AMPS: Deep Stochastic Models for Space-Time Weather-Driven Grid Simulations

随着分布式光伏和具有需求响应潜力的负荷等大量分布式能源的引入，电力系统正在迅速发生变化。增加不确定和可变的分布式能源需要更好地了解对电网运行的潜在影响。在电网上直接测试是不可能的，而依赖于计算机模拟不同程度的分布式能源渗透是必要的。伴随的数学挑战是在客户层面生成供应和负载的现实场景，通常具有高空间和时间分辨率，涉及数万个相关时空变量的联合生成。此外，由于负载中的电气设备切换和分布式光伏电源中的微尺度天气变化，这些过程受到不连续性的困扰。该项目代表了端到端的努力，开发新的数学工具，将应用于现实的电网测试平台。本研究将开发一个时空跳跃-扩散过程的框架，以捕获高时间频率下分布式能源和负载的真实非高斯行为。这些模型将对分布式能源对当前和未来电网的影响产生前所未有的见解。新框架和方法的测试平台是最先进的配电模拟网络数据集Smart-DS。通过研究电网对不同分布式能源采用水平的响应，本研究的结果将对基础设施和能源规划产生潜在影响。这些统计模型将适用于金融数学、环境计量学、地理学和生态学等各个领域。此外，这项研究将支持统计和能源科学交叉领域的学生培训，他们将在国家可再生能源实验室的夏季实习期间获得与领域科学家的跨学科和合作经验。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 power system is changing rapidly with the introduction of large amounts of distributed energy resources such as distributed photovoltaics and loads with demand response potential. The addition of uncertain and variable distributed energy resources requires better understanding of potential impacts on grid operations. Direct testing on the grid is not possible, and the reliance on computer simulations for varying degrees of the distributed energy resource penetration is necessary. The companion mathematical challenge is in generating realistic scenarios of supplies and loads at the customer level, often at high spatial and temporal resolutions involving joint generation of tens of thousands of correlated space-time variables. Moreover, such processes are plagued by discontinuities due to electrical device switching in loads and microscale weather variations in distributed photovoltaics supply. This project represents an end-to-end effort developing new mathematical tools that will be applied to realistic electricity network testbeds.This research will develop a framework for space-time jump-diffusion processes that capture realistic non-Gaussian behavior of distributed energy resources and loads at high time frequencies. Such models will yield unprecedented insights into distributed energy resource implications on current and future grids. The testbed for the new framework and methodology is a state-of-the-art distribution simulation network dataset, Smart-DS. Results of this research will have potential impacts in infrastructure and energy planning by studying grid responses to varying levels of the distributed energy resource adoption. The statistical models will be applicable to a variety of fields including financial mathematics, environmetrics, geography and ecology. Moreover, this research will support student training at the intersection of statistics and energy science, who will gain interdisciplinary and collaborative experience with domain scientists during summer internships at the National Renewable Energy Laboratory.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.