The National Science Foundation (NSF) project: “Crosslayer Optimization of Energy and Cost through Unified Modeling of User Behavior and Storage in Multiple Buildings,” awarded last year to computer science assistant professor Simone Silvestri and electrical and computer engineering professor, L. Stanley Pigman Chair in Power and PEIK Director Dan M. Ionel, has recently been expanded with a Research Experiences for Undergraduates (REU) supplement. This will allow electrical engineering major Rosemary Alden and computer science major Nelson Penn to participate in research alongside Ph.D. students and their faculty advisors. Both Penn and Alden are in the University Scholars Program.
This NSF project studies newly proposed transformative concepts of "human-in-the-loop," social-behavioral models, machine learning and large-scale modeling of buildings and electric power distribution systems, leading to the development of highly efficient and reliable smart homes and grids. The research is expected to also strengthen PEIK's partnership with regional utilities, the Electric Power Research Institute (EPRI) and National Labs.
The abstract for the project is below.
Crosslayer Optimization of Energy and Cost through Unified Modeling of User Behavior and Storage in Multiple Buildings
The building sector is the largest energy consumer in the world, and in the United States it accounts for more than 40 percent of the total energy consumption and greenhouse gas emissions. Therefore, it is economically, socially, and environmentally important to reduce the energy consumption of this sector. The goal of this collaborative proposal is to develop novel machine learning-based algorithms to address the problem of energy optimization at the building and district levels. These algorithms are integrated within a simulation framework that combines user behavior with the collaboration between buildings equipped with photovoltaic arrays, energy storage systems, and smart grid meters. The proposed research is expected to lay the foundation for robust multi-objective optimization for next-generation district level distribution systems. The proposed research is closely integrated with a broad and diverse education and outreach plan aimed at inspiring young women to pursue careers in STEM through summer programs for middle school. Additionally, the project will train the next generation of engineers and researchers by involving graduate and undergraduate students through the proposed research as well as through classes taught by the PIs encompassing the proposed research methodologies. Overall, the outcomes of this proposal are expected to significantly advance the areas of energy optimization, electric power systems, and smart grid design, as well as to have a positive impact on the academic and industrial communities and society.
The project proposes and integrates, within the same software tool, novel machine learning models for complex user behavior at the individual building level, for energy load prediction and energy storage systems scheduling at the district level, and for cost reduction via energy peak spreading. These models are used to formulate and construct algorithmic solutions based on reinforcement learning, recurrent and deep neural networks, and deep reinforcement learning suitable for implementation in the future generation Virtual Power Plants. The methodologies employed for energy reduction and cost minimization include: 1) alter user behavior through personalized recommendations regarding changes in the appliance states (e.g., heating and air conditioning settings), 2) district-level scheduling of energy storage systems among buildings equipped with photovoltaic arrays and smart grid meters, and 3) building-level scheduling of energy consumption events for smart appliances equipped with smart Internet-of-Things controllers to take benefit of different energy prices.