Validation of methodology for utility demand prediction considering actual variations in inhabitant behaviour schedules

Abstract

A data set of myriad and time-varying inhabitant behaviour schedules with a 15-min time resolution, generated by the authors in a previous study, is validated through a comparison analysis. The key idea of generating a set of raw schedule data from the restricted statistical information is called the ‘generate and kill’ concept, which is commonly used in the fields of artificial intelligence and multi-agent simulation. In the present study, we show three comparisons. The first and second compare the estimated demand with a time series of measured utility demand. These comparisons indicate that the generated data and the algorithm, as described by the authors, have the required robustness. Another comparison between the estimate and the annually averaged daily water demand of a residential area, consisting of 9327 residences, also shows an acceptable consistency.