Design and Optimization of a Hybrid Energy System: Towards the Full Decarbonization of a Small Mediterranean Island

The hybrid energy system (HESs) paradigm is a promising way to enhance renewable energy source (RES) penetration, particularly in remote Mediterranean regions which struggle in achieving energy independence. This dissertation focuses on addressing the challenges of designing and managing HESs and explores innovative computational and design strategies to optimize both the operational and structural aspects of renewable energy systems. The research, conducted in collaboration with Eunice Energy Group under the PON program, uses the energy system installed on the Greek island of Tilos as a case study. The Tilos HES comprises a wind turbine, a photovoltaic (PV) field, and a high-temperature battery, and can operate in two modes i.e., as a hybrid power station (HPS) and a self-sustained microgrid. The system currently faces operational challenges, including energy curtailment due to strict constraints coming from the grid operator and the strong seasonal mismatch between renewable production and demand. This research aims at solving both problems, first by enhancing operational efficiency through robust energy dispatch algorithms and then by improving system design for complete energy autonomy. Operational optimization was first addressed by employing data-driven methods to enhance RES production forecasts, improving their accuracy and reducing bias errors. Machine learning (ML) techniques proved to be particularly effective for solar forecasts but faced limitations when applied to wind production. To mitigate residual forecast uncertainties, robust optimization (RO) based on mixed-integer linear programming (MILP) was implemented. Following the robust approach, the optimization framework was structured to simultaneously optimize the HPS behavior in several production scenarios. For achieving energy autonomy, the research examined the optimal expansion of renewable generation and hybrid storage systems. Given the seasonal nature of energy demand, hybrid storage combining lithium-ion batteries with hydrogen storage technology emerged as the most cost-effective solution. An in-depth market research was conducted to obtain average component prices considering the current situation and projections to 2030 and 2050. An optimization framework based on stochastic optimization was developed to find the solution that reaches the full-autonomy objective with the lowest possible cost of energy. The optimization in turn includes a simulation framework able to estimate the long-term performance of a specific combination of generation and storage, considering the effect of degradation of electrochemical cells. This research thus provides a comprehensive toolbox for addressing the operational and design challenges of HES, offering insights into the transition to sustainable, autonomous energy systems for island communities. Key findings underline the potential of data-driven and optimization techniques to enhance RES integration. ML models, when optimized for training window size and hyperparameters, demonstrated substantial improvements in forecast accuracy. The robust MILP approach outperformed traditional dispatch strategies, significantly enhancing system performance in grid-constrained environments. Results have also shown significant improvements in system reliability and reduced economic penalties associated with forecast errors. Furthermore, hybrid storage systems incorporating P2G2P and batteries offered the lowest cost of energy, emphasizing the importance of power-energy decoupling for achieving full self-sufficiency. Finally, scenarios incorporating future price reductions for PV and storage technologies revealed that maximizing cheap RES generation is more convenient than overinvesting in storage.