HVAC energy improvements - best way to structure the tests?

[MattMcLeod]

Hi all,
I am working on a project to see whether we can improve energy consumption in a large building by switching off the HVAC overnight when attendance is low (its an airport terminal, 24/7 operation).

Clearly I can turn things off easily but I don't know how "hard" the system will have to run to "catch up" in the morning.

I have lots of data from the Building Management System logs so I can estimate electricity consumption via the plant room equipment that is running (I dpon't have enough meters to do it that way). If we make a change to the HVAC operation schedule we can log the same data points and collect plenty of measurements.

However, the ambient weather conditions are obviously a factor. If its "abnormally" hot or cold, I may not determine any difference in energy consumption if I just look at the "headline" figures.

So I think my problem statement is something like: "Energy Consumption after the HVAC schedule changes is lower PROVIDING there is no significant change to the ambient temperature profile".

If I have a distribution of energy figures before and after the HVAC schedule changes, I can test for differences in the means (for example Ho: Mean Energy After = Mean Energy Before). But how do I account for the ambient temperature?

Any ideas would be appreciated.

 

[GregLocock]

PMA (principal moments analysis) allows you to pull out the effect of different measured factors from noisy data. It needs to be used with caution.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[IRstuff]

Is that the same as Principal Components?

TTFN
faq731-376

Need help writing a question or understanding a reply? forum1529

 

[GregLocock]

Ack, yes, so far as I know.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[IRstuff]

PMA/PCA might be a bit of an overkill.

My naive thought is that your airport should already have ambient temperature and RH logs, so it's hypothetically possible to mash the power data with temp/RH data to create a surface plot of power, which should allow you to run small perturbation analyses, at least, and possibly, if you have sufficient data, to extrapolate to startup conditions

TTFN
faq731-376

Need help writing a question or understanding a reply? forum1529

 

[GregLocock]

You should always step back from analysis and do multi dimensional trend plots, Antaeus being the granddaddy of trend plotting.

This allows you to focus on the important factors, and allows you to check whether relationships are linear.

In the case of HVAC I think you'd be faced witha parabolic demand curve against external temperature, at some ambinet temp below 18 deg C you'd have a minimum power requirement, either side of that you'll be needing to cool or heat.



Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[MattMcLeod]

Thanks for your comments guys (and girls?).

Yes we have plenty of channels of data being sampled at 15 minutes intervals. I have no problem collecting data at all. Its just "how do I best use it?" without drawing the wrong conclusion.

Thanks again.