Assessing the Applicability of the Bartlett-Lewis Model in Simulating Residential Water Demands

This paper presents the set-up and application of the Bartlett-Lewis clustering mechanism to simulate residential water demand at fine, i.e. sub-hourly, time scales. Two different variants of the model, i.e., the original and the random-parameter model, are examined. The models are assessed in terms of preserving the main statistical characteristics and temporal properties of demand series at a range of fine time scales, i.e., from 1-min up to 15-min. The comparison against the typical Poisson rectangular pulse model showed that clustering mechanism enables a better reproduction of demand characteristics at levels of aggregation other than those used in the fitting procedure.