An unscented Kalman filter approach for the plant-model mismatch reduction in HVAC system model based control

The building sector is rising rapidly which is not only the consequence of the world's increasing population but also due to higher requirements with regards to comfort. Buildings along with HVAC systems to run these buildings account for a large amount of the world's global energy consumption. Hence operating buildings as well as HVAC systems in an energy-efficient way plays the key role in reducing energy consumption and therefore making the contribution towards mitigation of the climate change. Studies about advanced control techniques for buildings and systems to run those have shown promising results with regards to lower energy consumption and the optimal use of renewable energy sources. In contrast to the conventional approaches (e.g.: PID, two-position controller, etc.) for building and HVAC control, model based concepts take advantage of the direct knowledge of the system behaviour using an image of the system. Model based control concepts facilitate the systematic approach towards the energy-efficient operation of building and HVAC systems. However, the quality of the model based controller strictly depends upon the quality of the model employed to describe the dynamic behaviour of the system being subject to control. The model-plant mismatch can negatively impact the energy-efficient control operation. This paper presents the design of an unscented Kalman filter (UKF) approach for the purpose of state and parameter estimation for solar thermal HVAC system control. This UKF concept incorporates the possibility of model update using measurements and therefore reduces model-plant mismatch which in turn improves the control quality and can be regarded an asset with respect to the energy-efficient control operation.