Data-driven adaptive robust operation of PV generation in distribution systems

本研究的目的是开发一种新的数据驱动的决策支持系统（DSS），以确定有效的短期运营策略，以适应大规模光伏发电并减轻其对配电网可靠性和安全性的不利影响。本文提出的分布式决策支持系统（DSS）将改善配电网光伏发电的时空变异性和不确定性量化；2)确定光伏发电可调节的可变性和不确定性边界，以保证配电网的经济效率和安全；3)提出具有成本效益的动态解决方案，考虑配电网络中供需的时空变化和不确定性；4)捕捉微电网、分布式能源（DERs）和可控需求等自治实体之间的相互作用；与配电系统运营商（DSO）合作。这项研究计划有助于将所产生的知识迅速传播给研究和教育界。具体而言，它促进了研究生和本科生之间的创新合作，为当前配电网运营中的挑战提供有效的解决方案。该项目通过为本科和研究生项目引入新课程、促进跨学科合作、招募未被充分代表的少数民族和女学生以及开展K-12外展活动，确保最高质量的综合研究和教育，以满足美国能源部门新兴的劳动力和教育需求。本研究的具体目的如下：a)开发一种可扩展的数据驱动方法，利用多任务深度学习框架，为配电网中的大规模光伏发电提供改进的时空不确定性措施。B)将灵活性措施量化为三级调节服务，并形成分布自适应鲁棒优化问题，以量化可容纳的时空变异性和不确定性。C)为不平衡配电网的非凸风险规避短期运行问题提供了紧凸松弛。可行性集的非凸性主要是由于引入了用于切换和承诺决策的整数变量以及不平衡的交流潮流约束。d)开发分散的优化框架，以捕获自治实体和DSO做出的动态时间决策之间的空间相互依存关系。

The objective of this research is to develop a novel data-driven decision support system (DSS) to determine efficient short-term operation strategies for accommodating large-scale PV generation and mitigating its adverse effects on distribution network reliability and security. The proposed DSS will 1) improve the spatiotemporal variability and uncertainty quantification for PV generation in distribution networks; 2) determine the accommodated variability and uncertainty boundaries of PV generation to ensure the economic efficiency and security of the distribution networks; 3) propose cost effective dynamic solutions that incorporate the temporal and spatial variability and uncertainty of demand and supply in the distribution networks; 4) capture the interactions among autonomous entities such as microgrids, distributed energy resources (DERs), and controllable demands; with distribution system operator (DSO). This research plan facilitates rapid dissemination of the generated knowledge to the research and education community. Specifically, it promotes innovative collaboration among graduate and undergraduate students to provide effective solutions for the current challenges in the distribution network operation. This project ensures the highest quality of integrated research and education to meet the emerging workforce and educational needs of the U.S. energy sector by introducing new curriculum for undergraduate and graduate programs, promoting interdisciplinary collaboration, recruiting underrepresented minorities and female students, and developing K-12 outreach activities.The specific objectives of this research are as follows. a) develop a scalable data-driven approach that leverages a multi-task deep learning framework to provide improved spatiotemporal uncertainty measures for the large-scale PV generation in the distribution network. b) quantify the flexibility measures as tertiary regulation services and form distributionally adaptive robust optimization problems to quantify the accommodated spatiotemporal variability and uncertainty. c) provide a tight convex relaxation for the non-convex risk-averse short-term operation problem for the unbalanced distribution networks. The non-convexity in feasibility set is as a result of the introduced integer variables for switching and commitment decisions as well as the unbalanced AC power flow constraints. d) develop decentralized optimization framework to capture the spatial interdependence among the dynamic temporal decisions made by the autonomous entities and the DSO.