Traditional building automation controllers are having low performance in dealing with non-linear phenomena. In recent years, model predictive control (MPC) has become a notable control algorithm for building automation system capable of handling non-linear processes. Performance of model-based controllers, such as MPC, is depending on reasonably accurate process models. For a building using baseboard radiator heater, a non-linear model is a more reliable representation of heat distribution system. Therefore, this study aims to present a non-linear gray-box model for a residential building connected to the local district heating network that is equipped with radiator heat emitters. The model is supposed to forecast the indoor air temperature as well as the radiator secondary return temperature. The model is validated using measurements collected from a building in Västerås, Sweden. In addition to a better accuracy, another motivation behind using a non-linear heating circuit model is to enhance its generalization performance. With the added benefits of accuracy and generalization, this model is expected to extend practical MPC implementation for such buildings.
Keywords Metal recovery, printed circuit board, machine learning, environmental sustainability, circular economy