Including fairness and uncertainties in the design and operation optimization of local multi-energy systems

In the context of “local-to-regional” Multi-Energy Systems (MES), this Thesis aims to i) evaluate the weight of uncertainties of input data and boundary conditions in the design optimization of local MES and ii) analyse the aspects influencing the optimal aggregation of end users in Energy Communities (ECs). According to the second objective of this Thesis, the optimal aggregation of end users is studied in two different reference cases, neglecting uncertainties. In the first reference case, Mixed Integer Linear Programming (MILP) models are developed to optimize the daily operation of the local power systems of the Citizen Energy Community (CEC) and the Renewable Energy Community (REC). Results shows that the complementarity between the energy demand and generation profiles of the EC members leads to relevant cost savings (e.g., daily cost savings are 15-20 % higher in ECs with a higher degree of complementarity between members). A novel cost allocation mechanism is also proposed to encourage end users using free-of-charge and non-dispatchable RES to join the EC. The second reference case goes beyond the optimal aggregation of users and examines the “fairness” of the distribution of the total economic benefit among the members of an EC. A MILP model is developed to optimize the daily operation of the local power system of a REC, and the “Shapley value” mechanism is applied to fairly allocate the optimal total profit. This mechanism distributes higher economic benefits to the prosumers than to the consumers, as they contribute significantly to increasing the total economic benefit of the system. The third case study, in line with the first objective of this Thesis, focuses on the weight of uncertainties associated with RES and electricity demands in the design optimization of a local MES serving a REC. A novel framework is developed to perform the design-operation optimization of the system in the absence of and under uncertainties. The framework includes MILP and Stochastic Programming (SP) optimization models without and with uncertainties, solved for each day and for a set of daily stochastic scenarios of the uncertain parameters in one year, respectively. Different sets of stochastic scenarios are obtained by applying a clustering technique for each year of a training dataset. The “best” set of stochastic scenarios is then found by searching for the minimum standard deviation of errors between the solutions of the MILP and SP models without and with uncertainties, respectively, over the years of the training dataset. The SP model with the “best” set of stochastic scenarios is solved for each year of a testing dataset, resulting in “stochastic forecast” solutions. Finally, the “stochastic forecasts” are compared with the “perfect forecast” solutions obtained by solving the MILP model in the absence of uncertainties for each year of the testing dataset. A key finding is that the “stochastic forecasts” predict the optimal life cycle cost of the system quite accurately, with an average error of 4 % compared to the “perfect forecasts”. Overall, this Thesis shows interesting results in the design-operation optimization of local energy systems, despite some necessary assumptions such as the geographical boundaries of the systems and the type of EC members. Several optimizations are performed to achieve general guidelines that lead to the optimal aggregation of end users in ECs, also ensuring a possibly fair distribution of the total economic benefit. The proposed framework for optimizing the design of local MES under uncertainties provides a reliable approach to assess the accuracy of the results under uncertainty with respect to those in the absence of uncertainty, allowing uncertainties to be weighted in the optimal design of the system.