System identification for control of temperature and humidity in buildings

HVAC systems are widely used to provide a good indoor air quality in buildings. Buildings stand for a substantial part of the total energy consumption in developed countries, and with an increased focus on cost reductions and energy savings, it is necessary to use intelligent and energy-efficient controllers. Accurate models describing the dynamics of the building system is a good basis for intelligent control. In countries like Sweden there are large seasonal variations in the outdoor climate, and these variations interfere with the indoor condition and thus affects the control. In this thesis the seasonal variations are investigated, and the aim is to determine how these differences affect identified models for control of temperature and relative humidity in buildings. Two MISO (Multiple Input-Single Output) systems and one MIMO (Multiple Input-Multiple Output) system is used to describe the mean room temperature and relative humidity in a selected room in the Q-building at KTH, Stockholm. The models are identified following the black-box approach, and data from four different months during 2014, representing the winter, spring, summer and autumn season respectively, are collected and preprocessed. The validation of the identified models for the humidity and temperature, shows that it is possible to use identical orders and input delays for all seasons, with good results. Based on the results one would not recommend using models with the same model parameters throughout the year, since the conditions varies too much from season to season