Login to EVO

Questions & Answers

Q: I'm working on a project where we're looking at an Option C for electricity at a manufacturing site where the ECMs are all focussed around HVAC / lighting. There are so many changing product lines (different lines in baseline and reporting period!) that it's difficult find a consistent and reliable production variable, but, we do have sub-meter data for the production. Is it theoretically possible to use the production sub meter data as an independent variable for the whole site data? The thought is that the activity of the production lines has an impact on the HVAC / lighting loads (as well as other variables - degree days etc.) and so we're just using energy consumption of the production lines as a proxy for production activity. However, it doesn't feel comfortable to us as the sub-meter data is not independent of the dependent variable (i.e. the whole site data)!! Does anyone have any views / experience with this? Is this a regression analysis faux pas? Very grateful for any thoughts.

A:  I agree that the method you describe is not quite typical. It is more common to use a production metric as the independent variable. That said, there are definitely times when no clean production variable is available. It may not be recorded or product lines may change so quickly that there is no common element. In those cases, I think you have hit upon a clever and acceptable approach. It works in this case because you are not attempting to estimate savings on the production lines.

You say that “the sub-meter data is not independent of the dependent variable (i.e. the whole site data)”. That would be true if the dependent variable was the whole facility use. However, in your case I think you can isolate on the HVAC and Lighting loads. I recommend subtracting the submetered use from the whole facility use. That will give you the Lighting and HVAC use, plus some noise. Try regressing that net load against your production variable (kWh) and whatever other variables you are considering.

One uncertainty to consider is how much (if any) non-production load is actually served by the "production" submeter? If any of your lighting or HVAC is on that meter, you will miss those savings.

Q: I am currently working on the reduction of electrical or thermal power in building rehabilitation. I would have liked to know if you had any documentation on this subject. In particular, either feedback on similar actions or methods of measurement and verification.

A: Assuming that "building rehabilitation" means significant changes and modernization or improvements to the building and/or systems. (In the United States, this is referred to as "deep retrofits" which is not any clearer, in my opinion.) If that is the correct meaning, then a likely approach for M&V would be IPVMP Option C. Software that can help with the analysis for IPMVP Option C is ECAM, which is available in French as well as English. Contact Paul Calberg-Ellen at Biomasse Normandie or Sidonie Michel at Observatoire Régional Energie Climat Air Normandie to find out how to obtain the French version of ECAM.

Q - I am making a multiple regression for C EVO model. Constant values (m1,m2 and m3) according to the equation y(consumption)= m1(days)+m2(activity hours)+ m3(HDD)+b results with negative values; What should I do?

1.- If constant b<0 ; is it possible the building had negative consumption if no other variables affect?
2.- If constant m1<0
3.- If constant m2<0
4.- If constant m3<0

What is the procedure in these cases?

A - Unless there is energy production (e.g. solar panels) within the measurement boundary, the estimated consumption should never be less than zero.

The coefficients should always make physical sense. Without know the full meeting of your equation, I would expect m1 and m2 to be positive. For example, it wouldn't usually make sense that energy use could go down as the number of days increased.

m3 should usually be positive. Possible issues include an incorrect base temperature for HDD or allowing negative values for HDD.

Last, but not of least importance, you should not be extrapolating outside the range of independent variables in your data. If you extrapolate, in some cases you can get negative estimates of energy use. For high frequency data (e.g. daily or hourly) we are having discussions about if or when minor extrapolation could be allowed, but right now IPMVP does not allow extrapolation. (It is certainly possible that a reporting period could have a month with slightly more (or fewer) HDD than was present in any month in the baseline even with a full year baseline, and so even at present it might make sense to have VERY MINOR extrapolation, but it should be avoided.

The CMVP course online is currently available only for US citizens. For all other countries the course is available only on site. In both cases, the CMVP exam is to be taken on site and cannot be done online.

The CMVP exam is not available online. You should attend the exam session after any of the trainings but you should register in advance. You are allowed to attend the exam within three years after the training.