Multi-objective optimization of the design and operation of an energy hub for the Empa campus

Abstract

This paper presents a comparison of two multi-objective optimization processes used to simultaneously select and size the components of an energy hub and to determine their optimal operati on according to net present value and carbon emissions. The first is a singlelevel optimiz ation process that uses a mixed-linear integer programming (MILP) model based on the energy hub concept in which time-varying demands and supply availabilities must be matched using conversion and storage options. The second is a bi-level optimization process composed of a multi-objective genetic algorithm (GA) as the upper level to optimize selection and sizing of components. A linear programme is nested within the GA as the lower level to optimize the operation of each proposed system. The study uses measured data from the Empa research campus in Dubendorf, Switzerland for the heating, cooling and electricity demands that must be met. Appropriate values for solar availability, energy prices and equipment costs were used. The optimization process is conducted for a whole year, allowing the consideration of seasonal storage. The energy hub includes electricity, gas, solar power, and medium temperature and high temperature thermal networks. The technologies considered include boilers, chillers, photovoltaic panels, combined heat and power plants, heat pumps and storage. Results pr esented give trade-off fronts of the competing objectives (carbon emissions and discounted costs) that reveal a set of optimal design solutions, including their optimized hourly operational schedules. The effectiveness of the two approaches is compared, including the convergence of the optimization, necessary computing time and the identification of solution characteristics. It is shown that the single-level optimization finds a better Pareto front in much shorter time than the bi-level approach for this problem instance.