Question-and-Answer Resource for the Building Energy Modeling Community
Get started with the Help page
Ask Your Question
1

Sampling of input parameters for Sensitivity Analysis

asked 2018-11-05 15:57:40 -0500

Vasco Zeferina gravatar image

updated 2018-11-06 11:37:37 -0500

I intend to execute a sensitivity analysis (SA) of 16 input parameters in an Energyplus model using JEplus.

At this moment, I have chosen the type of SA that I want to perform and I also have defined the input variation range of the input parameters to be varied (16 in total). Thereafter, I am struggling to make the sampling of my input files accordingly my future analysis (Morris and SOBOL).

Firstly, I want to make a screening analysis (morris method) of all my variables and then reduce my analysis to 6 parameters and conduct a SOBOL global SA. To generate my input files for Energyplus/JEplus, I have tried to use SIMLAB 2.2. However, the program continuously crashes, and it does not permit to create my sampling instruction. I am now trying to create my samples using some Matlab functions, and I have also tried looking at R-functions.

I would like to know, which method have you been using to perform your sensitivity analysis. How have you been generating your samples and calculating the respective sensitivity indexes? Which matlab and R functions are available. Which programs have embedded sensitivity routines (screening and global SA) to proceed these analysis?

Regards, Vasco Zeferina

edit retag flag offensive close merge delete

1 Answer

Sort by ยป oldest newest most voted
1

answered 2018-11-05 17:20:12 -0500

You tagged this as JEPlus, But OpenStudio PAT provides those analytic capabilities for you as well.
The underlying R algorithms come from the R 'Sensitivity' package. Specifically the Morris and the various Sobol methods are used to generate the solution space and are then batched run in parallel.

edit flag offensive delete link more

Comments

Thanks for your answer!

I have limited experience with PAT. For me, it seemed to provide poorer systematic approach than JEPlus does. Do you reckon this?

I have been looking to the R 'Sensitivity' package and I couldn't find a function to generate my random input parameters to import in my job list to make simulation in EnergyPlus/JEplus. The functions in R 'sensitivity' seem to require a model as an input. However, I may be wrong, and it is possible to export a sample and import results to calculate indexes in R 'sensitivity' functions.

Let me know if you have experience on using R.

Vasco Zeferina gravatar imageVasco Zeferina ( 2018-11-06 11:36:16 -0500 )edit

I personally think PAT is more powerful since its built off of the functionality of OS Measures as well as leveraging all the algorithms of OS Server, but I am a bit biased. For the R algorithms, just use a NULL model and use the returned sample space however you want. Thats what we do in OS Server (or PAT)

BrianLBall gravatar imageBrianLBall ( 2018-11-06 14:17:21 -0500 )edit

Here's a documentation example using R sensitivity with an external model example

BrianLBall gravatar imageBrianLBall ( 2018-11-07 10:04:54 -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

1 follower

Stats

Asked: 2018-11-05 15:57:40 -0500

Seen: 76 times

Last updated: Nov 05 '18