Smart home energy management with mitigation of power profile uncertainties and model errors.

The paper proposes a two-stage home energy management system accounting for optimal energy dispatch, power profiles predictability, and building thermal behaviour. The first look-ahead stage aims at minimizing the energy exchanges with the upstream grid based on load and weather predictions. The second step adapts the controls in real-time in order to mitigate forecast errors while remaining as close as possible to the look-ahead commitments. The originality of the paper lies on the use of different granularities of the building models. Especially, this allows to account for the approximations in the system equations embedded in the predictive controllers. Furthermore, a real-time tuning of those model equations is performed based on the deviations observed on the state variables – errors between controller predictions and measurements. Ultimately the proposed approach allows to cope with two generic drawback of traditional model predictive control: i) environment forecast errors and ii) models approximations. Several simulations are carried out to under different weather conditions, demonstrating an average energy exchange reduction of at least 25% and better system predictability. The work is completed with sensitivities analysis of the controller settings, the system parameters and the quality of weather/load forecasts. [ABSTRACT FROM AUTHOR]