Accelerating long-term building energy performance simulation with a reference day method

In response to the growing necessity for rapid simulation techniques for long-term building energy forecasting, this study establishes a ‘reference day’ method. This method significantly alleviates computational load in intricate simulation tasks by minimizing the redundancy of meteorological data. By employing a selected number of reference days to represent the meteorological profile over an extended period, we can estimate the total long-term energy consumption of buildings. The Finkelstein–Schafer statistic is utilized to identify these reference days. To evaluate the effectiveness of this proposed methodology, we analyzed sixteen prototype buildings—comprising two residential and fourteen commercial structures—and thirty years of meteorological data from Denver, USA. The findings indicate that the reference day approach effectively identifies days with representative weather conditions, enabling accurate energy consumption predictions while considerably reducing computational demands. Our case study suggests that selecting nine reference days strikes a good balance between predictive accuracy and computational efficiency over a long time span, even a 25-year period. In such a period, the margin of average error for predicting electricity and gas consumption was remarkably low, at −0.7% and −3.0%, respectively. It is important to note that the building’s operational schedule can significantly influence energy consumption. Hence, different sets of reference days should be designated for varied building operation categories. In summary, considering the high computational costs and lengthy durations of work associated with standard building simulations, our proposed reference day method could play a pivotal role in rapid energy consumption assessments. The efficacy and applicability of this method warrant further investigation.