Data-Driven Voltage VAR Optimization Enabling Extreme Integration of Distributed Solar Energy

点击翻译按钮获取中文摘要

The increasing penetration of solar energy poses significant challenges on the safe and reliable operation of power distribution systems. This project will leverage data-driven and machine learning techniques to address voltage fluctuations induced by volatile solar generation. It will significantly advance the state-of-the-art of voltage regulation, enable utility companies to address overall voltage issues, and ultimately support the large-scale solar integration in power distribution grids, thus providing higher-quality, more reliable and cleaner electricity to millions of customers across the United States. The incorporation of electrical engineering, data analytics, statistics, and optimization knowledge will foster the multidisciplinary education of graduate and undergraduate students, promote teaching and training of future workforce, and improve scientific and technological understanding through dissemination of findings to academia, industry, and the general public.Conventional voltage VAR optimization (VVO) algorithms, which are model-based, computationally intensive, offline and non-scalable, cannot meet the operation requirements of a modern power system. Important technical issues such as rapid changes of two-way power flows, coordination of new and legacy VVO devices, and lack of accurate system circuit models, will need to be resolved to accommodate a very high penetration level of solar energy. This project will develop a comprehensive data-driven VVO framework that leverages voluminous sensor and meter data to identify real-time system models, perform online prediction of nodal voltages, and orchestrate voltage control devices across different time scales to address severe voltage violations and fluctuations induced by reverse power flows and volatile renewable outputs. The new data-based VVO technique is distinguished from existing methods as it is exempt from the requirement of detailed circuit models, and can achieve high scalability and a significant speed-up of computation time without sacrificing the robustness and accuracy of VVO commands thanks to the linear superposition nature in the proposed modeling and optimization methods. The effectiveness and readily application of the developed techniques will be validated using practical distribution system models and operation data obtained from utility collaborators. The project will benefit from the PI's strong collaboration with utility companies to ensure a pathway for the successful implementation of the outcomes in the real world.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.