Caractérisation des performances énergétiques des systèmes thermiques innovants pour le bâtiment au travers d'essais de courte durée en régime dynamique

摘要

Solar thermal systems combined with a backup system such as a boiler, a heat pump or incorporating an absorption chiller, can play an important role in reducing buildings energy consumption for heating, cooling and hot water production needs. In this sense, characterizing the energy performance of thermal systems is crucial.Currently available methods of system characterization are either based on several separate physical tests of system components to be evaluated, which do not take into account the real interactions between them, or on physical models that can be complex and difficult to identify especially because systems nowadays are compact and prefabricated in the factory. Due to the lack of a reliable method to estimate the performance of solar thermal systems before their integration into buildings, their market faces a lot of impediment to be developed.In this context, it becomes essential to develop a generic methodology that can be applied to different types of systems which overcomes the difficulties encountered by the current ones.The proposed evaluation approach in this manuscript is composed of four main steps: determining a test sequence, testing the system in a semi-virtual test bench according to predetermined sequence, data acquisition and identifying an artificial neural network (ANN) of the system and finally the model simulation in order to estimate the system consumption in the desired boundary condition. Using a completely "black box" model of the whole system using the ANN makes the methodology totally "non-intrusive". No prior knowledge about the systems internal parameters (yields, thermal conductivities, regulation etc.) is necessary to apply the proposed approach.The methodology validation was performed through several numerical experiments for seven systems coming from three different typologies. During the validation process, ANN estimates were compared with calculations of physical models in several different conditions (quality of building, climate and collector area). The developed approach was applied to five real systems as well. The application results allowed the confirmation of the methodology relevance.