Data-Driven Distributionally Robust Hierarchical Coordination for Home Energy Management

Home energy management (HEM) is conventionally formulated using central control or local control methods assuming known accurate probability distributions of uncertainties. Without coordination between the central and the local control hierarchies, mutual impacts cannot be properly considered and addressed, which undermine operation efficiency of HEM. Moreover, the probability distributions always have errors, even in emerging data-based methods. To address these unsolved issues, firstly, this paper proposes a new hierarchical coordination method for HEM. On the central hierarchy, a central controller optimizes schedule of non-thermal loads to minimize daily operating cost. On the local hierarchy, automatic local controllers respond to real-time variations of their corresponding thermal zone temperatures to fulfill customer thermal comfort requirements. Secondly, using a finite training dataset, a data-driven distributionally robust optimization (DRO) method is proposed to guarantee solution robustness against worst probability distribution of multiple uncertainties. The hierarchical coordination method is formulated as a DRO model using the Wasserstein metric, which can be solved with high computing efficiency. Numerical simulations validate high solution robustness and overall optimum of the proposed method, fully addressing mutual impacts and uncertainties. Moreover, the proposed method could efficiently save the daily electricity bill by 11%, compared to conventional methods.