are being generated (at least toolbar and gh produce identical results ;-))…
Any hint available?
Maybe there's someone out there who is willing to run these two simulations on his machine and to share his results with me. The two files are prepared and attached.
Settings first run: weather file Stuttgart (attached), view row1_west, no dynamic shading, annual glare -ab 4 (layers are preset, materials out of standard palette).
The annual and point-in-time glare metrics involve different methodologies, so it's not surprising that the results don't match. The annual method starts with a vertical illuminance simulation using Daysim, which takes care of term 1 in the DGP equation. (The second term -- involving individual glare sources -- is evaluated for specular reflections using -ab 0 point-in-time renderings.)
Since Daysim interpolates the direct contribution using 4 "nearby" sample positions for the sun (out of a set of ~65), the approximation of the first term is inherently less accurate in the annual runs. This also explains the dramatic discrepancy for particular hours, where small shifts in solar position could easily force glare sources just in, or out, of view.
Of course, before you go off and treat the point-in-time results as gospel, it's worth asking yourself how much pinpoint sun-position accuracy is worth when (a) you rarely know exactly where an occupant's eyeballs will be, and (b) sampling merely one sun position per hour is already arbitrary and incomplete.
As for the static/dynamic shades, I'll have to take a closer look at the attached files...