An Improved Procedure for Developing Calibrated Hourly Simulation Models

T. E. Bou-Saada, J. S. Haberl

OAKTRUST (Texas A&M University) · 1995 · 39 citations · 7 references

Concepts

Abstract

In order to improve upon previous calibration
\ntechniques, this paper presents new calibration
\nmethods including a temperature bin analysis to
\nimprove hourly x-y scatter plots, a 24-hour weatherdaytype
\nbin analysis to allow for the evaluation of
\nhourly temperature and schedule dependent
\ncomparisons, and a 52-week bin analysis to facilitate
\nthe evaluation of long-term trends. In addition,
\narchitectural rendering is suggested as a means of
\nverifying the building envelope dimensions and
\nexternal shading placement. Several statistical
\nmethods are also reviewed to evaluate the goodness-offit
\nincluding percent difference calculations, mean bias
\nerror (MBE), and the coefficient of variation of the
\nroot mean squared error (CV(RMSE)).
\nThe procedures are applied to a case study
\nbuilding located in Washington, D.C. Nine months of
\nhourly whole-building electricity data and site-specific
\nweather data were measured and used with the DOE-
\n2. ID building simulation program to test the new
\ntechniques. Use of the new calibration procedures
\nwere able to produce an hourly MBE of -0.7% and a
\nCV(RMSE) of 23.1% which compare favorably with
\nthe most accurate hourly neural network models.

References

7