Publications

The opportunities created in the Greek electric vehicle (EV) market have allowed potential investors to participate in the development of the EV charging infrastructure of the country. Yet, the decision process of relevant stakeholders for strategic investments is challenging, involving the identification of the most promising charging sites from a set of multiple alternative locations of various features that may significantly affect their business competitive advantage. This paper attempts to facilitate decision making in such settings using a comprehensive, yet thorough multi-criteria decision analysis framework. The proposed approach is validated considering ten Greek municipalities of different characteristics. The results show- case the overall strengths of the proposed approach and its utility in the strategic planning process of potential investors.

Energy behaviours will play a key role in decarbonising the building sector but require the provision of tailored insights to assist occupants to reduce their energy use. Energy disaggregation has been proposed to provide such information on the appliance level without needing a smart meter plugged in to each load. However, the use of public datasets with pre-collected data employed for energy disaggregation is associated with limitations regarding its compatibility with random households, while gathering data on the ground still requires extensive, and hitherto under-deployed, equipment and time commitments. Going beyond these two approaches, here, we propose a novel data acquisition protocol based on multiplexing appliances’ signals to create an artificial database for energy disaggregation implementations tailored to each household and dedicated to performing under conditions of time and equipment constraints, requiring that only one smart meter be used and for less than a day. In a case study of a Greek household, we train and compare four common algorithms based on the data gathered through this protocol and perform two tests: an out-of-sample test in the artificially multiplexed signal, and an external test to predict the household’s appliances’ operation based on the time series of a real total consumption signal. We find accurate monitoring of the operation and the power consumption level of high-power appliances, while in low-power appliances the operation is still found to be followed accurately but is also associated with some incorrect triggers. These insights attest to the efficacy of the protocol and its ability to produce meaningful tips for changing energy behaviours even under constraints, while in said conditions, we also find that long short-term memory neural networks consistently outperform all other algorithms, with decision trees closely following.

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.

The transformation of energy and power systems is impacted by several factors. The introduction of smart grid technologies is changing the structure and direction in which the distribution and transmission electricity systems are operating. The development of power systems technologies also pushed the operational limits of the elements to a higher level. Since the power system backbone is well established and new investments are rare, the Transmission system operators (TSOs) need to maintain the system stable and provide a secure supply is facing new challenges. The way the system operates and is controlled followed the development of the IT and Power sectors and the approach to system maintenance is undergoing major changes as well. With the implementation of a new IT solution, new detailed monitoring is enabled and the TSOs have a big arsenal of new data, which should be utilized in planning and maintenance planning activities as well. The well-established time-based approach to system maintenance needs revision and upgrades. This paper will describe the updates in the maintenance planning process of the Slovenian transmission system operator ELES as the pilot location within the Horizon 2020 research project Big Data for Next Generation Energy (BD4NRG), which deals with the implementation and connection of big data systems in the power systems sector. The innovative approach aims to provide a faster alternative for the maintenance outage planning of the transmission system elements, with the implementation of newly available data that was not used to the full extent in the established planning process. The main benefit of the designed tool is the automated and repetitive process of maintenance outage planning, with consideration of the locational factors, monitored data, and external actor requests.

Increasingly, homeowners opt for photovoltaic (PV) systems and/or battery storage to minimize their energy bills and maximize renewable energy usage. This has spurred the development of advanced control algorithms that maximally achieve those goals. However, a common challenge faced while developing such controllers is the unavailability of accurate forecasts of household power consumption, especially for shorter time resolutions (15 minutes) and in a data-efficient manner. In this paper, we analyze how transfer learning can help by exploiting data from multiple households to improve a single house’s load forecasting. Specifically, we train an advanced forecasting model (a temporal fusion transformer) using data from multiple different households, and then finetune this global model on a new household with limited data (i.e., only a few days). The obtained models are used for forecasting power consumption of the household for the next 24 hours (day-ahead) at a time resolution of 15 minutes, with the intention of using these forecasts in advanced controllers such as Model Predictive Control. We show the benefit of this transfer learning setup versus solely using the individual new household’s data, both in terms of (i) forecasting accuracy (∼15% MAE reduction) and (ii) control performance (∼2% energy cost reduction), using real-world household data.

High penetration of renewable energy sources brings both opportunities and challenges for Smart Grid operation. Due to their high contribution to energy consumption, aggregated load flexibility of small residential and service sector consumers has a potential to address the intermittency challenge of distributed generation. Predicting aggregated load flexibility of this consumer sector involves access to sensitive smart meter data, raising data collection and sharing concerns. Federated Learning, a decentralized machine learning technique that uses data distributed on user devices to construct an aggregated, global model, offers potential solutions to tackling this challenge. This paper explores the potential of using Federated Learning for flexibility prediction in Smart Grids through an analysis of its opportunities and implications for different stakeholders involved, as well as the challenges faced. The analysis shows that Federated Learning is a promising approach for building privacy-preserving energy portfolios of aggregated demand data.

Today’s energy systems and smart grids produce large volumes of data on a daily basis. These data are exploited by their owners through specific applications. However, the utility of these data does not approach its full potential. This is because, on the one hand applications are intended to perform a specific task, while on the other hand, often times the owners of these data lack technical and analytical skills to get, process, and analyze the data. Moreover data silos, caused by technological barriers and/or organizational culture prevent organisations that have significant amounts of data and know-how from sharing them with other interested parties even if it would be beneficial for the common good. Last but not least, the combination of these data with sources from other domains could significantly improve the accuracy of several services. However, for the time being, this is not feasible due to standardization and interoperability issues. This publication proposes a technical solution that enables visual analytics, AI models development, and efficient querying of heterogeneous (Big) Data sources for the Electric Power and Energy Systems (EPES) sector. This solution has been designed under the context of the EU funded H2020 research project BD4NRG that aims at addressing the emerging challenges in big data management for energy and enabling Business to Business data sharing to unlock new market opportunities.

Energy efficiency (EE) projects are often fragmented, of high transaction costs, and fall below the minimum value that many private financial institutions are willing to consider. The finance community is lacking a tested, evidence-based platform, providing decision makers with support regarding the impacts of various investment criteria, risk aware assessment, and performance applied on a pool of EE investments. The capability offered by emerging near big data analytics to integrate cross-domain financial and energy consumption is key to building the necessary market confidence in EE projects and making them an attractive investment asset class. The availability of comparable, anonymized historical data pooled from major market segments, structured along major project characteristics, can encourage greater EE investment flow. The aim is to present data-driven applications based on machine learning methods that can attract and mobilize private funding on such projects, providing investors/financiers (e.g., commercial/green investment banks, institutional/insurance funds, etc.) and project developers (public/local authorities, energy providers, ESCOs, construction companies, etc.) with data and tools to identify sustainable investment pathways and decrease the EE investment risk. Extensive data processing is applied to elaborate and categorize financing instruments and risk mitigation strategies, and to identify best practices on private financing as a basis for benchmarking.

The accurate forecasting of photovoltaic (PV) power generation is of great significance in renewable energy systems, as it enables optimal energy management and grid stability. Despite the importance of this issue, substantial limitations still exist in the majority of existing research initiatives, which employ shallow machine learning algorithms. Recently, some studies have proposed employing convolutional and long short-term memory neural networks (LSTMs) in conjunction with transfer learning techniques; however, these approaches require that the production of PV systems is known during training. To overcome these limitations, we present the first study in the task of PV power forecasting utilizing unsupervised domain adaptation methods. Specifically, we employ two unsupervised methods, namely Domain Adversarial Neural Network and Margin Disparity Discrepancy. Both approaches use a source and a target domain during training, where the target labels of the target domain are unknown during training. We use production and weather data from seven PV systems with nominal capacities ranging from 23.52 kW to 271.53 kW, located in different areas. The findings demonstrate that our proposed architectures improve root mean squared error (RMSE), normalized RMSE, and 𝑅2 scores over the smart persistence model across all the PV systems used for testing. Furthermore, our approaches improve the performance of the smart persistence model, with a forecast skill index reaching up to 45.35%. Our extensive experiments demonstrate that our introduced approaches offer valuable advantages over state-of-the-art ones, as the target variable of the target domain is unknown during training. We also demonstrate the robustness of our approaches by conducting a series of ablation experiments.

Accurate estimation of energy savings is crucial for the effective implementation of energy conservation measures (ECMs). Simultaneously, the integration of Artificial Intelligence (AI) has revolutionized software engineering by imbuing software with intelligent capabilities and autonomy. In this paper, we propose an ensemble model for precisely estimating baseline energy consumption within the realm of AI-empowered autonomous software. The ensemble model combines predictions from three tree-based Machine Learning models, namely Random Forest, XGBoost, and LightGBM. Notably, our model emphasizes the provision of explainability, granting transparency and insights into the key factors influencing baseline energy consumption. To validate its effectiveness, we conduct experimental evaluations on a diverse cluster of real-world buildings in Latvia. The results demonstrate the superiority of our proposed ensemble model over individual models and even a deep learning network tailored for energy consumption estimation. These findings underscore the efficacy of AI-empowered models in the energy sector, offering a robust and interpretable solution for estimating energy savings.

This paper presents a novel development methodology for artificial intelligence (AI) analytics in energy management that focuses on tailored explainability to overcome the “black box” issue associated with AI analytics. Our approach addresses the fact that any given analytic service is
to be used by different stakeholders, with different backgrounds, preferences, abilities, skills, and goals. Our methodology is aligned with the explainable artificial intelligence (XAI) paradigm and aims to enhance the interpretability of AI-empowered decision support systems (DSSs). Specifically,
a clustering-based approach is adopted to customize the depth of explainability based on the specific needs of different user groups. This approach improves the accuracy and effectiveness of energy management analytics while promoting transparency and trust in the decision-making process. The
methodology is structured around an iterative development lifecycle for an intelligent decision support system and includes several steps, such as stakeholder identification, an empirical study on usability and explainability, user clustering analysis, and the implementation of an XAI framework.
The XAI framework comprises XAI clusters and local and global XAI, which facilitate higher adoption rates of the AI system and ensure responsible and safe deployment. The methodology is tested on a stacked neural network for an analytics service, which estimates energy savings from renovations,
and aims to increase adoption rates and benefit the circular economy.

The monitoring and control of Critical Energy Infrastructure (CEI) is nowadays entrusted to Smart Grids (SGs). SGs rely on massive data and services to provide “awareness” about the status of the system. To do that distributed computing schemes have been applied based on decentralized communications, data collection, extractions, loading and analysis. These schemas are totally aligned with the Edge Computing (EC) paradigm. EC is an emerging paradigm that provides capabilities for processing and analyzing data away from the cloud, at the edge of the network closer to the source of the data. It offers multiple benefits including improved application performance, network latency reduction, and data locality. These characteristics reinforce EC is expected to have great impact on SG. However, a crucial aspect in implementing EC is company’s foundational technology to really progress in cyber, digital, and cloud moves for SGs. The authors strongly believe that the foundation for the successful implementation of cloud/edge-based solutions strictly depends on employing new core architectures based on modern advanced cloud-native solutions, i.e., patterns, tools, techniques, and technologies derived from cloud-based design. As a result of this statement an Edge Platform-as-a-Service (PaaS) has been designed, developed, and deployed and used as the foundation of a flexible data platform at the Edge made up of fast-deployable, open source, and free-to-use PaaS services

Energy management is crucial for various activities in the energy sector, such as effective exploitation of energy resources, reliability in supply, energy conservation, and integrated energy systems. In this context, several machine learning and deep learning models have been developed during the last decades focusing on energy demand and renewable energy source (RES) production forecasting. However, most forecasting models are trained using batch learning, ingesting all data to build a model in a static fashion. The main drawback of models trained offline is that they tend to mis-calibrate after launch. In this study, we propose a novel, integrated online (or incremental) learning framework that recognizes the dynamic nature of learning environments in energy-related time-series forecasting problems. The proposed paradigm is applied to the problem of energy forecasting, resulting in the construction of models that dynamically adapt to new patterns of streaming data. The evaluation process is realized using a real use case consisting of an energy demand and a RES production forecasting problem. Experimental results indicate that online learning models outperform offline learning models by 8.6% in the case of energy demand and by 11.9% in the case of RES forecasting in terms of mean absolute error (MAE), highlighting the benefits of incremental learning.

The rising digitisation of the energy system and related services is unveiling an enormous opportunity for energy stakeholders to leverage on Big Data & AI technologies for improved decision making and coping with challenges emerging from an increasingly complex and interconnected energy system. Initiatives in the field of Big Data Reference Architectures, like IDSA, GAIA-X or FIWARE provide generic frameworks to share, manage and process Big Data. Through alignment among them and integration of missing aspects, an interoperable and secure framework for the energy comes into view. The Reference Architecture presented in this paper moves towards this goal and will be instantiated in a set of concrete use cases within the European Energy Sector. Structurally inspired by SGAM and the BRIDGE Reference Architecture, it puts concrete analytics processes and data source components into context, taking important issues of Data Governance, Security, and Value Creation into account.

Energy Efficiency projects are often fragmented, of high transaction costs, and fall below the minimum value that many private financial institutions are willing to consider. The availability of comparable, anonymised historical data pooled from major market segments, structured along major project characteristics, can encourage greater investment flow in energy management and efficiency. The aim of this paper is to identify investment financing patterns in a pool of provided projects in Latvia and discover possible Grand Financing Plans (GFP) for future use. These GFPs could improve the procedure of decision making in energy sector in terms of the percentage of grand financing per project. The improvement of the process of grand financing can attract and mobilise private funding on such projects, providing investors/financiers (e.g., commercial/green investment banks, institutional/insurance funds, etc.) and project developers (public/local authorities, energy providers, ESCOs, construction companies, etc.) with data and tools to identify sustainable investment pathways and decrease the investment risk

Interoperability within a data space requires participants to be able to understand each other. But how do you get data space participants to use a common language? According to the IDS Reference Architecture Model (IDS-RAM)1 , the main responsibility for this common language lies with an intermediary role called a vocabulary provider. This party manages and offers vocabularies (ontologies, reference data models, schemata, etc.) that can be used to annotate and describe datasets and data services. The vocabularies can be stored in a vocabulary hub: a service that stores the vocabularies and enables collaborative governance of the vocabularies. The IDS-RAM specifies little about how vocabularies, vocabulary providers and vocabulary hubs enable semantic interoperability. The hypothesis that we address in this position paper is that a vocabulary hub should go a step further than publishing and managing vocabularies, and include features that improve ease of vocabulary use. We propose a wizard-like approach for data space connector configuration, where data consumers and data providers are guided through a sequence of steps to generate the specifications of their data space connectors, based on the shared vocabularies in the vocabulary hub. We illustrate this with our own implementation of a vocabulary hub, called Semantic Treehouse.

Mainstreaming energy efficiency financing has been considered a key priority during the last decade among several stakeholders. The capability offered by Multicriteria Decision Analysis to integrate cross-domain financial and energy consumption data, combined with statistical analysis techniques and data abundance, contributes to building the necessary market confidence in energy efficiency projects and make them an attractive investment asset class. In this context, the aim of this paper is to propose a solid methodological framework in order to support the financing procedure of energy efficiency investments, and to identify improved grant financing plans, considering a series of factors which are of vital importance for the sustainability of such actions and the limitation of investment risk.

A decision support tool, developed in Python, is presented which implements the suggested methodology, improving the decision making for the investor in terms of the percentage of grant financing per project. The developed methodology has been applied on a reliable dataset of energy efficiency projects from several cities in Latvia, where the actual performance of the investments is exploited. The application of the methodology has resulted in a financing plan which achieves about the same energy savings, while bringing 15% reduction of the energy efficiency investments’ cost.

Accurately forecasting solar plants production is critical for balancing supply and demand and for scheduling distribution networks operation in the context of inclusive smart cities and energy communities. However, the problem becomes more demanding, when there is insufficient amount of data to adequately train forecasting models, due to plants being recently installed or because of lack of smart-meters. Transfer learning (TL) offers the capability of transferring knowledge from the source domain to different target domains to resolve related problems. This study uses the stacked Long Short-Term Memory (LSTM) model with three TL strategies to provide accurate solar plant production forecasts. TL is exploited both for weight initialization of the LSTM model and for feature extraction, using different freezing approaches. The presented TL strategies are compared to the conventional non-TL model, as well as to the smart persistence model, at forecasting the hourly production of 6 solar plants.

Despite the large number of technology-intensive organisations, their corporate know-how and underlying workforce skill are not mature enough for a successful rollout of Artificial Intelligence (AI) services in the near-term. However, things have started to change, owing to the increased adoption of data democratisation processes, and the capability offered by emerging technologies for data sharing while respecting privacy, protection, and security, as well as appropriate learning-based modelling capabilities for non-expert end-users. This is particularly evident in the energy sector. In this context, the aim of this paper is to analyse AI and data democratisation, in order to explore the strengths and challenges in terms of data access problems and data sharing, algorithmic bias, AI transparency, privacy and other regulatory constraints for AI-based decisions, as well as novel applications in different domains, giving particular emphasis on the energy sector. A data democratisation framework for intelligent energy management is presented. In doing so, it highlights the need for the democratisation of data and analytics in the energy sector, toward making data available for the right people at the right time, allowing them to make the right decisions, and eventually facilitating the adoption of decentralised, decarbonised, and democratised energy business models.

This study introduces an energy management method that smooths electricity consumption and shaves peaks by scheduling the operating hours of water pumping stations in a smart fashion. Machine learning models are first used to accurately forecast the electricity consumed and produced by renewable energy sources on an hourly level. Then, the forecasts are exploited by an algorithm that optimally allocates the operating hours of the pumps with the objective to minimize predicted peaks. Constraints related with the operation of the pumps are also considered. The performance of the proposed method is evaluated considering the case of a Greek remote island, Tilos. The island involves an energy management system that facilitates the monitoring and control of local water pumping stations that support residential water supply and irrigation. Results indicate that smart scheduling of water pumps in a small-scale island environment can reduce the daily and weekly deviation of electricity consumption by more than 15% at no monetary cost. It is also concluded that the potential gains of the proposed approach are strongly connected with the amount of load that can be shifted each day, the accuracy of the forecasts used, and the amount of electricity produced by renewable energy sources.

Energy efficiency is critical for meeting global energy and climate targets, requiring however significant investments. Due to the lack of mature decision-support systems and the utilization of traditional investment mechanisms that focus on the economical aspects of the energy efficiency projects and neglect their environmental impact, such projects can experience difficulties in being funded. In the interim, the impact of the digitization era is more apparent than ever, as algorithms and data availability and quality have significantly improved. This study aspires to bridge the gap in energy efficiency financing with the development of a data-driven methodology that labels energy efficiency investments based on their expected utility in terms of renovation cost and energy savings. Various machine learning classification methods are deployed and combined through a meta-learning model with the objective to improve overall classification performance and determine the funding that each investment should receive according to its particular characteristics. The proposed methodology is evaluated using a set of 312 projects that have been completed in Latvia. Our results indicate that the meta-learner outperforms all baseline classifiers, effectively identifying projects of high and medium potential and successfully distinguishing low from high potential ones.