Analysis of optimal energy management in smart homes using MPC

Advanced building management systems utilize future information, such as electricity spot prices, weather forecasts, and predicted electric loads and hot water consumption, to reduce the maximum electric power consumption and energy cost. A model predictive controller (MPC) is implemented for a household with one hour sample intervals, including hot water usage, charging of an electric vehicle, and domestic heating, but also an accumulator water tank to be used as an additional thermal energy storage. Both the maximum total power used in the house and the energy cost are included in the cost function to evaluate how these properties are affected by different system designs. The MPC solution is compared to the global optimal solution using dynamic programming indicating comparable performance. The robustness of the MPC is evaluated using a prediction of the future household electric consumption in the controller. Results also show that a significant part of the cost reduction is achieved for as small prediction horizons as five hours. Analysis shows that including an accumulator tank is useful for reducing the total energy cost, while reducing the peak power is mainly achieved by increasing the prediction horizon of the MPC.