多故障场景下基于主动学习的暂态电压安全评估与安全裕度估计  

Active learning‐based transient voltage safety assessment and safety margin estimation for multi‐fault scenarios

作　　者：陈阳[1] 陆超[1] 刘迪 贺静波 CHEN Yang;LU Chao;LIU Di;HE Jingbo(Department of Electrical Engineering,Tsinghua University,Beijing 100084,China;National Power Dispatching&Control Center,State Grid Corporation of China,Beijing 100032,China)

机构地区：[1]清华大学电机工程与应用电子技术系,北京100084 [2]国家电网有限公司国家电力调度控制中心,北京100032

出　　处：《武汉大学学报（工学版）》2023年第11期1336-1346,共11页Engineering Journal of Wuhan University

基　　金：国家电网总部科技项目(编号:52094021N00N)。

摘　　要：对于电力系统暂态电压安全评估与安全裕度估计的问题,传统的分析方法难以应对新型电力系统的复杂性与多样性,基于数据驱动的方法为暂态电压安全在线评估提供了新思路,但也存在着物理意义模糊、训练复杂度高、可拓展性差等问题。针对上述问题,提出了一种面向多故障场景的基于主动学习的暂态电压安全边界快速近似与安全裕度高效估计方法。首先,基于动态安全域的概念,针对训练复杂度高的问题,提出了基于主动学习的暂态电压安全评估模型与安全裕度估计的数据驱动方法,在保持较高准确率的同时极大地降低了标记样本集的规模,同时得到了解析的安全边界函数;然后,针对不同的故障信息,提出了基于shapelet变换的层次化事故聚类方法,将相似的故障聚集在一起,为每一类故障场景构建安全评估模型,使得暂态电压安全评估模型在一定程度上摆脱了故障依赖,提升了安全裕度在线估计的效率与精度;最后,通过IEEE39节点系统验证了所提方法的正确性,实现了多故障场景下暂态电压安全性的快速辨识与安全裕度的快速估计,有效降低了离线仿真的时间,提升了面向多故障场景的安全评估模型的部署效率。Regarding the issues of transient voltage safety assessment and safety margin estimation of power systems,traditional analysis methods are difficult to cope with the complexity and diversity of new power systems.Data-driven methods provide new ideas for online transient voltage safety assessment,but there are problems such as fuzzy physical meaning,high training complexity,and poor scalability.In order to solve these challenging problems,this paper proposes an active learning-based fast approximation of transient voltage safety boundaries and efficient estimation of safety margins for multi-fault scenarios.First,based on the concept of dynamic safety region,in order to solve the problem of high training complexity,a data-driven method of transient voltage safety assessment model and safety margin estimation based on active learning is proposed.This method greatly reduces the size of the labeled sample set while maintaining a high accuracy,and at the same time obtains an analytical safety boundary function.Then,aiming at the different fault information,a hierarchical fault clustering method based on shapelet transformation is proposed.This method clusters similar faults together to build a safety assessment model for each type of fault scenarios.This enables the transient voltage safety assessment model to get rid of fault dependence to a certain extent and improves the efficiency and accuracy of online estimation of safety margins.Finally,the correctness of the method proposed in this article is verified through the IEEE39-bus system.The proposed method achieves rapid identification of transient voltage safety and rapid estimation of safety margin in multi-fault scenarios,effectively reduces the time of offline simulation,and improves the deployment efficiency of safety assessment models for multi-fault scenarios.

关 键 词：安全域 安全裕度 数据驱动 主动学习 shapelet聚类学习 

分 类 号：TM769[电气工程—电力系统及自动化]