An Improved Procedure for Developing Calibrated Hourly Simulation Models

In order to improve upon previous calibration techniques, this paper presents new calibration methods including a temperature bin analysis to improve hourly x-y scatter plots, a 24-hour weatherdaytype bin analysis to allow for the evaluation of hourly temperature and schedule dependent comparisons, and a 52-week bin analysis to facilitate the evaluation of long-term trends. In addition, architectural rendering is suggested as a means of verifying the building envelope dimensions and external shading placement. Several statistical methods are also reviewed to evaluate the goodness-offit including percent difference calculations, mean bias error (MBE), and the coefficient of variation of the root mean squared error (CV(RMSE)). The procedures are applied to a case study building located in Washington, D.C. Nine months of hourly whole-building electricity data and site-specific weather data were measured and used with the DOE2. ID building simulation program to test the new techniques. Use of the new calibration procedures were able to produce an hourly MBE of -0.7% and a CV(RMSE) of 23.1% which compare favorably with the most accurate hourly neural network models.