CAREER: Computation-efficient Algorithms for Grid-scale Energy Storage Control, Bidding, and Integration Analysis

点击翻译按钮获取中文摘要

Energy storage is a cornerstone in future low-carbon power systems for reducing carbon emissions and enhancing power system reliability against extreme events. This NSF CAREER project aims to develop new computation tools aiding grid integration of energy storage with the overarching goal to provide affordable and reliable electricity supply in sustainable power systems. The project will bring transformative change to enable power system operators and storage owners to more accurately bid and dispatch a variety of existing and emerging storage technologies. This will be achieved by developing a novel computation framework combining model-based optimization with machine learning, achieving both reliable performance and computation efficiency. The intellectual merits of the project include developing novel control algorithms for complex storage energy models and investigating approaches to integrate existing and emerging storage technologies into electricity markets. The broader impacts of the project include developing university curricula and K-12 outreach programs on incorporating data science into energy decarbonization and climate change education, and developing an outreach program to promote community solar plus storage deployments with a focus on disadvantaged neighborhoods in New York City.The project simultaneously addresses several technical challenges in energy storage grid integration including multi-stage uncertainties, nonlinear and nonconvex storage models, and computation scalability over a large number of networked storage resources. The project will i) develop a fully open-source analytical algorithm without proprietary commercial solvers tailored for energy storage to solve nonlinear stochastic dynamic programming with extreme computation speed; ii) develop new market models and pricing schemes inspired by the opportunity value function from dynamic programming to economically manage storage state-of-charge in grid dispatch; iii) combine machine learning with dynamic programming into a two-stage learning model to more efficiently analyze and manage a large number of storage resources participating in electricity markets. The results of this project will benefit power system operators and storage owners to develop energy management system software for storage resources that more accurately reflect the storage operating characteristics and future uncertainties, and aid education and outreach activities related to energy storage deployments for energy sustainability and resiliency.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.