I want to ask about pollination cloud computing services.
My study consists of a parametric analysis with 175.692 variations. I want to analyze these variations parametrically using Colibri. When I examine the plans in the Pollunation cloud computing service, I see that the professional option is limited to 2400 runs and 2000 CPUs. When I watch the parametric analysis video, I see that there is a separate run for each variation, how can I do this analysis within these limits?
In addition, in the current free version of Pollunation Cloud Computing, one analysis with 3 CPUs takes about 20 minutes. I guess it will be faster because there will be more CPUs in the professional version, but how can I calculate if this time is enough to do all my analyses?
This sounds like too many simulations. Can you explain a bit more about what you are trying to achieve? We can probably help you to reduce the number of simulations.
You have to break down your studies into smaller batches and compile the results together. But for this many iterations, you probably want a custom recipe that also includes the post-processing step so you don’t have to download all the results.
Yes. Generally speaking, It will be faster because you will use more CPUs. Calculating the time for all of your studies depends on many factors. You can usually get a good sense by running several batches of the same model.
Two more items:
I can’t see any studies under your account, but I can see the model. Is this the model that you are trying to parametrize?
I am pleased to use this unique platform that contributes to our academic studies, thank you.
I aim to make these analyses to create a data set that I can use in an ml work.
My analyses consist of 3*3 m sized zones (rooms) in different building forms and I need to get the energy results of each of these points. For this, I have created separate boxes for each zone, so the analysis time may be longer. I would be very grateful if there is a method to get regional analysis results without physically zoning a building.
Since there are 12 orientations for each building form and 11 WWR values for each facade , there are 12x11x11x11x11x11x11=175.692 variations in total.
Yes, this is my model, my work consists of this and similar building models separated by 3*3 m units of different forms.
I did not understand this part of your answer, is the current common-energy OSM recipe not enough for me?
Can’t you just sample the rooms? Many of them will give you the same results anyways. For example:
For the rooms that are facing outwards, you can just pick one room in each direction to represent the rest of the rooms.
You can get rid of one of the floors. You only need the ground, middle, and top. You can also model all the hallways as a single room. or just not model them.
For the rooms that are facing the courtyard the shading pattern can be different but you can study that by introducing external shades instead of modeling the rooms.
If I were you I would spend more time thinking about the input parameters, and how they will be used eventually before running all of these models. You might get better results by simplifying this problem to a single shoebox and running the study for the shoebox in different orientations and with different shading geometry.
What are the output metrics that you are looking for? Is it only the EUI or do you need some postprocessing to get the values that you need?
