324 Energy Project Financing: Resources and Strategies for Success
least to the daily level to control the number of independent variables
required to produce a reasonable model of the baseyear, without signifi-
cant impact on the uncertainty in computed savings (Katipamula 1996,
Kissock et al. 1992). Scatter found in daily data is often attributable to
the weekly cycle of most facilities.
Many models appropriate for Option C are possible. To select the
one most suited to the application, statistical evaluation indices should
be considered, such as R^2 or CV (RMSE) (see Appendix B). Additional
information concerning these selection procedures can be found in
Reynolds and Fels (19 88), Kissock et al. (1992, 1994) and in the ASHRAE
Handbook of Fundamentals (1997).
Statistical validity of the selected model should be assessed and
demonstrated by reference to published statistical literature.
In certain types of facilities (such as schools) where there is a sig-
nificant difference between the facility’s energy use during the school
year and summer break, separate regression models may need to be devel-
oped for different usage periods (Landman and Haberl 1996a; 1996b).
3.4.3.5 Option C: Computation of Routine Adjustments
The following steps are used to calculate the Adjustments term in
Equation 1 for Option C.
1 Develop the appropriate model for the baseyear energy data and
selected significant driving conditions (see Chapter 3.4.3.2 and
Chapter 3.4.3.3).
2 Insert the post-retrofit period’s independent variables (e.g. ambi-
ent temperature, metering period length) into the baseyear model
from 1, above. This process derives the energy use that would
have happened under post-retrofit conditions if the ECM had not
been installed. (Note if some other set of conditions is selected
for reporting savings (Chapter 3) the independent variables for
this set of conditions would be used in place of the post-retrofit
independent variables.)
3 Subtract the baseyear’s energy use from the result of 2, above, for
each month.
3.4.3.6 Option C: Cost
The cost of Option C methods depends on whether the energy