Data-driven Buiding Climate Control Using Model Prediction and online Weather Forecast Data

This paper proposes a multi-unit building model, in which the parameters are obtained via an online identification process. The identification process is carried out on-the-fly so it can update the best model of the building units. A model predictive controller (MPC) is also employed that uses the prediction of the building model, as well as the weather forecast data and acts on the heating boiler in an optimal fashion. In addition, since the controller is designed for a multi-unit building, it is crucial to estimate the amount of the delay that takes the hot flow to reach the units. This paper presents a very simple method for the delay identification based on unscented kalman filter. For one scenario, the results of system identification as well as the design of the controller are shown in Monte Carlo simulation, and compared with the results of a traditional controller. The results show that the proposed method can save 13% in the energy consumption. The results are also obtained for 1000 simulations for different temperature scenarios, and came up with reaching up to a 25% save in the energy consumption.