Collaborative Research: Learning-Assisted Estimation and Management of Flexible Energy Resources in Active Distribution Networks

该NSF项目旨在开发基于学习的新方法来估计太阳能、太阳能和存储等网格边缘资源(GERS)或智能温控器的灵活性，并基于多智能体和分布式优化方法设计公平的资源协调和管理方法。该项目将为配电网GER管理领域带来革命性的变化，将机器学习(ML)和人工智能(AI)与此类资源的基于物理的模型相结合，以估计电网级别的地理空间灵活性，并开发一种用于GERS协调的多时间尺度分布式优化方法，以提供网格服务。该项目的结果预计将对电网的可靠性和复原力产生重大影响，同时为客户提供新的财务和资金机会。该项目的智能优势包括新的基于物理/数据驱动的混合GERS灵活性评估方法及其不确定性，以及创建可配置的、多时间尺度的分布式优化，以根据客户的计算和通信能力提供快速和缓慢的网格服务。该项目的更广泛影响包括通过印刷媒体、广播新闻和互联网整合教育公众，并为代表性不足的学生提供教育和研究机会。该项目将在以下四个方向推进配电网灵活能源的管理。第一个方向将是利用生成性ML技术，并利用来自附近可观测的表后(BTM)太阳能和存储资产的空间、时间和通道信息来解决数据差距。这种方法增强了对这些BTM单元的可用性和灵活性的估计。第二个方向将是开发一种地理空间灵活性评估方法，以改进智能恒温器负载的特征。该方法结合了基于物理的模型和数据驱动的模型，以获得预期的功率和能量调整以及相关的不确定性。第三个方向将是构建一个可配置的多时间尺度分布式协调框架，将BTM的灵活性打包为快和慢的网格服务。使最终用户能够提供多时间尺度的电网服务，提高了电力系统的弹性并增加了客户收入。最终的方向将是通过考虑他们在多智能体协调程序中的计算和通信限制来促进服务不足的客户的参与。这一进步将更好地分配社会福利，并释放未得到充分利用的BTM资产的潜力。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 project aims to develop novel learning-based approaches for estimating the flexibility amount of grid edge resources (GERs), such as solar, solar and storage, or smart thermostat, and then design equitable resource coordination and management methods based on multi-agent and distributed optimization approaches. The project will bring transformative changes to the area of GER management in distribution electricity networks by combining machine learning (ML) and artificial intelligence (AI) with the physics-based models of such resources for estimating geo-spatial flexibility at the grid level according, and also by developing a multi-time scale distributed optimization method for GERs coordination to provide grid services. The outcome of this project is expected to have significant impacts on grid reliability and resilience, while providing customers with new financial and monetary opportunities. The intellectual merits of the project include new hybrid physics-based/data-driven flexibility estimation methods for GERs along with their uncertainties, and creation of configurable, multi-time scale, distributed optimization for providing fast and slow grid services according to the customers’ computation and communication capabilities. The broader impacts of the project include integrating educating the public through print media, broadcast news, and the Internet, and providing educational and research opportunities for underrepresented students.This project will advance management of flexible energy resources of distribution grids in the following four directions. The first direction will be in utilizing generative ML techniques and leveraging spatial, temporal, and channel-wise information from nearby observable behind-the-meter (BTM) solar and storage assets to address data gaps. This approach enhances the estimation of availability and flexibility of these BTM units. The second direction will be in developing a geo-spatial flexibility estimation method that improves the characterization of smart thermostat loads. This method combines physics-based and data-driven models to obtain expected power and energy adjustments and associated uncertainties. The third direction will be in building a configurable multi-time scale distributed coordination framework to package BTM flexibilities as fast and slow grid services. Enabling end-use customers to provide multi-time scale grid services increases power system resilience and boosts customer revenue. The final direction will be in facilitating participation of underserved customers by accounting for their computation and communication limitations in multi-agent coordination procedure. This advancement will better distribute societal welfare and unlock potentials of underutilized BTM assets.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.