Multiple energy carrier systems stem from the need to evolve traditional electricity, gas and other energy systems to 10 more efficient, integrated energy systems. An approach is presented, for controlling multiple energy carriers, 11 including electricity (AC or DC), heat, natural gas and hydrogen, with the objective to minimise the overall cost 12 and/or emissions, while adhering to technical and commercial constraints, such as network limits and market 13 contracts. The technique of multi-agent systems (MAS) was used. The benefits of this approach are discussed and 14 include a reduction of more than 50% in the balancing costs of a potential deviation. An implementation of this 15 methodology is also presented. In order to validate the operation of the developed system, a number of experiments 16 were performed using both software and hardware. The results validated the efficient operation of the developed 17 system, proving its ability to optimise the operation of multiple energy carrier inputs within the context of an energy 18 hub, using a hierarchical multi-agent system control structure. It is necessary to evolve traditional electricity, gas and other energy systems to more flexible, integrated energy 34 systems [11], referred to as multiple energy carrier, or multi-carrier systems. The points of interaction between 35 different energy carriers have been described as "energy hubs" [5], [12], which present an integrated approach for 36 optimizing systems with multiple energy carriers, such as electricity, hydrogen, or natural gas networks [13]. Devices 37 are incorporated in an energy hub with the purpose of converting from one carrier to another, e.g. a CHP unit 38 converting natural gas to electricity and heat. Storage elements such as batteries or thermal storage may also be 39 considered. The energy carrier inputs to the energy hub are optimised and controlled in order to supply a given set of 40 energy carrier loads / outputs, thus achieving whole-system optimization [12], [14]. In (1), the backward coupling 41 matrix (Dnm) which links the inputs (Pm) with the outputs (Ln) is shown, as this is a formality that is used in Section 2.2. 42The elements of the Dnm matrix are constructed using the conversion efficiencies of individual devices in the energy 43 hub [12]. Matrix dimensions are × 1, × 1 and × , for Pm, Ln and Dnm respectively. 44
