CyberSEES Type 2: Achieving Clean Power System Flexibility: Sensing, Modeling, and Optimal Control

不断增长的可再生能源发电能力正在导致电力系统运行出现前所未有的变数和不确定性，电力系统灵活性的缺乏威胁到全球电力部门进一步提高全球可持续性。该项目寻求聚合和控制需求侧资源，以通过分布式控制框架提供这种灵活性，该框架对网络基础设施中的故障具有健壮性，并且可以跨机构动态、能源市场和电力系统开发基础设施级别进行扩展。该提案中的研究目标将推动当前在计算、建模和控制方面的应用和理论，这项工作的有效性和可扩展性将通过尼加拉瓜的实地试验得到严格证明。该项目将产生新的工具，以促进全球可持续能源系统，一个开放的数据平台，加强全球研究人员为整合可再生能源所面临的挑战制定全球解决方案的能力，以及教育方面的进展，以使劳动力配备新的工具，以应对可持续发展的挑战。设想的系统采用集中式计算基础设施来为各个恒温控制负载产生最优控制律。控制律可以在有或没有可靠的网络基础设施的情况下实时运行；在没有网络基础设施的情况下，控制只能基于使用时间和本地频率测量。控制律将由中央服务器发布，并由本地情报机构在每次加载时定期访问。这项工作面临的挑战包括创建能够主动分发控制规则并验证参与的计算基础设施，对真实世界的负载有效的建模和控制方法，以及确定具有有效激励机制的控制策略以供消费者参与。在项目结束时，研究人员将有：(1)建立一个框架，借以使一个名为Mezuri的创新的云端数据管理平台能够有力地跟踪所有数据的来源和处理，并传播控制规律；(2)开发用于大型热控负荷状态估计和预测的分层统计模型，其中直接包括人类行为的影响，并调整模型以使所需的观测量尽可能少；(3)设计新的随机最优控制公式，其中考虑到人类活动，并兼顾系统和地方一级的目标。总而言之，这些贡献构建了一个新的框架，用于控制需求侧资源，该框架在短时间内可在全球范围内扩展。

Growing renewable generation capacity is leading to unprecedented variability and uncertainty in power system operations, and a lack of power system flexibility threatens further increases in global sustainability in electricity sectors across the globe. This project seeks to aggregate and control demand-side resources to provide this flexibility through a distributed control framework that is robust to failures in cyber-infrastructure, and is scalable across institutional dynamics, energy markets, and levels of power system development infrastructure. Research objectives in this proposal will advance current applications and theories in computing, modeling and control, and the effectiveness and scalability of this work will be rigorously demonstrated through a field trial in Nicaragua. This project will result in new tools to facilitate sustainable energy systems throughout the globe, an open data platform that strengthens the ability of global researchers to develop global solutions to challenges underling the integration of renewable sources, and advances in education to equip the workforce with new tools to tackle sustainability challenges. The envisioned system employs a centralized computing infrastructure to produce optimal control laws for individual thermostatically controlled loads. The control laws can operate with or without reliable cyber-infrastructure in real time; in the absence of cyber-infrastructure, control can be based on time of use and local frequency measurements only. Control laws will be published by a central server and periodically accessed by local intelligence at each load. Challenges to this work include creating computing infrastructure that can actively distribute control laws and validate participation, modeling and control methods that are effective on real-world loads, and identifying control strategies with effective incentive mechanisms for consumers to participate. At the conclusion of the project the investigators will have: (1) built a framework by which an innovative, cloud-based data management platform called Mezuri can robustly track the provenance and processing of all data and disseminate control laws; (2) developed hierarchical statistical models for state estimation and prediction of large fleets of thermally controlled loads that directly include the effects of human behavior, and adapt models to require as few observations as possible; and (3) devised new stochastic optimal control formulations that account for human activity and accommodate both system- and local-level objectives. Taken together, these contributions build a new framework for control of demand-side resources that is globally scalable in a short time horizon.