LB Incident Radiation Analysis Doesn't Match Revit Solar Analysis

Rhino 7.20.22193.9001 LB1.5.0
RVT 2022

For a tall tower in a southern US climate with 77,000 sqft of surface…Graphics look remarkably similar and spot range of about 0-6 kWh/m2 for each seem right. However, I got these results for total incident radiation…

       LB     RVT

Dec21 182,102 136,373

Mar21 209,450 180,654

Jun21 141,403 170,429

Geolocations match
North match
Date and time periods match
Geometries match (I even manually removed the bottom of the rhino brep to see if that are was swinging the numbers…it wasn’t)
Daylight savings unchecked in RVT for each period and I’ve left LB locational TZ unchanged.

My gut tells me that LB is correct even thought the results look counter-intuitive. As LB indicates, I believe that the winter total should probably be higher than summer total because the winter angle of incidence on the tall facades is less grazing and therefore radiation less diffuse. If my belief is correct then the RVT numbers aren’t just off but really wrong because their summer total is way higher than their winter total. It wouldn’t bother me so much if the results were just a bit dissimilar but this is wild.

@mostapha any thoughts?

I would check the weather data first and make sure they match. Revit uses a different source for weather data. Did you use the climate-based sky when running the study in Ladybug Tools?

I don’t know if Revit’s approach is validated at all. Hard to tell.

1 Like

I didn’t think that one could actually run the LB Incident Radiation component without a Sky Matrix, but yes, I’ve got one plugged in and its got the same north and is deriving its vital stats from the same EPW just about a mile from the site… the same location, direct normal radiation and diffuse horizontal radiation.

Do you disagree with my assumption about better angle of incidence in winter?

@rnarracci, I agree with your assumption on solar gain in winter - climates I usually study have peak daily radiation around the winter to equinox period, I’m sure some of it will depend on sun up times but in general lower winter angles result in higher peak and incident solar.

Without digging into how Revit is doing the analysis I would trust the LBT result first. Looking into the LBT calculation method it makes sense to me roughly considering the physics principles.

If you’re really keen to get to the bottom of the difference I’d recommend digging into how each do their calculations - solar calc methods are generally fairly simple to compare. On top of that I’d be checking the climate inputs are the same, particularly hourly radiation values, and also how each method considers ground reflectance - I believe the LB method just flips the sky matrix and multiplies it by the ground reflectance (0.2 default I think)

1 Like

Thanks Charlie,

So far as I’ve read, RVT uses the Perez Solar Model and I assume weather data is gotten when one geolocates the model with its Location tool. The Sun path diagram looks very similar to LBs. Otherwise it’s pretty opaque. I find it hard to believe that they’d be that wrong and will continue to dig and see if there is actually a weather input somewhere that I’m missing.

Ok so the RVT location selector just grabs the nearest weather data file and I’ve read that those prefaced with the number 59 contain Typical Meteorological Years, whereas the others are from Autodesk’s “climate server” (source unknown)…so (seeing an actual weather data selector right next to the location tool (head slap)) I switched to the nearest one prefaced with 59 and that just happened to be the same location where I retrieved an EPW for LB. This is the results with RVT numbers revised…

      LB     RVT

Dec21 182,102 188,079

Mar21 209,450 223,861

Jun21 141,403 155,207

Still off by 3%, 7% and 10% respectively. A lot better than being off by 20% but the path I took to get there is still a bit disturbing…at least the number gradient roughly matches LB.

1 Like

@mostapha does every available EPW indicate somewhere therein whether it is a TMY2, 3, or x?