CAREER: Synthesis of Feedback-based Online Algorithms for Power Grids

点击翻译按钮获取中文摘要

This CAREER proposal focuses on power grids, and aims to translate foundational theory and algorithms into breakthrough real-time optimization and control approaches for distributed energy resources (DERs). In this context, the overarching goal is to overcome current technological and operational barriers associated with the large-scale integration of DERs, where: (a) the deployment of DERs with business-as-usual practices has decreased power-quality and reliability, (b) existing network optimization approaches may fail to provide solutions at a time scale that matches the dynamics of power systems with DERs, and (c) synthetic models for users’ preferences and comfort may not capture the users’ goals truthfully. The research plan seeks a shift from a paradigm with a time-scale separation between economic optimization and local control – predominant in today's distribution grids, where corrective and localized rules serve as a basis for real-time voltage regulation and ancillary-service provisioning – to operations where DERs actively partake into grid operations and leverage real-time network-level coordination to seek increased efficiency and reliability. DER coordination is engineered so that DERs can learn to maximize users' preferences, while aiding system-level frequency and voltage control. An integrated education and outreach plan will engage middle- and high-school students through a summer Science, Technology, Engineering and Mathematics (STEM) Research Academy and lectures for the Pre-Collegiate Development Program. To bridge research and education, the PI will develop courses on the themes of online optimization for networks and optimization of power systems, and will promote undergraduate student research. A working group on optimization and learning will be created at the University of Colorado Boulder in synergy with the Autonomous Systems Interdisciplinary Research Theme, to bring together faculty and students across the campus and stimulate multi-disciplinary research and education.The proposed research leverages time-varying optimization models for networks operating in dynamic environments, and seeks to develop real-time optimization architectures with tightly-integrated feedback and learning components. The proposed feedback-based online algorithms have the following key attributes: i) Principled algorithmic steps employ measurements from the network to bypass the need for a network model; ii) Algorithms include humans in the loop by learning the users' utility functions from users' feedback during the execution of the online decision algorithm; iii) Algorithms are implemented in closed loop with the power network to acknowledge dynamics and effectively act as feedback controllers; and, iv) Algorithms promote low-complexity, distributed, and scalable architectures. Fundamental tradeoffs between convergence rate, tracking of time-varying optimal solutions, maximum constraint violation, and computational complexity of the algorithms will be offered.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.