A randomised approach to multiple regression analysis of building energy simulation

Multiple regression analysis (MRA) is useful for developing energy prediction equations from the results of building energy simulation. However, if many design parameters are involved, a very large number of simulations is needed to generate data for the MRA. To tackle this problem, a randomised approach to MRA is proposed so that less simulations will be needed to generate the data. This paper presents a research study in Hong Kong that investigates building energy performance using a randomised MRA method. A parametric study has been conducted using the DOE-2.1E building energy simulation program to derive a set of energy equations relating 12 design parameters. The parameters are selected from a sensitivity analysis and are taken from three groups: load, system and plant. It is found that the energy equations generated from a set of 100 randomised simulations can do energy predictions quite well. It is believed that this approach has good potential for reducing the efforts required for MRA and making MRA easier for average building design professionals.