Ομαδοποίηση οικιακών φορτίων ηλεκτρικής ενέργειας για την διαχείριση ζήτησης με αλγορίθμους μηχανικής μάθησης

Abstract

This research focuses on analyzing household energy consumption patterns to enhance demand-response (DR) strategies, essential for optimizing energy use relative to supply conditions. Analyzing data from 5,567 London households participating in the UK Power Networks' Low Carbon London project, we processed and engineered the data to reveal nuanced energy behaviors. We evaluated various machine learning clustering algorithms, including K-Means++, Fuzzy C-means, Hierarchical clustering, SOMs, BIRCH, GMMs, Spectral Clustering, and notably, Ensemble Clustering. Using metrics such as the Silhouette Score and Davies-Bouldin Score, we identified distinctive energy consumption patterns that facilitate the tailoring of DR strategies. Our findings highlight the effectiveness of Ensemble Clustering and the role of Explainable AI (XAI) in providing deeper insights into energy use, suggesting future research into sociodemographic influences and XAI methodologies for refined energy management.