Question-and-Answer Resource for the Building Energy Modeling Community
Get started with the Help page
Ask Your Question

PAT 2.9.1 Algorithmic mode is not creating alternatives

asked 2020-01-23 17:16:50 -0500

sborle's avatar

updated 2020-01-25 14:21:44 -0500

I am relatively new to PAt and this is my first time trying the algorithmic mode. For the current project, I need to test combinations of a range of values for the following variables - 4 different wall insulations, 4 different roof insulations, 6 values for window u-value, 5 values for window SHGC - which total to 19 different variables. I am trying to run a combination of all these options using the Design of Experiments - full_factorial sampling method. The setup runs successfully on the cloud but with zero data points and a results csv file with only titles of my measures and output variables but no values, essentially, a blank results file. Did I miss something in my parametric setup?

Questions -

  1. Algorithmic method - Is DOE the correct method for my analysis?
  2. Algorithm settings - Are the number of samples the total number of distinct variables I have viz. 19?
  3. OpenStudio Measures - is this setting wrong? I am using "replace exterior window construction" and "swap construction in construction set". Variable setting is discrete, static/default is the existing construction, measure inputs are the options to replace with. What are the weighing factors in measures for? I have left it blank.
  4. Reporting Measures - OpenStudio results with the variable setting of argument (i also tried discrete). I also need the peak cooling and heating values, where do I get it from?
  5. I have selected the variables I want from OS Results under the output tab with variable type "integer", is that correct?

Answer to even to one of my questions is very appreciated! Thanks in advance!

edit retag flag offensive close merge delete

1 Answer

Sort by ยป oldest newest most voted

answered 2020-01-24 10:23:03 -0500

updated 2020-01-24 10:28:46 -0500

  1. If you need to do every possible combination, then the DOE is correct as it will create a full factorial design of experiments for you. You can also test your workflow using the LHS sampling method with a few samples to ensure you have your problem formulated correctly before you do a larger DOE run.

  2. The Number of Samples argument is for continuous variables. This will discretize them according to the number of samples you give, and those discrete values will be used in the DOE.

  3. Weighting factors for discrete variable describe the distribution for the variable (i.e., probabilities). If you dont specify them, then the values are assumed to be normally distributed and will get equal weights that sum to one. For reference we use qdiscrete and its implemented here.

  4. Variable settings for reporting measures should be left to Argument. However, you assign the Outputs you care about as Objective Functions on the Outputs tab and they will show up in the plots/reports

  5. I would make your outputs type 'Double'. Also set Objective Function to True.

There are example PAT projects in this git repo that you can clone and run.

I'm also assuming you are running this on AWS?

Also the log showen with the red arrow has information on how the algorithms are setup and running. It might have some insight as to your issue as well.

image description

edit flag offensive delete link more


Thanks, Brian. Your responses were very helpful. I was able to run the project using LHS, not DOE though. And yes, I am running this on AWS.

For the results, in an attempt to get the peak cooling load, I added additional variables from e+ .rdd file "zone list sensible cooling rate" with the frequency set at "hourly . I used "AddOutputVariable". In the output tab, I don't see the variable in "select outputs" after adding the "AddOutputVariable" measure. Also, how do I add outputs in "OpenStudio Results"? Simply copy-pasting the variable name from .rdd isn't working

sborle's avatar sborle  ( 2020-01-24 17:11:46 -0500 )edit

Another question - is there a way to download all the generated zip files for alternatives from server at the same time. Or do I have to download them one by one. Just in case if I have to figure out another way of getting the results for output variables i need.

sborle's avatar sborle  ( 2020-01-24 17:14:20 -0500 )edit

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer


Question Tools



Asked: 2020-01-23 17:16:50 -0500

Seen: 147 times

Last updated: Jan 24 '20