|Metrix » Tuning|
Last Revised: September 29, 1999
SRC Systems' recommendations regarding tuning have changed in accordance with the latest thinking in ASHRAE's GPC-14P Monitoring and Verification Committee.
Below are an up-to-date set of Tuning Guidelines:
Determination of M&V Approach
- To use the Whole Building approach (Utility Bill Analysis), expected savings should exceed 10% of measured energy use or demand.
- Tune bills for periods in multiples of roughly 365 days.
- Use the outdoor temperature graph to determine balance point. Balance point should not deviate by more than a few degrees from the temperature indicated in the outdoor temperature graph.
- The cooling balance point, however, should still always be below the indoor setpoint or below the economizer setpoint, if any.
- Energy, Demand, and different TOU categories do not need to have the same balance point temperature.
Minimum Degree Days / Day and Excluded Points
- Great care should be taken in selecting an appropriate minimum degree-day / day setting. See the topic: How do you select an appropriate minimum degree-day / day setting?
- Tunings that include more than one variable should have minimum degree day / day settings of 0.
- Deselect as few points as possible. However you may exclude up to 7 points out of 12 (via the minimum degree-day / day setting).
- If more than 3 points are deselected due to the min DD/day setting, then all excluded points should be billmatched with a replicating modification.
- If many points are excluded, make sure that the remaining points span both mild and severe seasons.
- Excluded points should preferably be in sequential order (i.e., Dec., Jan., Feb.).
Manually Excluded Points
- Points should only be manually excluded if there is a reason for it (one-time, anomalous conditions). Technically, the values of R2 & T should not be used as sole criteria to decide if a point is "convenient" or not.
- If you manually exclude a point, determine and document why the data does not follow the trend, (i.e., bldg. was evacuated, meter was broken). Manually excluded points for one-time, anomalous conditions need not be billmatched with a replicating modification.
- R2 value > 0.75
- |T statistic| >2
- |Mean Bias| < 0.5% (this may change)
- |CV(RMSE)| < 25% for energy and |CV(RMSE)| < 35% for demand (this may change)
- For one variable regressions, baseload (constant * #days) values should be close to bill values in which independent variable = 0.
- Tuning baseload coefficient can be negative (if you use bill-matching modification for deselected points).
- When in doubt, include fewer Variables.
- User-Defined Variables should be normally distributed.
- User-Defined Variables should be independent of each other.
- Highs and Lows of a User-Defined Variable should span at least 2 to 1.
Otherwise feel free to contact our Tech Support staff at (805) 329-6565, or via email at email@example.com.