计及海量异构数据的电力系统电压稳定状态在线深度认知  

Online deep cognition of power system voltage stability state with massive heterogeneous data

作　　者：许洵 黄宗武 胡浩[1,2] 樊友平[1,2] 范建斌 舒印彪 XU Xun;HUANG Zongwu;HU Haol;FAN Youping;FAN Jianbin;SHU Yinbiaol(School of Electrical Engineering and Automation,Wuhan University,Wuhan 430072,China;Institute for New Power Systems and International Standards,Wuhan University,Wuhan 430072,China)

机构地区：[1]武汉大学电气与自动化学院,湖北武汉430072 [2]武汉大学新型电力系统与国际标准研究院,湖北武汉430072

出　　处：《武汉大学学报（工学版）》2024年第10期1349-1359,共11页Engineering Journal of Wuhan University

基　　金：国家重点研发计划项目国家质量基础设施体系专项(编号:2022YFF06010600)。

摘　　要：为充分利用电网中潜在关联的多组特征和高维多源异构数据,提出一种基于多变量合作学习、最小绝对收缩和选择算法(leastabsolute shrinkage and selection operator,LASSO)的电压稳定裕度在线评估方法。首先,利用合作学习算法在无功储备、节点电压等系统异质运行参数之间形成最佳融合模式,并通过局部加权LASSO回归工具建立评估模型。然后,设计数据库自动更新系统,实现模型对运行工况的检测与自动更新。最后,采用IEEE30节点和1951节点系统对所提方法进行验证。结果表明该方法在功率增长方向、运行方式等变化情况下,具有良好的准确性及泛化性。In order to fully utilize the potentially correlated multi-set features and high-dimensional multi-source heterogeneous data in power grids,an online assessment method for voltage stability margin based on multivariate cooperative learning and least absolute shrinkage and selection operator(LASSO)is proposed.Firstly,the cooperative learning algorithm is used to form the best fusion mode between the system heterogeneous operating parameters,such as reactive power reserve and node voltage,and then an evaluation model is established by locally weighted LASSO regression tool.Then,the database automatic updating system is designed to realize the model's detection and automatic update of operating conditions.Finally,the proposed method is validated by using IEEE 30-node and 1951-node systems.The results show that the proposed method has good accuracy and generalization under the changes of power growth direction and operation mode.

关 键 词：静态电压稳定裕度 异构数据融合合作学习 最小绝对收缩和选择算法 局部线性回归 

分 类 号：TM712[电气工程—电力系统及自动化]