|Metrix » Tuning|
The following are a few general rules for tuning to weather:
Tune in multiples of 1 year. If you tune for 18 months, then you could be tuning for two summers and one winter, and as a result, the plethora of summer points could unfairly bias the best fit line.
Once you have selected the tuning period, the first and most important step is determining the meter's balance point. The heating balance point is the temperature at which the building starts to heat, while the cooling balance point is the temperature at which the building starts to cool. Do so with the outdoor temperature graph. While you tune, try to stay close to the balance point that you determined with the outdoor temperature graph.
The R2 value should never be less than 0.75. Any R2 value less than 0.75 belies a statistically invalid correlation. If you used poor correlations on a job, and if you were evaluated by a third party, you would have a difficult time explaining why you chose to adjust your baseline for weather when there was no clear relationship between weather and utility usage. (Personally, in my prior job as an energy analyst, I have been called to task by a third party consultant for proposing such poor regressions and it is not a pleasant experience. It also doesn't reflect well on your or your company's expertise in M&V.)
Finally, your T statistic for each variable should have an absolute value more than 2.0. If not, you have a poor correlation for that variable, and you should exclude it.
Metrix may exclude bills from the Tuning Period due to the minimum degree days per day value. You may exclude up to 7 points in this manner. However, if Metrix excludes more than 3 points, you should billmatch the excluded points.
If you choose to manually exclude bills from your tuning period, it would behoove you to have a reasonable justification for doing so (such as, the chiller was broken that month, and they rented an inefficient backup chiller). If you do not have reasonable justifications for doing so, again, you may be called to task by either the customer or a third party consultant. In general, the more bills you exclude (even though the R2 value may increase) the more questionable the correlation between weather and utility usage. If the manually excluded bills are due to some extraordinary circumstance (such as equipment breakdown), then you need not billmatch the manually excluded bills.
You may include negative baseloads. However, if you do so, you should probably billmatch the deselected points (below the degree day/day threshold).
If you cannot meet the above criteria for a good fit, you should use bill-matching or look for other variables to tune with.
Otherwise feel free to contact our Tech Support staff at (805) 329-6565, or via email at firstname.lastname@example.org.