CAREER: Stochastic Optimization and Physics-informed Machine Learning for Scalable and Intelligent Adaptive Protection of Power Systems

这个NSF职业项目旨在通过设计数据驱动的自适应保护平台(APP)来提高电网的弹性。该项目将通过设计智能和自适应保护方案来带来变革性的变化，以应对不同运行模式和电路拓扑的现代电网以及基于逆变器的资源(IBR)的高渗透率下的挑战。这些挑战可能会恶化常规保护计划的性能，并可能导致广泛停电等有害影响。因此，有必要重新设计传统的保护系统，使其适应当前电网的条件。这将通过设计一个可扩展的APP来实现，该APP可以在输配电网中采取自适应保护行动。该项目的智力优势包括通过结合软件和硬件解决方案来解决IBR高渗透率带来的保护挑战，这些解决方案可以提高基础保护系统的可靠性、选择性、敏感性和安全性。该项目的更广泛影响包括扩大代表不足的群体对电力工程的参与，并将实用和现实世界的概念纳入电力工程的现有课程。这将通过在新墨西哥大学(UNM)为代表不足的K-12和大学生组织夏令营和其他外展活动，并为本科生和研究生设计新的课程主题来实现。该项目的研究目标是(I)设计一个使用随机优化算法来优化继电保护整定值以响应极端事件的自适应保护平台，(Ii)创建无通信和自适应的本地保护模块。提出的研究将建立一个多阶段随机优化问题，以确定满足极端事件引起的不同电路拓扑场景下的继电保护协调时间间隔约束的可行继电保护整定。另一方面，局部自适应保护模块将使用无监督条件生成对抗网络(C-GAN)进行故障检测，并使用物理信息机器学习算法进行故障定位。物理信息机器学习算法将利用故障后顺序元件网络的方程来正规化估计的故障位置和电阻。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

This NSF CAREER project aims to improve the resilience of power grids by designing a data-driven adaptive protection platform (APP). The project will bring transformative change by designing intelligent and adaptive protection schemes in response to challenges associated with modern power grids with different operational modes and circuit topologies and under high penetration of Inverter-based Resources (IBRs). These challenges can deteriorate the performance of conventional protection schemes and may result in detrimental impacts like widespread blackouts. Therefore, there is a need to redesign the conventional protection systems and make them adaptive to the prevailing power grid conditions. This will be achieved by designing a scalable APP that can take adaptive protection actions in transmission and distribution electric power grids. The intellectual merits of the project include addressing the protection challenges that rise from the high penetration of IBRs by incorporating software and hardware solutions that improve the reliability, selectivity, sensitivity, and security of the underlying protection system. The broader impacts of the project include broadening the participation of underrepresented groups in power engineering and integrating practical and real-world concepts into the existing curriculum of power engineering. This will be achieved by organizing summer camps and other outreach activities for underrepresented K-12 and college students and designing new course topics for undergraduate and graduate students at the University of New Mexico (UNM).The research objectives of this project are (i) to design an adaptive protection platform that is responsive to extreme events using a stochastic optimization algorithm for optimizing protection relay settings, and (ii) to create communication-free and adaptive local protection modules. The proposed research will formulate a multi-stage stochastic optimization problem to identify feasible relay settings that satisfy the relay’s coordination time interval constraints for different circuit topology scenarios caused by extreme events. On the other hand, the local adaptive protection module will be designed using unsupervised conditional generative adversarial network (C-GAN) for fault detection and physics-informed machine learning algorithms for fault location. The physics-informed machine learning algorithms will utilize the postfault sequential component networks’ equations for regularization of estimated fault location and resistance.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.