Energy Data Analytics for Reliable and Efficient Electric Grid Operations

在整个北美，相量测量单元（pmu）和智能电表已经在整个电网中广泛部署。从这些设备收集的数据的数量和种类有可能实现广泛的电力系统监测和操作任务。具体而言，该项目将利用从pmu和智能电表收集的测量数据来改进[T1]实时暂态稳定性评估，以及[T2]需求响应的在线学习。暂态稳定性将是未来电力系统的一个重要问题，主要是由于可再生能源发电的迅速变化，预计可再生能源发电将逐渐但广泛地取代基于化石燃料的技术。基于机器学习的以数据为中心的方法已经被证明可以高精度地预测系统是否稳定。在[T1]中，我们将与我们的行业合作伙伴一起，扩展现有的工具，以确定不稳定的根本原因，以便电力公司能够及时修复损坏并恢复正常的系统运行。需求响应计划有助于提高电力系统的可靠性和市场效率，这可以通过实时定价来激励客户从需求高峰时期转移他们的用电量来实现。在这里，一个主要的挑战是预测人类的行为，这可以通过在线学习算法来解决，但它们通常忽略了电子网络效应。在[T2]中，我们将研究电网对需求响应策略的实时实现的影响，它施加了重要的物理和操作约束。与我们的行业合作伙伴密切合作，提出的工具将被原型化并实施到他们的商业级软件中，以扩展其对能源数据分析的适用性。通过利用整个电网收集的数据，项目成果有助于确保电力的可用性和质量，以应对高水平的可再生能源渗透和客户参与所带来的日益增长的不确定性。

Across North America, phasor measurement units (PMUs) and smart meters have been extensively deployedacross the electric grid. The volume and variety of the data collected from these devices have the potential toenable a broad array of power system monitoring and operations tasks. Specifically, this project will leveragemeasurement data collected from PMUs and smart meters to improve [T1] real-time transient stabilityassessment, and [T2] online learning for demand response.Transient stability will be an important concern in the future power system, primarily due to rapidly varyingrenewable generation, which is expected to gradually but widely displace fossil-fuel-based technologies.Data-centric approaches based on machine learning have been shown to predict whether or not the system willbe stable with high accuracy. In [T1], along with our industry partner, we will extend existing tools to alsoidentify the root cause of instability, so that electric utilities can repair damages and restore normal systemoperations in a timely manner.Demand response programs help to improve power system reliability and market efficiency, which can beachieved by incentivizing customers via real-time pricing to shift their electricity usage away from periods ofpeak demand. Here, a major challenge is in predicting human behaviours, which can be solved via onlinelearning algorithms, but they generally neglect electrical network effects. In [T2], we will investigate theimpact of the electric network, which imposes nontrivial physical and operational constraints, on real-timeimplementations of demand response strategies.In close collaboration with our industry partner, the proposed tools will be prototyped and implemented intotheir commercial-grade software to extend its applicability to energy data analytics. By leveraging datacollected across the electric grid, the project outcomes help to ensure power availability and quality in the faceof growing uncertainty arising from high levels of renewable penetration and customer participation.