Traitement des données de consommation d’électricité acquis d’un dispositif autonome de mesure préalablement conçu en vue de mieux gérer la demande

Abstract

This document provides an in-depth analysis and management of electricity consumption data collected by an autonomous device within a condominium equipped with electric vehicle (EV) charging stations. The literature review, covered in the first chapter, provides a comprehensive synthesis of the theoretical foundations related to time series, data classification, and anomaly detection, spotlighting a wide array of methodological approaches such as the use of auto encoders, the DBSCAN algorithm, and LSTM neural networks, which prove particularly relevant in the context of energy consumption. The methodology chapter outlines a structured and rigorous approach: beginning with data enrichment using contextual variables, followed by advanced anomaly detection techniques, including MDWS, auto-encoders, DBSCAN, and Isolation Forest, and subsequent correction using the K-nearest neighbor’s method. This methodology further extends to forecasting through models such as LSTM, SARIMA, and stacking, providing a holistic framework to address energy-related challenges. The case study, detailed in a subsequent chapter, focuses on a meticulous examination of data from three distinct condominium units. The results, presented and discussed in a dedicated chapter, highlight the varied performance of different detection and prediction techniques, with recognition of the effectiveness of specific methods in capturing irregularities and anticipating consumption trends. The ensuing discussions explore the profound influence of electric vehicles on consumption profiles, which are identified as a dominant factor in usage peaks, and pave the way for strategic recommendations, such as implementing dynamic pricing tailored to peak-demand hours, alongside a thoughtful expansion of charging infrastructure to support a sustainable and balanced energy transition.

Date	27 Feb 2026
Original language	French
Awarding Institution	École de technologie supérieure
Supervisor	Julio Cesar Montecinos (Supervisor) & Tony Wong (Co-supervisor)