Calibration Criteria Calibration Calibrated Index Limit Calibrated Type Parameter Model IPMVP FEMP Error, (2002) (2008) % Hourly Cooling energy NMBE
-- 10 3 CvRMSE 20 30 15 Hourly Heat transfer NMBE
-- -- 2 coefficient CvRMSE -- -- 8 Hourly Panel surface NMBE
-- -- -4 temperature CvRMSE -- -- 6 Table 4.
Statistical indicators for Monthly Calibration of the Sample Buildings Sample Building NMBE
, % CVRMSE, % 1 -2.
24% and an NMBE
of 11%, while the default model has a CVRMSE of 14.
We have attempted to reduce the scatter by plotting the percentage savings not against a combined kWh-kW statistic (namely, the GOF_Total) but against the GOF_kWh and GOF_kW (for example, GOF_kWh includes only the NMBE
and CV of the calibrated simulation trial with respect to kWh).
For the charging case, the fit results in a NMBE of 0.
The NMBE and the CV(RMSE) are defined as follows (IBPSA 2015):
The models of the carpet floored room are not as accurate, especially at the peaks during ramping profiles, as seen in Figures 9 and 10b, yet the statistical indices are within the reasonable limits (CV-RMSE below 30% and NMBE
ASHRAE Guidelin 14 ASHRAE 2002) requires calibrated building energy models to have an error rate below specified thresholds in terms of both CV(RMSE) and NMBE
to meet guideline requirements.
The CV and NMBE
results for the WBE energy channel are plotted in Figures 2 and 3 for the large hotel building in Chicago, IL.
A model that has a low NMBE
(%) can accurately predict the total energy consumption even if the prediction CV (%) is high.
= [MBE(residual)/Average_measured_energy] x 100% (6)
describes the variation between the mean measured and predicted values while the CVRMSE describes the variation in the pattern of the data.