Question-and-Answer Resource for the Building Energy Modeling Community
Get started with the Help page
Ask Your Question
2

cloud based cbecc-com / how to reduce run time

asked 2017-05-24 12:26:14 -0600

XChen gravatar image

Is there any online/cloud application for california cbecc-com for faster calculation and batch processing? Or any other suggestions on how to reduce the run time for cbecc-com on desktop computer?

edit retag flag offensive close merge delete

2 Answers

Sort by ยป oldest newest most voted
3

answered 2017-12-23 05:47:36 -0600

updated 2017-12-23 05:47:48 -0600

There is no specific cloud application for CBECC-Com models, but you could always purchase any cloud-based service (such as Amazon) and run it that way.

The best ways to reduce run time are to create a simple model. Use the least number of glazing surfaces you can (combine and simplify glazing geometry), and the least number of surfaces overall. Simplify your zoning and geometry wherever you can, you can typically combine all back-of-house or support spaces into one space, for example.

In addition, you can run the "Quick Analysis" option, just to debug and get through errors in early stages. But you should aways run a full annual analysis to determine compliance, as the quick analysis option will sometimes show compliance when a full analysis won't.

edit flag offensive delete link more

Comments

I can second @Anna Osborne Brannon's comments above. Simplifying the model and 'Quick Analysis' are really the only two ways to 'speed up' the simulations. I have done testing on running CBECCcom on web/cloud services; both MS Azure and Amazon Web Server. The gains were underwhelming to disappointing. There are two main issues ( as I have been informed). The first is that the underlying simulations are in energyplus, which is single threaded. So, it doesn't matter how much multi-core power is available, it all gets back to the speed of said core.

dradair gravatar imagedradair ( 2017-12-25 09:04:23 -0600 )edit

this post link text disusses the hardware side of the EnergyPlus stuff. The second issue is that (I've been told),the CBECCCom application has some other call-out/pause in its simulations which further slows down simulations. I cant say I understand it....hearsay only.. A third issue which can lead to significant delays is that the CBECCcom model is translated from CBECCCom to OpenStudio to EnergyPlus. On a very large model which took ~48-50 hours to run, this was 4-6 hours of translation time.

dradair gravatar imagedradair ( 2017-12-25 09:07:40 -0600 )edit

Final thought, I haven't seen a 'native' way to batch process within CBECCcom. I am sure it is possible to write an application which would edit the (text) input files, launch CBECCcom and start a simulation...but that's something I haven't seen.... Long simulation times are an issue we have run into in multiple projects and something we have just learned to accept....

dradair gravatar imagedradair ( 2017-12-25 09:10:14 -0600 )edit

48-50 hours runtime is just "something we have just learned to accept..."? I think you're accepting the unacceptable! Based purely on the economics, I do not think a compliance calculation should take more than an hour, and ideally less than 10 minutes, which would have been ample for a program like DOE-2 or eQUEST. There are many deleterious effects from the long runtimes that I wish the "authorities" would realize, including not fixing mistakes nor doing many (any?) parametrics. I doubt the CBECC-Com simulations being done now are any more accurate and probably less so than before.

Joe Huang gravatar imageJoe Huang ( 2017-12-29 06:38:52 -0600 )edit

dradair, I continue to be amazed at your comment about a very large model taking 48-50 hours to run, with 4-6 hours of translation time. I currently have a small project to evaluate the use of CBECC-Com for NZE building design, and want to make sure such long run times are not an urban myth. I'd appreciate getting more details about your project and experiences in using CBECC-Com in general. You can reach me via e-mail at yjhuang@whiteboxtechnologies.com.

Joe Huang gravatar imageJoe Huang ( 2018-01-01 04:53:27 -0600 )edit
1

answered 2017-12-29 06:56:02 -0600

Joe Huang gravatar image

I think going to cloud-based computing to speed up CBECC-Com runs will introduce other problems of its own like cost and a learning curve, and avoids the issue why CBECC-Com is so slow?

One idea I've had for several years that would speed up CBECC-Com by at least a factor of 4 is to run simulations with a reduced weather file containing only 12 weeks instead of the full 52 weeks, i.e., 1 week to represent each month. This approach is similar to the Quick Analysis method, except that instead of picking the simulation time periods by the calendar and thus getting a random sampling of the weather data, I propose to analyze the weather data to pick the time periods most representative of each month just like how the months of a typical year weather file are picked. I wrote a paper on this method for SImBuild 2016 in Salt Lake City, where I showed that the results were indistinguishable from those using the full typical year weather file in terms of accuracy. This paper is available at the following link: http://ibpsa-usa.org/index.php/ibpusa...

edit flag offensive delete link more

Comments

This sounds like a great idea to me, although I've noticed the "quick" analysis isn't actually that much quicker (drops my current model's 8 hour run-time to 5 or 6 hours).

Luckily for the 2019 code cycle, the draft Standards include language that will allow performance compliance calculations to be done using alternative engines (i.e. non-CBECC, and even non-EnergyPlus). If these alternative software vendors step up to develop the ruleset and reporting, I bet people will drop CBECC-Com overnight.

Greg Collins gravatar imageGreg Collins ( 2018-02-18 11:24:28 -0600 )edit

I think the relative lack of speed improvement with the "quick" mode in CBECC-Com is because EnergyPlus has to re-initialize whenever there's a gap in the run-period. If the re-initialization is 10 days, then doing 4 15 day run-periods is making each of them really 25 days or 100 days in total instead of 60 days. I totally agree that reinstatement of ACMs may spell the doom of CBECC-Com.

Joe Huang gravatar imageJoe Huang ( 2018-04-06 14:26:17 -0600 )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: 2017-05-24 12:26:14 -0600

Seen: 104 times

Last updated: Dec 29 '17