Publications
- A forecast-driven, comfort-aware load shifting optimization tool for optimizing solar self-consumption in island energy communities
Date: 2025
Publication type: Conference Paper
Author(s): Efstathios Sarantinopoulos, Vasilis Michalakopoulos, Nektarios Matsagkos, Eleni Kanellou, Elissaios Sarmas, Vangelis Marinakis
Abstract: Remote and non-interconnected island communities face persistent energy management challenges due to geographic isolation, constrained infrastructure, and dependence on imported fossil fuels. The integration of renewable energy sources (RES), particularly photovoltaics (PV), presents a promising pathway toward energy autonomy. However, the intermittent nature of solar energy complicates its efficient utilization. This study introduces a novel self-consumption optimization framework designed to facilitate intelligent load shifting at both the community and household levels, with a specific focus on the island of Chalki, Greece. The proposed system integrates PV generation and electricity consumption forecasting using Long Short-Term Memory (LSTM) models, coupled with non-intrusive load monitoring (NILM) techniques for disaggregating coolingrelated demand. To ensure user comfort, thermal modeling is employed using customized Predicted Mean Vote–Predicted Percentage Dissatisfied (PMV–PPD) metrics. At the community scale, a heuristic algorithm identifies optimal time windows for load shifting based on projected PV surplus. Concurrently, household-level cooling demand is managed through a multiobjective optimization approach that balances thermal comfort with the availability of renewable generation. The framework is implemented as a user-centric web application, providing both a centralized energy management dashboard for the community and individualized control tools for end-users. The system aims to enhance renewable energy self-consumption and promote sustainable, resilient energy practices in insular communities. Index Terms—Photovoltaic systems, self-consumption optimization, energy management, load shifting, renewable energy, thermal comfort, machine learning, island microgrids
Link
- A Key Performance Indicator Framework for Activated and Data-Driven Energy Communities
Date: 2025
Publication type: Conference Paper
Author(s): Nektarios Matsagkos, Efstathios Sarantinopoulos, Vasilis Michalakopoulos, Eleni Kanellou, Vangelis Marinakis
Abstract: A series of unprecedented energy and geopolitical shocks led the European Union to accelerate its commitment to a just green energy transition. Efforts for diversification of energy supply, shifting from centralised to decentralised energy systems, and increasing the Renewable Energy Sources while reducing the reliance on fossil fuels, led to the upgrowth of energy communities. Within this evolving landscape, energy communities have emerged not only as tools for enhancing energy resilience and sustainability but also for citizen empowerment and energy democracy. The aim of this work is to introduce a list of 25 indicators in order to evaluate and assess the impact of implemented services and strategies by different energy communities, and thus the impact of these communities, from a social, economic/technical and environmental perspective. The Key Performance Indicator framework is applied to four diverse energy communities from different European Union member states, each with distinct characteristics and missions. This research offers insights into the understanding of energy communities' evaluation but also underlines the importance of refining Key Performance Indicators to ensure holistic evaluation.
- An Integrated ML-Ops Framework for Automating AI-based Photovoltaic Forecasting
Date: 2023
Publication type: Conference Paper
Author(s): Symeon Chorozoglou, Elissaios Sarmas, Vangelis Marinakis
Abstract: Energy digitization holds significant importance for various energy applications, encompassing aspects like production, consumption, and distribution within power grids. The digital transformation of energy plays a pivotal role in enhancing the integration of Artificial Intelligence (AI) into energy management systems, leveraging extensive datasets. The development of AI systems and the utilization of Machine Learning (ML) techniques empower users with precise predictions related to renewable energy production, thereby expediting the shift towards clean energy. Nevertheless, the effective use of data demands a high level of expertise, thereby excluding energy stakeholders from the benefits modern technologies offer. In this paper, we introduce an AI forecasting system designed to bridge the knowledge gap in data processing methods and ML models for energy stakeholders. This system focuses on delivering a user-friendly interface for photovoltaic (PV) production forecasting by automating the entire ML operations pipeline. Consequently, users can obtain data-driven model results without the need to manually code all the requisite steps for model training and fine-tuning. To demonstrate the system’s capabilities, we provide an experimental application using real PV data from a Portuguese aggregator.
- Democratise Energy Through Energy Activated Citizens and Data-Driven Communities: The ENPOWER Approach
Date: 2024
Publication type: Conference Paper
Author(s): Nektarios Matsagkos, Eleni Kanellou, Afroditi Fragkiadaki, Vasilis Michalakopoulos, Vangelis Marinakis, Haris Doukas
Abstract: The need to ensure a just energy transition and shift towards decentralised energy systems is becoming crucial. The European Commission outlines the need for energy independence, based on two key pillars: (a) diversifying gas supplies; (b) reducing faster the use of fossil fuels in homes, buildings, industry, and power system, by increasing energy efficiency gains and renewables. Energy communities can be a pivotal mechanism in achieving this transition. However, the current social practices, technologies adoption and business models are not sufficiently mature to enable energy citizenship's practices scaling up and replication. There is a need to transform traditional passive energy consumers into active energy citizens enabling them to take full control of their energy usage. To do so engaging with all the principal actors of the energy value chain, and allowing them to achieve energy savings is needed, while increasing their energy efficiency and self-consumption optimisation and transforming energy consumers and communities to digitally enhanced and grid-friendly ones.
- Energy allocation and settlement in collective selfconsumption
Date: 2024
Publication type: Conference Paper
Author(s): João Mello, Luís Rodrigues, José Villar, João Saraiva
Abstract: Energy allocation rules are one of the core aspects of collective self-consumption (CSC) regulations. It allows final consumers to share their surplus generation with other CSC members, while keeping their full rights as consumers, i.e., maintaining a supply contract with the retailers of their choice. Some European Union member states regulations use allocation coefficients so that local allocations are integrated with wholesale settlement and directly affect the retailers’ billing. Several AC methods have been proposed so far, each one adapted to distribution system operators’ settlement procedures with specific rules that can impact the benefits that each CSC member obtain. This paper analyses, assesses and compares two relevant AC methods, namely pre-delivery fixed AC and post-delivery dynamic AC, by developing a settlement formulation for a community with members with flexible assets and different opportunity costs. AC policy recommendations based on findings are provided.
- Explainable AI-Based Ensemble Clustering for Load Profiling and Demand Response
Date: 2024
Publication type: Journal article
Author(s): Elissaios Sarmas, Afroditi Frangkiadaki, Vangelis Marinakis
Abstract: Smart meter data provide an in-depth perspective on household energy usage. This research leverages on such data to enhance demand response (DR) programs through a novel application of ensemble clustering. Despite its promising capabilities, our literature review identified a notable under-utilization of ensemble clustering in this domain. To address this shortcoming, we applied an advanced ensemble clustering method and compared its performance with traditional algorithms, namely, K-Means++, fuzzy K-Means, Hierarchical Agglomerative Clustering, Spectral Clustering, Gaussian Mixture Models (GMMs), BIRCH, and Self-Organizing Maps (SOMs), across a dataset of 5567 households for a range of cluster counts from three to nine. The performance of these algorithms was assessed using an extensive set of evaluation metrics, including the Silhouette Score, the Davies–Bouldin Score, the Calinski–Harabasz Score, and the Dunn Index. Notably, while ensemble clustering often ranked among the top performers, it did not consistently surpass all individual algorithms, indicating its potential for further optimization. Unlike approaches that seek the algorithmically optimal number of clusters, our method proposes a practical six-cluster solution designed to meet the operational needs of utility providers. For this case, the best performing algorithm according to the evaluation metrics was ensemble clustering. This study is further enhanced by integrating Explainable AI (xAI) techniques, which improve the interpretability and transparency of our clustering results.
- DATA ACCESS AND EFFECTIVE SOCIAL ENGAGEMENT: A PATHWAY TO RESILIENT AND CONSUMER-ACTIVATED COMMUNITIES
Date: 2025
Publication type: Conference paper
Author(s): Mário Couto, Mariana Jimenez, Peter Richardson, Alessio Coccia, Alexandra Revez, Julia Blanke, Deirdre de Bhailís, Jonathan Sandham, John Walsh
Abstract: Social engagement and data sharing have a critical role in transitioning to sustainable energy systems. Leveraging the work developed in the ENPOWER and Crete Valley projects, this research focuses on the Dingle Peninsula, where smart meter data is used to characterize energy behaviour and identify opportunities for load shifting and optimizing renewable energy use. Firstly, two interlinked Data-Sharing Workshops were conducted with local representatives and community members. Participants highlighted the importance of financial benefits, community independence from major energy providers, and contributions to the local sustainability. Concerns about data handling and privacy were also raised. Secondly, towards the energy behaviour characterization of the local community, historical active power measurements from 49 household smart meters, covering various periods between 2019 and 2022, were analysed. Households were categorized based on their behind-the-meter assets. Also, Monte Carlo simulations were applied to model electric vehicle (EV) charging patterns, accounting for variability and uncertainty, by using the data collected through the EV infrastructure available in Dingle.
- A hyperparameter-space clustering methodology of residential electricity loads
Date: 2025
Publication type: Journal article
Author(s): Vasilis Michalakopoulos, Ioannis Papias, Efstathios Sarantinopoulos , Elissaios Sarmas, Vangelis Marinakis, Dimitris Askounis
Abstract: Clustering residential electricity consumption patterns is a crucial step toward scalable and interpretable energy forecasting. Traditionally, clustering relies on direct analysis of load time series, which may obscure deeper behavioral or structural similarities between households. This study introduces a novel approach that shifts the focus from using raw consumption data to using the hyperparameters of stacked hour-ahead deep learning forecasting models—specifically based on LSTM, Bi-LSTM, and GRU architectures. By optimizing models independently for each household and clustering based on the resulting hyperparameter configurations, we reveal functional similarities in forecasting behavior that are not always evident in the original data. This method enables the formation of new, interpretable consumer segments, which can support the development of tailored forecasting models per cluster. Utilizing over three years of data from 200 households located in London, we evaluate the proposed hyperparameter-based clustering against traditional time-series clustering, applying standard metrics and explainable AI techniques (SHAP and LIME) to interpret the results. Findings demonstrate a strong alignment between the mean and median consumption patterns of clusters for both approaches, validating the effectiveness of the proposed method. Moreover, the analysis highlights that the feature transformation layers play the most critical role in shaping cluster formation, underscoring its significance in capturing underlying consumption behaviors. Beyond its analytical value, the method also offers a privacy-preserving advantage by requiring only the exchange of model hyperparameters rather than raw consumption data.
