This paper describes the development of a load prediction algorithm for the energy demand of a large, commercial, all-electric building during a summer cooling season. This load-prediction algorithm is based on an extensive multiple linear regression analysis of the independent variables influencing the HVAC system, and it enables a building operator to predict energy consumption and peak usage up to four hours in advance. Application of several sets of statistical criteria yielded two roughly equivalent four-variable load-prediction models, each having an accuracy of plus or minus 2.5 percent for electrical demands predicted three to four hours in advance.

The load-prediction algorithm forms an integral component of an adaptive control scheme for building HVAC systems. The adaptive control strategies utilize the predicted loads to modify interactively any of the well-known building control processes and thereby minimize building energy costs.