As the principal primary energy consumer, buildings could contribute significant energy savings and carbon emission reductions through the design and management of building energy systems. The residential sector contributes a large fraction of building energy consumption due to the huge number of residential buildings, and most energy consumption involves electricity, especially in urban cities. For example, nearly 21% of Hong Kong’s total energy consumption in 2016 was consumed in the residential sector, and over 70% of the residential sector’s energy demands were supplied by electricity.

This study aims to develop the critical technologies for electricity-centric smart home energy management (SHEM) systems. Current literature, however, takes an ad-hoc approach to SHEM. There is no unified framework for standard-based SHEM development and implementation, in contrast to the current building efficiency evaluation code (i.e., the LEED and BEAM building efficiency rating systems). Moreover, efficient and economical SHEM deployment is impeded by several technical barriers. First, the home energy consumption details lay the foundation of home energy management, as they tell us where the invisible electricity comes from and goes away. Current household-level smart meters do not capture the appliance-level energy consumptions and are mainly used for logging the energy consumption data monthly for remote bill-charging purposes. Second, by forecasting future consumption, residential energy systems can be scheduled proactively for economic benefit or occupant comfort. The low predictability of household demands, however, reduces the benefit of proactive scheduling. A robust, scalable, and accurate household-level short-term load forecasting model is still needed. Third, the residential photovoltaic (PV) –battery systems are becoming efficient and economically viable to reduce the dependency on the grid system. However, sizing and operating residential PV–battery systems are still technical barriers for household owners: the uncertain household load and PV output, and the battery degradation, bring uncertainties to the PV–battery systems' worth. Fourth, coordination among multiple SHEMs has prompted significant research interests. While a range of coordination structures and approaches have been proposed and discussed, few researchers focus on the vertical comparison among various coordinating applications, such as peak shaving, demand response (DR) aggregation, local energy transaction. The benefits of evaluating different coordination applications are two-fold. It can not only provide support for selecting employed coordination structures and approaches but also help conduct systematic cost-benefit analyses.

This research addresses the above challenges by proposing a pyramid taxonomy for standard-based development and implementation of SHEM. The research sub-objectives in different development stages are reviewed and organized by a pyramid structured into monitoring, analyzing & forecasting, scheduling, and coordinating layers. Then, a specialized event-based non-intrusive load monitoring (NILM) model is developed for non-linear appliance energy consumption monitoring, which can be combined with other NILM models for more accurate non-intrusive load monitoring in households. Next, a Kalman filter-based, bottom-up approach is proposed to improve day-ahead household load forecast accuracy based on the disaggregated energy consumption details. Then, the sizing and operating problems of residential PV–battery systems are investigated based on an abstract mathematic model. Two two-stage stochastic programming models are established to account for the uncertain load/PV forecasts and battery degradation effects, and to deal with decision-making in different stages (sizing vs. operating, day-ahead scheduling vs. real-time scheduling). Finally, three typical coordination applications with different mechanisms are presented and compared. 

The core contribution of this work lies in the identification of the critical SHEM functions, including monitoring, inferencing and forecasting, scheduling, and coordinating, and their integration into a single study. Then, possible or partial solutions were proposed to address the technical challenges in each functional layer. This research makes it possible to non-intrusively monitor typical nonlinear home appliances’ energy consumptions, infer occupant preference of appliance usages, and to optimally schedule and coordinate the energy consumption behaviors of residential homeowners. The results offer six useful outcomes: 1) a technical pathway for standard-based development and implementation of SHEM; 2) an integrated study for the main SHEM functions: monitoring, analyzing (user preference inference), forecasting, scheduling, and coordinating; 3) a specialized, event-based, non-intrusive load monitoring model for identifying the usage of nonlinear appliances; 4) a Kalman filter-based, bottom-up approach for improving household load forecast accuracy; 5) an optimal solution for the sizing, day-ahead and real-time scheduling of grid-connected residential PV–battery systems; 6) three mechanisms to coordinate multiple SHEMs for different applications: household peak shaving, PV power-sharing, and distribution network voltage regulation. Overall, this study offers a conceptual model for new researchers and practitioners to gain an overview and to find a standard development and implementation path through the multi-faceted SHEM problem. House owners could be guided in investing economically in the residential PV–battery systems, and in operating their building energy systems efficiently. In the long term, this study aims to promote the utilization of solar energies in the residential sector in order to transition conventional residential buildings toward active and efficient energy systems.