摘要
为解决配电网量测数据多重共线性导致电压-功率灵敏度估计精度低的问题,提出融合拓扑信息的数据驱动方法.首先,将电压-功率灵敏度矩阵分解为主成分矩阵和次成分矩阵两部分,其中主成分与配电网拓扑密切相关,次成分为主成分与实际值的误差.然后,分两阶段依次估计主成分和次成分,分别建立基于二次规划的数据驱动模型.第一阶段模型的关键是基于配电网拓扑信息的约束,第二阶段模型的关键是次成分与主成分比值为微小量的约束.最后,采用实际配电网量测数据和IEEE 33节点系统验证所提方法的精度和效率,并与常规最小二乘回归、岭回归、LASSO回归方法对比,仿真结果表明所提方法的精度具有数量级的显著提升.
The multicollinearity of measurement data leads to the low accuracy of the data-driven methods for estimating voltage-power sensitivity in distribution networks.In this paper,a data-driven method embedded with topological information is proposed to address the problem.First,the voltage-power sensitivity matrix is decomposed into principal and secondary components,where the principal component is closely related to the distribution network topology and the secondary component is the error between the principal component and the actual value.Then,the principal and secondary components are estimated sequentially in two stages,and their data-driven estimation models based on quadratic programming are established,respectively.The key of the model in the first stage is the constraint based on the distribution network topology information,and the key of the model in the second stage is the constraint that the ratio of the secondary component to the principal component is tiny.Finally,the accuracy and efficiency of the proposed method is validated in the IEEE 33-bus system with a set of measurement data,and comparisons are made with ordinary least square regression,ridge regression,and LASSO regression.The simulation results show that the accuracy of the proposed method is significantly improved by orders of magnitude.
作者
刘舒
周敏
高元海
徐潇源
严正
LIU Shu;ZHOU Min;GAO Yuanhai;XU Xiaoyuan;YAN Zheng(State Grid Shanghai Electric Power Research Institute,Shanghai 200437,China;Key Laboratory of Control of Power Transmission and Conversion of the Ministry of Education,Shanghai Jiao Tong University,Shanghai 200240,China;Shanghai Non-Carbon Energy Conversion and Utilization Institute,Shanghai 200240,China)
出处
《上海交通大学学报》
EI
CAS
CSCD
北大核心
2024年第6期855-862,I0001,共9页
Journal of Shanghai Jiaotong University
基金
国网上海市电力公司科技项目(52094022003R)。
关键词
电压-功率灵敏度
配电网拓扑
多重共线性
数据驱动
主成分
voltage-power sensitivity
distribution network topology
multicollinearity
data-driven
principal component
