Automated optimization based synthesis of distributed energy supply systems

摘要

The conceptual synthesis of distributed energy supply systems (DESS) is an inherently challenging problem that is characterized by time-dependent constraints (e.g., energy demands, ambient temperatures curves, etc.), economy of scale of equipment investment, limited capacities of standardized equipment, and part-load performance characteristics of the considered energy conversion units. Moreover, for optimal DESS synthesis, multiple redundant units are generally to be expected. Optimization-based synthesis methods offer great potentials for the synthesis of cost-effective, energy-efficient, and sustainable systems. However, a lack of adequate, user-friendly methods has so far hindered routine application of optimization in engineering practice. In research, most commonly, superstructure-based synthesis is performed for optimal systems synthesis. In this approach, a user-defined superstructure is analyzed using mathematical programming techniques to identify the optimal solution among the alternatives encoded in the superstructure. Current optimization software facilitates the use of superstructure-based synthesis, e.g., by enabling easy problem definition through graphical superstructure modeling. However, the a priori definition of the superstructure remains a serious obstacle for the use of superstructure-based synthesis in industrial practice: On the one hand, the manual superstructure definition bears the risk to exclude the optimum from consideration; on the other hand, the use of excessively large superstructures causes prohibitively large computational effort. To circumvent these drawbacks, superstructure-generation methods and superstructure-free synthesis methods have been proposed. Superstructure-generation methods automatically define a superstructure for a given synthesis problem. Superstructure-free methods avoid the use of a superstructure by enabling simultaneous alternatives generation and optimization. Available approaches involve several drawbacks that impede their use for the optimal synthesis of distributed energy supply systems: Superstructure-generation methods neglect major DESS characteristics; superstructure-free methods require the manual definition of many technology-specific replacement rules, which is equally difficult as the definition of an appropriate superstructure. In this thesis, two novel synthesis methodologies are proposed to facilitate the use of optimization for efficient and reliable DESS synthesis, thus making optimization accessible for practitioners: The automated superstructure-based and the superstructure-free synthesis methodology. The proposed methodologies avoid both the a priori definition of a superstructure and the manual definition of many technology-specific replacement rules while accounting for the major characteristics inherent to DESS synthesis problems. The superstructure-based framework relies on an algorithm for automated superstructure-generation. The method employs a successive superstructure expansion and optimization strategy that continuously increases the number of units included in the superstructure until the optimal solution is identified. The superstructure-free approach combines evolutionary optimization and deterministic optimization for simultaneous alternatives generation and optimization. A knowledge-integrated mutation operator is proposed that relies on a hierarchically-structured graph, the so-called energy conversion hierarchy (ECH). The ECH efficiently defines all reasonable replacement rules, thus avoiding their manual definition. The mutation operator performs structural replacements for the evolutionary generation of solution alternatives. Both synthesis methodologies use a generic component-based modeling framework, thus making the methodologies independent of the employed mathematical programming formulation. In this thesis, a robust MILP formulation is used that allows to simultaneously optimize the structure, sizing, and operation of distributed energy supply systems accounting for time-varying load profiles, continuous equipment sizing, economy of scale of equipment investment, and part-load equipment performance. In this thesis, it is shown that both synthesis methodologies proposed in this thesis enable practitioners to perform optimization-based synthesis of distributed energy supply systems. It should be pointed out that the use of the proposed synthesis methodologies only requires energy-related expert knowledge that is usually prevailing among engineers active in the field of energy systems synthesis. In particular, no expert knowledge on mathematical programming is required. Finally, this thesis provides the foundation for future research as discussed in the next section. Last but not least, based on the experience gained during the work on this thesis, the author comments on the necessity of optimization for the conceptual DESS synthesis.