CAREER: Learning and Control Algorithms for Electricity Demand Response with Humans-in-the-Loop

点击翻译按钮获取中文摘要

Electricity Demand response (DR) technologies aim to inform electricity end-users about the operational state of the power grid and enable them to take an active role in grid operations. Specifically, DR programs can incentivize price-responsive customers to consume more electric energy when supply is abundant and less when supply is scarce, hence increasing the customers' ability to opportunistically consume wind and solar energy and relieve sources of grid stress. However, two main challenges of balancing demand and supply in the grid under such scenarios is the unpredictability of renewable energy outputs, as well as the need to know customers' price response behavior, i.e., how they change their electricity consumption patterns given different prices. Such information is not available to electricity retailers and cannot be easily solicited from the customers either. The goal of this project is to advance scientific knowledge on the design of DR mechanisms that operate in the absence of reliable generation forecasts and customer price response models. To demonstrate the value of our proposed algorithms for a specific type of price responsive electricity demand, we will highlight the application of our methods for electric vehicle smart charging in public parking lots and fast charging stations. The proposed CAREER plan integrates research with teaching and training activities that provide exposure to exciting opportunities in power systems engineering to students at University of California - Santa Barbara, both at the undergraduate and graduate levels. Our outreach efforts will introduce smart grid concepts to local middle and high school students.The intellectual merit of this project is to build on algorithmic techniques for online learning and control of stochastic systems with unknown parameters in order to generate novel systematic tools for demand response and retail market operation in the presence of high levels of uncertainty. Due to the direct involvement of humans in the DR control loop and high levels of renewable integration, a principal challenge we now face is how we can optimize system operations and dispatch resources in the absence of: 1) algebraic models of the system, e.g., due to unknown price response of customers; 2) a characterization of the uncertainty faced by the system in the future, e.g., due to renewables. We will develop real-time dispatch and pricing solutions that provide optimality guarantees in the face of uncertainty about users demand flexibility as well as grid conditions. Our proposed learning and optimization techniques will be integrated with grid reliability and cyber-security constraints. Furthermore, to study the practical impact of this proposal for an important application, we study online learning and dispatch methods for the mobility-aware real-time electric vehicle (EV) demand management problem, which suffers from both of the challenges highlighted above.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.