Seasonal autoregressive modelling of water and fuel consumptions in buildings

In order to account for variations in the performances of buildings, it is necessary to construct explanatory models of water and energy consumptions. In this paper, a modelling approach is illustrated for those parts of the variances in consumptions of water and energy that are neglected in conventional monitoring and targeting procedures. It is shown that these parts of the consumption variance need not be random and that identifying an autoregressive component can generate better models. Additionally, conventional procedures do not exploit the seasonality that is common in many buildings. Such improved models permit the more reliable detection of significant changes in a building's performance, and more accurate estimations of the effects of changes, whether the result of plant faults or operator intervention.