Small impact of splitting floor space and adding perimeter zones on EUI?

As shown below, I did a test to check the impact of splitting concave zone and adding perimeter zones on total building EUI using a building with volume variations.

Four scenarios were tested as shown below from A to D, with different number of zones as a result of different floor space subdivision methods.

All four scenarios were test under two conditions by changing the definitions for interior walls only, one using the default solid wall, and the other using Air Wall (which is still a kind of thin solid wall).

The rest settings are controlled to be the same.

The results as shown in the table indicate that:

  1. for the solid interior wall condition, there is only very small difference in EUI across the four models.
  2. for the air wall condition, there is only a maximum of 4% difference between scenario A and C.

Considering the significant increase of simulation time, there seems to be very little gain by subdividing floor space and adding perimeter zones, if our primary concern is the estimation of annual EUI for the entire building, especially for energy performance evaluation in early stage of urban planning and urban design when architectural design details are not developed yet.

I’m not sure if this is related to the weather of the location considered here, which is tropical Singapore with relatively little seasonal variations, because the results are not consistent with what was reported in Timur Dogan’s study in which he found about 13% difference in EUI between the single zone and perimeter zone models (… not sure which weather file is used in his study)

Dogan, T., Reinhart, C., & Michalatos, P. (2014). Automated multi-zone building energy model generation for schematic design and urban massing studies. Paper presented at the eSim 2014.

Well, the concave zones cause severe errors in solar distribution calculation as reported in the err file. So, I need to do another test by setting solar distribution as full exterior as suggested and see if there is a difference, and probably using weather file of a different location.

Nevertheless, I’d like to hear your comments on:

  1. whether the results are reasonable.
  2. your experience in balancing the accuracy and efficiency of multi-zone energy modeling.

Thanks.

1 Like

Hi @Grasshope,
Nice topic. I don’t like specially the perimeter alternatives. You are introducing a big amount of thermal zones just to deal with the convex issue. You can change the solar model and still remain with the zones you really need. Even split the geometry to receive just convex polygons.
As the topic is something one of my students is going to deal with (really soon) i recommend you to check this paper (there are more):

Reinhart, C.F., Davila, C.C., 2016. “Urban building energy modeling. A review of a nascent field”. Build. Environ. 97, 196–202.

Where the authors propose a method to simplify geometry modeling for simulation on urban scale.
I don’t believe this is a topic that weather files will lead to some good conclusions. I believe more that this can be more related to investigating the building topology(s). This will be part of the research of my student BTW.
Please update about your findings. Is interesting.
-A.

Great discussion @Grasshope @AbrahamYezioro . I would add two points.

First, there is a sliding scale of building conditions that will require more or less zone splitting to accurately reflect energy conditions. So the comparison of total EUI can vary widely based on where our building lies within those conditions. Secondly, when we do need to split zones, we can still be very computationally efficient if we are precise about identifying which conditions we need to model, and which we can interpolate from the performance of other zones in order to reduce the need to brute force simulate all zones in a building.

Essentially, I think it’s more important to model precisely the energy dynamics in detail of representative or sample spaces, and then multiply or interpolate that out across similiar spaces, rather then do the reverse by generically (and inaccurately) modeling the energy dynamics of all spaces in the building then brute-force simulating them all.

So for the first point, there are a couple of different factors that I believe will impact the total building EUI when comparing split vs non-split floors. I’ll list the conditions in which it’s important to split the zones here, and try and explain why they should result in a different EUI from a single zone.

  • Higher surface area to volume ratios (i.e If you have a high SA:V building then the building energy use will be driven by external loads rather then internal loads. This means capturing the shifting external loads,by orientation should impact the EUI difference.
  • Higher WWR. For a similiar reason as above, the building will be driven by external loads, and will be very sensitive to orientation of facade, contextual shading etc.
  • More use of passive energy strategies. Natural ventilation, passive solar strategies are primarily targeting perimeter zones.
  • Having highly varied, and significant wind and solar exposure around the building. Again, this will differentiate perimeter vs core zone EUI.
  • Less internal loads in program. I.e offices tend to have high internal loads from occupants and equipment that will dominate the total EUI.
  • Less (or more regulated) mass and energy transfer through your interior zone walls. If you have an open office space you’re modeling with airwalls, then there’s little difference between a single zone and multizone. If the interior walls are adiabatic, then there’s a huge difference, and you will need to split your zones. Or, if you have a more “regulated” transfer between zones (i.e. you have a system of fans circulating air through multiple interior zones) then again, you’ll need to split zones to capture this effect.

That’s all I can think of for now, but I’ll edit this list if I come up with more.

To the second point: you don’t have to model all the floors, or even all the rooms in your building. A floor level that is 3.5m higher or lower then its neighboring floors is not going to have significantly different energy loads from them (assuming they all have the same program). You can just model the floors that will have significantly differences, due to program, external conditions, or boundary conditions, and then multiply the results to estimate total energy. E+ has a good documentation on how to do this with Zone Multipliers and changing the boundary conditions of internal zone surfaces. For example this is how they model a multi-story residence:

Note that, I’m not sure if Zone Multipliers can be used in Honeybee (@chris?) so you may have to go outside of grasshopper if you want to implement it in a more rigorous fashion the way E+ does it.

In summary, I think we can still efficiently simulate larger buildings without taking up too much computational resources if we increase the resolution of simulation in certain area (via zone splitting) and reduce the resolution of simulation in other areas (via linear interpolation or multiplying loads).

  • S
2 Likes

@AbrahamYezioro thanks for your comments and the recommend paper.

Yes, my focus is primarily on the impact of building typology and urban morphology on various types of building performances, too, such as energy, daylight availability, outdoor comfort, air movement, solar energy harvesting potential, etc. I attached my recent papers in this regard for your and your student’s reference.

The general finding is that building typology matters to a great extent in relation to many performance areas under a fixed density. This suggests that performance evaluation and optimization shall begin and will play a crucial role at the early stage of urban planning and urban design because the performance gain at this stage could be big, and typology with inferior performance at this stage might be very difficult and costly to remedy later at detailed architectural design development stage.

One natural question for urban scale of analysis is to what extent can we simplify the urban geometry to obtain relatively meaningful and reliable results efficiently in performance evaluation based on certain acceptable and realistic premises on common design factors.

This is why I want to check if more detailed zone subdivision may results in significantly different EUI estimation as I was using the “one zone per floor” method in the previous studies due to constrain in time. If the difference is not much, then maybe we can rely on relative simple zone definition for efficient estimation.

Nevertheless, more tests are needed to examine if this is the case regardless of building typology or it is typology dependent.

Will update if there are new findings.

2015 PLEA.pdf (836.3 KB)

Zhang 2016 IC2UHI.pdf (1.5 MB)

Zhang 2017 ABS.pdf (1.9 MB)

@SaeranVasanthakumar

Thanks for your comments which are quite thorough.

The premises here are primarily related to the early stage of urban planning and urban design in which we only have incomplete or no information on the following items:

  1. internal spatial division of the floor area which is related to the geometric definition of thermal zones
  2. function and program of each zone which are related to the definition of internal loads
  3. fenestration design which is related to glazing ratio and external loads
  4. constructions of building surfaces which is also related to the envelope heat gain and loss.

Without knowing too much about these design and operational factors, we may have to assume a standard “template” for them, especially for comparative study between different design options characterized as typology or building massing.

For a building with relatively the same plane across all floors, maybe the between-floor difference is not much. But for alternative building forms with changing profile or floor planes, the vertical variation in external and internal loads might be quite large. So, I think a floor by floor division of thermal zones may still be the bottom line option.

Nevertheless, I do agree that we shall model the thermal zones in relatively precise manner provided that information on the representative areas is known.

Thanks @Grasshope for the papers.
I see that you were inn Bologna and i missed your presentation :frowning:
Another point to think about making one floor zone is that it may distort the orientation influence. E+ documentation doesn’t recommend to to that. On the contrary the spike about defining thermal zones according to orientation. So i will take this point with special care.
In any case, you need to define all loads as they are in “reality” as all other parameters.
@SaeranVasanthakumar idea of using multipliers can save tons of time, but you need to have a good control of the system. Though is not implemented in HB (yet), as he said.
Another point is the simulation time step (hourly, monthly, yearly). The last can save a lot of time. Here it depends what kind of resolution are you looking for. My understanding from what you write is that yearly can be the one to use.
Keep updating @Grasshope. This is interesting.
-A.

@chris Is honeybee able to apply this zone multiplier function like E+?

I’ve yet to use it, but yes it’s in there: https://mostapharoudsari.gitbooks.io/honeybee-primer/text/components/Set_Zone_Properties.html

S