Digital Twins of Smart Energy Systems: Models and Optimization Algorithms

In 2022, urban areas in Europe accounted for 80% of energy consumption [1], and, as the global population grows and living standards improve, the worldwide energy demand will not diminish in the coming years. Renewable Energy Sources (RES) are part of the solution toward decarbonisation, however, due to the irregular availability of renewable energy from solar and wind, achieving a balanced and resilient energy system is a challenge. Therefore, there is a critical need for innovative methods and models to create an economically sustainable energy sector which includes Smart Energy Systems (SESs) where various energy sources and consumers are integrated. For these purposes, The European Union actively promotes the digitalisation of these systems via various initiatives [2], and Digital Twins (DTs) represent the opportunity to combine scientific progress with real-world application. Using these synchronised replicas, operators may acquire real-time simulation results by monitoring the system. This would allow to determine the most optimal actions for improving system operation and safety, as well as reducing costs and emissions. The focus of this PhD research project is the development of DT-oriented algorithms to model and optimize different energy systems. The current state-of-the-art (SoA) of DTs in the energy field is presented, emphasizing the applications of this technology for various objectives and describing how to assess the maturity of a DT or its differences from other models. All the methodologies used are then explained, including data-driven and physics-based modelling, optimization algorithms, data processing, and visualization. Specifically, studies on District Heating Networks (DHNs), microgrids, Solid Oxide Fuel Cells (SOFCs) and buildings were conducted, detailing each methodology and the results in distinct chapters. These studies address various gaps in the existing literature, demonstrating the benefits and potential of DT technology. For instance, simulations on fourth and fifth-generation DHNs revealed significant energy and emissions reductions compared to third-generation networks or autonomous consumers. The twin of Rio Vena DHN, a third-generation network, confirmed that integrated optimization of boilers could reduce the environmental impact even when operating with natural gas. In the context of microgrid optimization, data from the Emotion facilities case study enabled the integration of a DT capable of maximizing Photovoltaic (PV) self-consumption by optimizing Electric Vehicles (EVs) charging and providing real-time suggestions. Therefore, by analysing over six months of data of a SOFC provided by the CNR, an "ageing" DT was developed to maintain an accurate Machine Learning (ML) model for optimizing the operation of four different SOFCs in a larger context. Finally, different studies on SolarLab laboratory were conducted to create a building DT capable of enhancing indoor comfort.