National Technical University of Athens (NTUA)
Data-driven methods for the integration of flexible loads in demand-side management schemes
Abstract
dc:descriptionThe transition towards a sustainable and resilient energy system in the European Union (EU) has accelerated the integration of Distributed Energy Resources (DERs), including renewable energy sources (RES), energy storage systems, and electric vehicles (EVs). While this shift is essential for achieving the EU’s ambitious climate objectives, it also introduces significant challenges to maintaining grid stability and reliability. Demand-Side Management (DSM) emerges as a critical strategy for balancing supply and demand by leveraging flexible loads. This doctoral dissertation introduces data-driven methods to enhance the integration of flexible loads into DSM schemes, aligning with the goals of the EU Horizon 2020 GECKO project, which aims to foster responsible and transparent Artificial Intelligence (AI) applications in the green energy transition. The doctoral dissertation investigates three primary questions: (1) How can data from smart meters and Internet of Things (IoT) devices be effectively utilized to propose optimal load scheduling strategies in connected home environments? (2) How do different data-driven methods, including statistical and AI-based models, perform in forecasting aggregated load for quantifying flexibility? (3) What strategies can maximize residential flexibility through efficient aggregation schemes in DR programs and novel energy markets? A multi-faceted approach is adopted, beginning with a comprehensive review of Particle Swarm Optimization (PSO) methods tailored for residential energy management. The review introduces a novel classification framework for PSO applications, emphasizing critical optimization objectives such as minimizing energy costs, maintaining user comfort, and enhancing thermal efficiency. Building on these insights, the dissertation develops two innovative smart EV charging strategies using Deep Q-Network (DQN) models. The proposed strategies offer end-users a choice between a "cost savings" policy, which prioritizes reducing electricity bills, and a "user-oriented" policy that aligns with historical charging habits. The latter approach achieves notable results, with cost savings of up to 49.83% while maintaining an 86.22% consistency with historical charging patterns. To address the challenge of Very Short-Term Load Forecasting (VSTLF) at the appliance level, a Long Short-Term Memory (LSTM) model with Feed-Forward Error Correction (FFEC) is proposed. The model predicts 15-minute aggregated loads for EVs and air conditioning (A/C) units. Benchmarking against state-of-the-art (SoA) methods such as XGBoost, LightGBM and N-BEATS, the LSTM with FFEC demonstrates superior accuracy in EV load forecasting, contributing to improved load management and flexibility quantification. LightGBM and XGBoost slightly outperform LSTM with FFEC in A/C load forecasting, demonstrating the highest prediction accuracy among the investigated models. The research further proposes a "Fair Demand Response (DR)" framework to enhance residential participation in DSM programs. The fairness-driven selection strategy ensures a balanced inclusion of participants, promoting reliability in delivering flexibility services. This approach addresses the limitations of traditional performance-based selection methods, which often concentrate opportunities among a small group of participants. In advancing the concept of capacity reserve markets, the dissertation explores the participation of EV aggregators in a newly proposed energy reserves market product. A mathematical formulation of energy reserves maximization for EV aggregators is proposed, taking into consideration EV users’ preferences. A novel Six-Criteria selection approach is introduced, enabling System Operators (SOs) to efficiently preselect aggregators, thereby minimizing the administrative burden associated with evaluating and dispatching all available flexibility providers. This methodology not only optimizes operational efficiency but also contributes to a more stable and balanced energy system. Overall, this dissertation highlights the potential of data-driven and AI-enhanced demand-side management strategies to unlock residential flexibility and facilitate the broader adoption of smart grid solutions in the energy transition. By combining technical innovation with principles of fairness, inclusivity and energy markets efficiency, the proposed methods offer a robust foundation for future research and practical implementations in the evolving energy landscape.
Degree
thesis:*- Grantor dc:publisher
- National Technical University of Athens (NTUA)
- Year dc:date
- 2025
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Menos-Aikateriniadis, Christoforos
- Μένος-Αικατερινιάδης, Χριστόφορος
Subjects
dc:subject × 50- Διαχείριση ζήτησης
- Απόκριση ζήτησης
- Οικιακή ευελιξία
- Τεχνητή νοημοσύνη
- Μηχανική μάθηση
- Βαθιά ενισχυτική μάθηση
- Μακρά βραχυπρόθεσμη μνήμη
- Πολύ βραχυπρόθεσμη πρόβλεψη φορτίου
- Αθροιστικό ευέλικτο φορτίο
- Ενεργειακή εφεδρεία
- Μέγιστη ποσοτικοποίηση εφεδρείας
- Ηλεκτρικά οχήματα
- Βέλτιστος προγραμματισμός φόρτισης ηλεκτρικών οχημάτων
- Ιδιοκατανάλωση
- Προτιμήσεις τελικών χρηστών
- Φορείς σωρευτικής εκπροσώπησης
- Μέθοδος φιλτραρίσματος
- Κατανεμημένοι ενεργειακοί πόροι
- Έξυπνο δίκτυο
- Δίκαια επιλογή συμμετεχόντων
- Demand-Side Management
- Demand response
- Residential flexibility
- Artificial Intelligence (AI)
- Machine learning (ML)
- Deep reinforcement learning (DRL)
- Long short term memory (LSTM)
- Very short-term load forecasting (VSTLF)
- Aggregate flexible load
- Energy reserves
- Reserves maximization
- Electric vehicles (EVs)
- EV optimal scheduling
- Self-consumption
- End-user preferences
- EV aggregators
- Filtering approach
- Distributed energy resources (DER)
- Smart grid
- DR selection fairness
- Περιβαλλοντική Μηχανική
- Επιστήμες Μηχανικού και Τεχνολογία
- Πράσινη και Αειφόρα επιστήμη και Τεχνολογία
- Επιστήμη Ηλεκτρολόγου Μηχανικού, Ηλεκτρονικού Μηχανικού, Μηχανικού Η/Υ
- Υπολογιστές, Υλικό (hardware) και Αρχιτεκτονική
- Environmental Engineering
- Engineering and Technology
- Green and Sustainable Science and Technology
- Electrical Engineering, Electronic Engineering, Information Engineering
- Computer science, Hardware and Architecture
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/59288