Development and evaluation of a RC network grey-box model for an MPC system of an office building.

Energy used for building thermoregulation plays an important role in the whole energy consumption in developed countries. Model predictive control is an innovative tool utilized to optimize the energy consumption and consequently to reduce the production of greenhouse gases. Model Predictive Control systems exploit a model representing the controlled system in order to simulate it and use the output to optimize the control strategy. Cause the wide variety of building stock there is not a common methodology for modelling buildings and hence the modelling task it is easily the most time demanding in all the process of designing a Model Predictive Control system. The aim of this thesis is to develop and evaluate RC network models of an office building and investigate in some implementation of the models. The office building has been designed on EnergyPlus while the RC model has been developed on Modelica and then the parameters have been learned by ModestPy. Two RC models have been done, both with three thermal resistances and three heat capacities, but one with an implementation of a block to calculate the sun irradiance components according to the geometry of the controlled building. Each model has been used to predict the internal building temperature of the office building for all the heating period of a typical year in Torino. The accuracy of the predictions in relation of the internal temperature of the office building has been evaluated with the root mean square error. The comparison of the results of the two RC models do not state one model better than the other.