Understanding how people use electricity in their homes is vital for determining their consumption requirements, improving energy management (the demand side), and supporting policies to control carbon emissions. This paper presents a unified Largest Triangle Three Buckets-Dynamic Time Warping-K-medoids (LTTB-DTW-Kmedoids) model for identifying household electricity consumption patterns using smart meter data. The suggested approach uses the LTTB algorithm to downsample high-resolution time series data, maintaining key load properties; the DTW algorithm to accurately estimate the similarity of electricity consumption sequences; and the K-medoids algorithm to generate robust, noise-resistant clusters. The 1000 households are clustered using the elbow criterion, based on one month of electricity consumption data from the Irish Intelligent Metering Customer Behaviour Trial. The research paper proposes a universal household electricity profiling technique and highlights the natural structural features of household electricity consumption, rather than relying on external or socioeconomic factors. Comparative studies show that the proposed model consistently outperforms benchmark models, achieving better clustering results as measured by the Calinski-Harabasz Index (CHI), Silhouette Index (SIL) and Davies-Bouldin Index (DBI). These findings demonstrate that the developed framework is a robust and realistic tool for establishing residential electricity consumption trends. • Proposes an integrated LTTB–DTW–Kmedoids framework for household electricity profiling. • Preserves key load characteristics while efficiently reducing smart meter data dimensionality. • Identifies robust and noise-resistant consumption patterns using sequence-based similarity. • Clusters 1000 households using real-world smart meter data without socioeconomic inputs. • Outperforms benchmark methods across multiple clustering validation indices.
Sun et al. (Sat,) studied this question.