Adaptive model predictive control for energy-efficient smart homes using a dynamic Kalman filter-bank

In order to be economically attractive the implementation of a smart home product should be simple and possibly done by craftsman. However, the model parametrization requires expert knowledge and the first commissioning has to perform well from the start. In this paper an adaptive model predictive control (aMPC) scheme for energy-efficient smart homes using a dynamic Kalman filter-bank is presented, which does not require on-line input from experts. Predefined smart home models are stored and simulated in parallel in the Kalman filter-bank. This allows for an efficient representation of the most important smart home dynamics and guarantees that the underlying aMPC model gives sufficient performance. The proposed scheme allows fast commissioning in smart homes and adds flexibility by extending the model set with additional predefined models. Switching models in the aMPC can represent time-varying behavior and ensure the best performance and energy-efficiency. The switching utilizes a hysteresis to avoid high-frequency switching and ensures a robust control scheme. Stability is analyzed, a simulation comparison to a Fuzzy MPC is performed, and the results show that the proposed method is both versatile and effective.