I am trying to run DA according to LEEDv4 where shades are closed when 2% of the area has direct sun. The idea was to divide the process into two steps.
First step: I divide the windows per room/grid and then I generate the schedules based on custom sensors (derived from each grid) that are applied to each window group and custom sensor count (2%) based on the amount of sensors per grid.
First thing I noticed here, was that even though I supply all inputs in the "'Windows" component as trees (pic 1), when I run the simulation somehow the component rearranges everything and I end up with the same schedule per orientation, regardless the room size or depth. Since the schedule is depended on sensor count with direct sun, the deeper rooms should have their windows open for more time. I managed to solve this by running separate simulations per grid as a batch with different names. The calculation takes milliseconds so even if I have 30 rooms and only 10 cores it is still doable. So I ended up with different schedules per room! success! (pic 2).
Second step: I took a new "Window"component and added the schedules as 8760 values (generated in the previous step) to each window (again as trees, with 104 paths each to be precise) and then connect my windows to the "Daylight" component and run the DA simulation. Of course this might end up in 100 annual runs after the pairing of grids and window groups but I am not there yet. I just run the simulation with only windows as output and I see that the schedule has changed, even though I defined it before hand. (pic 3).
So any ideas what might be the problem or how to solve it? I have approximately 100 rooms and I would like to avoid run each separately. Also is there a way to change the way grids pair to each window group?
PS. I also tried to directly connect the windows with schedule generated from the "Daylight" component in the first step to the second DA simulation but it only resulted in one annual simulation without shading.
PPS. I ve been trying to post this issue since last week with no luck. Hope this time it will be posted.
If you're trying mimic the LM-83 2% threshold, you should use auto sensors generated from the grid -- see my response to this post.
That said, I'd actually recommend doing exactly what you were trying to avoid, which is running all the rooms separately. If you run them together, you will end up increasing the runtime several fold, because DIVA will try to account for all the shading states of light spilling between rooms, down hallways, etc. These states would have some non-zero impact on real-world sensors, but potentially less than the size of the simulation error itself. Don't waste your precious CPU time on these scenarios.
Indeed, running each room separately is much more efficient. Batching the files makes the process quite fast. Thanks!