Development of Al-based models for the management of energy communities

The transition towards a new sustainable development paradigm and the promotion of carbon neutrality by 2050 envisages the application of several energy-saving measures also in the field of buildings, which account for more than a third of global consumption. The coupling of construction and renewable energy sources is one of the key points embedded in a suitably developed policy framework. Unfortunately, most of the renewable energy sources employed in the residential field are powered by solar or wind energy, characterized by a non-programmable nature. This could lead to mismatches between energy production and consumption, undermining the stability of power grids. In order to overcome these problems, a new model is increasingly being developed based on Energy Communities. These take the form of a group of electric and thermal utilities which share the same power generation systems and are able to consume a good deal of self-produced energy. Coordinated management of buildings belonging to an Energy Community is one of the key requirements for making an interaction with the power grid effective and impactful. This can be achieved through the deployment, at different scales, of Internet of Thing and Cloud Computing-based technologies, which enable long-term monitoring data. The subsequent development of a data-driven virtual simulation environment allows us to understand the role played by different integrated energy technologies at different scales, from individual buildings to entire aggregates. The results achievable through that model enables us to identify the ideal combination to meet multiple objectives including environmental, energy and economic ones. In this context, a useful tool for comparing different scenarios is the development of Key Performance Indicators, which may or may not have a temporal significance. This work proposes the development of Artificial Neural Networks applied on a district scale in order to simulate the thermal dynamics of buildings belonging to different Energy Communities. The performances of these data-driven supervised learning models are then compared whit those of a grey-box models developed by other researchers on the same database. Next, this research set out to use a virtual environment developed on a district scale to simulate real-world contexts. In more detail three different scenarios were investigated, concerning different Energy Communities, which may have different generation systems and consumption units. The results emerging from these analyses show significant advantages when energy production takes place on site through a centralized micro-cogeneration plant, sized to meet the entire Energy Community’s requirements. The efforts regarding these generation systems have led to the promotion of appropriate development drivers based on economic enhancements and consequent reductions in the cost of purchasing electricity. Moreover, taking a multi-energy approach, such as this, reveals better management of energy flows. Due the increasingly ambitious decarbonization goals, electric vehicle offer a set of important advantages in terms of sustainable mobility. However, it is important not to neglect electricity-related emissions for charging such power consumption units. Meeting their electricity demand through a micro-cogeneration system results in cheaper costs than full purchase from the grid, but also in a slight increase in greenhouse gas emissions to the atmosphere and an additional cost for fuel procurement.