Analysis of residential electricity consumption patterns utilizing smart-meter data: Dubai as a case study.

• Analysis of 15-minute sampled residential electricity consumption data of Dubai. • Households are classified using cooling consumption included in electricity bills. • K-Means clustering is used to group households based on their hourly consumption. • Consumption patterns helped to identify occupant behaviours in summer and winter. • Consumption patterns were cross-referenced to building type and occupant profiles. Analyzing residential load profiles and usage patterns is critical to making better decisions for demand-side management initiatives and designing strategies to get more people interested in energy savings. Therefore, analyzing load profiles, it is crucial to know how hourly consumption varies during summers, winters, weekdays, and weekends. In addition, determining the influence of occupancy, dwelling size, and building topologies on consumption is equally important to build predictive models. Therefore, this paper presents detailed research on the electricity consumption and profiles of the residential sector in Dubai based on the dwellings' characteristics and 15-minute resolution smart meter data. The data utilized includes the cooling systems, the number of occupants, and the physical characteristics of a dwelling, such as its typology, sizes, and number of bedrooms. First, using the K-Means clustering method, the authors grouped the households based on their consumption profiles. Second, the consumption patterns of each group of households were identified and organized based on similar consumption profiles over 24 h. Third, the authors applied several classification algorithms to assess the potential of using dwellings and occupants' characteristics to predict the patterns with which each household is associated. The analysis of consumption patterns showed that 43% of households with cooling included in their bills had the global peak demand at midnight during weekdays and weekends in summer. However, the global peak demand for cooling-excluded households occurs from 7:00 to 10:00 pm on weekdays and, in some specific cases, on the weekends, as early as 10:00 am. Finally, the methods used for classification were able to identify key characteristics driving the patterns of electricity demand and were well suited to this predictive modeling context. [ABSTRACT FROM AUTHOR]