Need help trying to optimize SHGC and U value with Colibri

Dear all,

Sorry in advance as this may be a long and difficult post.

I am trying to help select the most suitable glazing for my project. Specifically choose the right balance of SHGC and U value. Concern is that a high SHGC value will benefit in winter but hurt in summer.

To help find the optimum, I am trying to iterate an energy model and a daylight simulation with Colibri. I found a sample file posted on the GitHub shared by Mingbo and built my geometry in it. Unfortunately I have not been able to run a single simulation thus far. Please find my GH file attached. Please lock your solver before opening the file.

Question

  1. the Colibri Aggregator is giving an error “object reference not set to an instance of an object” how can I resolve this?
  2. energy simulation is taking too long. Can anyone advise how to simplify and run the energy calc faster. I know for daylighting we can reduce the ambient bounces is there something similar for energy calcs? Or should I limit analysis period to specific days like summer solstice etc.
  3. Can I run the same script on two computers, one each for energy and daylighting and combine the results to view in design explorer?
  4. Is there a alternate method to find the optimum shgc vs u-value combination for a project without iterating thought the energy and daylight model?

Thank you in advance
Best,
Ibby

Energy & Daylight_With_Colibri.gh (723.1 KB)

1 Like

Hi Ibby,

A couple of things that I notice -

The first thing is Energyplus will not let you simulate curved walls. If you really need to simulate curve walls you would need to break them down in to a series of straight sections. Another thing your zones should not be non-convex.

This image shows your current energy model on the left. Energy plus prefers the model to be like the one shown on the right. Broken down into four zones. Convex and nonconvex rules also apply to surfaces.

here is a good example of convex and nonconvex surfaces

image

also check out this post on the old GH3D

I rebuilt your model to be a four zone space. That has corrected your error from question number 1.

response to Question number 2 - you have over 8000 iterations simulating that for both energy and daylighting will take way too long. Try reducing your options. Break it into smaller steps. Maybe start with a fixed window to wall ratio. Wait on testing those window to wall ratios till after you have found your optimal U and SHGC. Another thing that would help is removing your HVAC unit and running an IDEAL loads run.

You can also reduce your time steps to 1 per hour for a total of 8760 hours/ timesteps(the default is 6 every 10 minutes for a total of 52560 steps for the whole year).

Question 3&4 -I don know the answer to these. I would like to know the answer to those as well.

Here I updated your model, best of luck!

Energy & Daylight_With_Colibri_v2.gh (692.0 KB)

1 Like

Hi @Ibby FabianPosadas,
What i would do for questions 3 and 4 is:
3. In the exportToOpenStudio component i’ll set the writeOSM to True and runSimulation to False. This will create the files to be run later on. Then use the Re-Run IDF to run the cases in parallel. I will also set the timestep_ input in the HB_EPOutput to “annual” (unless you want to get more detailed results, it will be just fine for your purposes to get the total values. It will run faster).
4. Assuming your want to get the less energy consumption when combining energy and daylight, why you don’t use Galapagos (or any other optimization solver)? Why you need to have all the 8000 cases? Maybe i’m missing something here in this question …
-A.

Hi Fabian,

Thank you for helping me with this. 1 & 2 are resolved. For the time being, to reduce the number of iterations, I have plugged in the SHGC slider into the VT as well. I think I will have to run a base case this way. Once I get a SHGC & U value optimum, I will have to rerun the simulation to adjust for VT based on a few actual glazing samples. If anyone knows of a better way, please advise.

Hi Abraham,

Thank you for helping me with this.

  1. If I write the OSM now and run it later. Once the simulation is complete, will I be able to combine the results with the daylighting data in Design explorer?
  2. I am a little not clear about using Galapagos, please advise. As I understand Energyplus & Radiance feedback is required to identify an optimum. Using Colibri, we get all the results and use design explorer to play with different scenarios. Galapagos I assume would use the same EP radiance feedback except reduce the number of iterations and I don’t get all the results to use with design explorer. Is this correct?

Thank you,

Hi @Ibby,

I am not sure how you set up each input slider, if you want to test a couple selected value within a certain range, you can use the value-list to reduce the iteration count image

To split the total simulation works into two or more computers, you can use “Iteration Selection”, and set Iteration domain for each computer to run each selected range, and you can manually put all data together (copy-paste) including combine all data.csv files once all computers finished their work

Hope this answers some of your questions.

Mingbo

Hi Mingbo,

Thank you for all you help. Yes, I use value lists to selectively input some of the values. I actually used your Colibri sample script from the GitHub page as a base for this exercise.

Question: instead of splitting the iterations, would it be possible to run daylight one computer, export OSM and run in parallel on another computer. Recombine results in CSV to populate in design explorer?

Thank you and keep up the good work.
Best,

It’s possible but it needs some work on your side. DesignExplorer simply reads a csv file. It doesn’t matter how the csv is created.

Also see this: Parallel EnergyPlus runs with Colibri or optimization components - #4 by mostapha

1 Like