BEopt Optimization Running Super Slowly

asked 2019-02-28 11:51:11 -0600

Alisa
71 ●1

updated 2019-03-02 12:52:24 -0600

__AmirRoth__

4436 ●5 ●17 http://bleedinggreenna...

Hello,

I am running a BEopt model on a remote server that I built for it to run quickly. The remote server has 8 VCPUS and 32 GB of memory and the CPU is running at 100%, yet the model hasn't even completed one iteration in 24 hours. Is there something I'm missing to make this run faster? At this speed, it would take over a month for the optimization to complete. Normally my multifamily models take a while, but not more than 10 days and with this high powered machine I would have anticipated it would be done after 4 days or less. I attached the model in question.

Thank You! Alisa

Comments

Do you happen to be modeling one stratified water heater per unit?

mdahlhausen ( 2019-02-28 12:37:40 -0600 )edit

Yes it is using the heat pump water heater option which includes a stratified tank. Unfortunately since tank stratification isn't something we can change within BEopt I'm not sure how to resolve this.

Alisa ( 2019-02-28 12:57:19 -0600 )edit

add a comment

answered 2019-02-28 13:05:06 -0600

Eric Wilson

1853 ●12 ●8 https://resstock.nrel.gov/

We are actively working on improving HPWH stratified tank runtime.

In the meantime, I would recommend running the optimization with an electric resistance water heater and then running a parametric to add in a HPWH to the point(s) of interest.

edit flag offensive delete link

Comments

You can follow the progress of this work at https://github.com/NREL/EnergyPlus/pu.... We expect this improvement to make it into EnergyPlus 9.1.0.

shorowit ( 2019-02-28 20:02:58 -0600 )edit

Thanks. I updated the model with this in mind, but am still having the same issue. My model has been running for 21 hours and is still on iteration 1 run 8. Is there something else that could be making it especially slow?

Alisa ( 2019-03-01 10:21:50 -0600 )edit

@Alisa, multifamily models with many units, smaller than normal timesteps (e.g., to support PCM analysis), and optimizations with many options selected could all be causing longer optimization times.

Eric Wilson ( 2019-03-01 10:32:58 -0600 )edit

I am already using 60 min timesteps to avoid this and it has been stuck on iteration 1 run 8 for 20 hours without making any progress (it did iteration 1 run 1-7 in the first hour). Is there any chance it's doing a calculation for this long? Seems unlikely, but I don't want to stop the run if that is the case and lose the 21 hours of work.

Alisa ( 2019-03-01 10:38:13 -0600 )edit

Will we ever get to see the benefits of these improvements? BEopt is currently stuck at EnergyPlus 8.8