Collaborative Research: AMPS: Deep-Learning-Enabled Distributed Optimization Algorithms for Stochastic Security Constrained Unit Commitment

目前，电力系统的运营格局正在经历一场深刻的变革，其驱动因素包括可再生能源的整合、对清洁能源经济的需求以及应对气候危机的紧迫性。该项目旨在奠定必要的数学基础，以充分利用深度机器学习方法的潜力，增强电力系统的运行，特别是在可再生能源方面，如风能和太阳能发电。该研究将开发一套新的分布式优化工具，使大规模电力系统运行能够有效地管理不确定性，同时有效地利用可再生能源。新的算法将有可能改变电力系统内的操作实践。同时，研究结果将提高利益相关者、监管机构、政策制定者和市场参与者之间的公众意识和理解。该项目的成功完成将使电力系统运营商能够采用先进的算法，显著提高可再生能源发电的运营实践。该项目将为两所院校的学生，特别是来自STEM领域代表性不足群体的学生提供培训和外展机会。该项目旨在开发和验证支持深度学习的分布式随机算法。这些算法将解决电力系统中大规模、随机安全约束的机组承诺问题。具体而言，该项目将侧重于以下目标：(i)设计一个整体的、三阶段的、基于深度神经网络的机器学习方法；（ii）基于混合分布参数系统控制理论的求解策略；（iii）使用大规模真实世界电力系统数据集对所提出的算法进行广泛验证。该研究将通过引入创新技术来解决不确定情况下电力系统运行的相关挑战，从而推动该领域的发展。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The operational landscape of electric power systems is currently experiencing a profound transformation driven by various factors, including the integration of renewable energy sources, the need for a cleaner energy economy, and the urgency to address the climate crisis. This project aims to lay the mathematical groundwork necessary to harness the full potential of deep machine learning approaches in enhancing power system operations, particularly in relation to renewable energy, such as wind and solar generation. The research will develop a new suite of distributed optimization tools that will empower large-scale power system operations to manage uncertainty while incorporating renewable energy resources effectively. The new algorithms will potentially transform operational practices within the power system. At the same time, the results will increase public awareness and understanding among stakeholders, regulators, policymakers, and market participants. The successful completion of this project will enable power system operators to adopt cutting-edge algorithms that significantly enhance their operational practices with renewable generation. The project will provide training and outreach opportunities to students from both institutions, particularly those from underrepresented groups in STEM. The project aims to develop and validate deep-learning-enabled distributed stochastic algorithms. These algorithms will solve large-scale, stochastic security-constrained unit commitment problems within power systems. Specifically, the project will focus on the following objectives: (i) the design of a holistic, three-stage, deep neural network-based machine learning approach; (ii) the solution strategies based on the hybrid distributed parameter system control theory; and (iii) extensive validations of the proposed algorithms using large-scale real-world power system datasets. The research will advance the field by introducing innovative techniques to address the challenges associated with power system operation under uncertainty.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.