Optimal sizing of Renewable Energy Communities under uncertainty

In recent years, the global landscape of energy production and consumption has witnessed a shift towards more sustainable and community-driven models. Renewable energy communities (RECs) have emerged as dynamic and participatory entities where individuals, businesses, and local organizations unite to collectively generate, manage, and share energy resources. This transformation turns them from passive consumers into active participants, or "prosumers", contributing to the transition towards clean energy and fostering decarbonization. Realizing a REC demands substantial investments in renewable energy sources (RES), and the economic feasibility of such projects is pivotal for their successful establishment and long-term sustainability. Recognizing this, institutions are actively pushing for their diffusion by adopting incentive policies. In the Italian context, RECs are promoted through the valorization of the shared energy, a performance indicator for diffused self-consumption configurations, strongly related to the synchronism between generation and consumption. This synchronism, in a pre-feasibility study, is often difficult to evaluate due to the strong variability in RES production and the users’ consumption patterns, leading to an unreliable assessment of the economic feasibility of the project. Within this paradigm, this thesis aims to develop a stochastic optimization model to determine the optimal design of the energy production plant from renewable sources, specifically photovoltaic (PV) technology, to maximize the Net Present Value (NPV) of the project, and to evaluate the possibility of the installation of an aggregate storage system within the REC. The problem is formulated as a two-stage stochastic MILP, where first-stage decision variables account for the sizing of PV and storage systems, while second-stage variables optimize energy dispatch operations. Uncertainty is introduced into the model following a probabilistic approach based on Monte Carlo simulations: different scenarios are generated, starting from historical data, in order to capture the variability of the uncertain parameters and to consistently represent their possible combinations. The uncertainty modeling mainly involves the photovoltaic generation, the energy load, and energy prices. The hourly PV production is modeled with a Gaussian mixture distribution, and daily PV profiles are reconstructed by sampling sequentially from the fitted random variables. A similar procedure is applied to generate load profiles and energy price scenarios. The impact of each stochastic parameter on the output is analyzed, and stopping criteria for Monte Carlo simulations are evaluated to ensure a tradeoff between accuracy and computational time. Economic and energetic Key Performance Indicators (KPIs), including payback time, self-sufficiency, and self-consumption, are then computed to assess the goodness and feasibility of the solution. Finally, the results of the stochastic problem are tested and compared against the deterministic solution, obtained by solving the problem considering the average values of the uncertain parameters. The solution to the stochastic problem can lead, depending on the configuration of the CER, to an increase of up to 15% in NPV, resulting in a drastic reduction of the payback time of the investment.