Deep Learning-Based Classification of Households for Domestic Consumption Balancing

Abstract

This thesis investigates the classification of household energy consumption using deep learning techniques, aiming to optimize energy management amidst rising demands and stagnant energy reserves in Algeria. The study begins by exploring deep learning fundamentals and progresses to contextualize energy consumption challenges. Methodologically, it focuses on dataset collection, preprocessing, and experimental setup involving REFIT and UK-DALE datasets. Results from classification experiments underscore the model's strengths and limitations in predicting consumption patterns. The research highlights the need for enhanced feature engineering, advanced time series techniques, and model refinements to overcome identified challenges. Ultimately, this work contributes to advancing energy efficiency and sustainability through innovative deep-learning applications in residential energy management.