动态拓扑下基于终端数据挖掘的配电网异常节点检测
Abnormal Node Detection in Distribution Network Based on Terminal Data Mining in Dynamic Topology
张津铭 1赵健 1唐维溢 1宣羿 2孙智卿 2张凯3
作者信息
- 1. 上海电力大学电气工程学院,上海 200090
- 2. 国网杭州供电公司,浙江杭州 310007
- 3. 上海电力大学计算机科学与技术学院,上海 200090
- 折叠
摘要
异常节点检测用于发现由窃电、设备故障等现象导致的功率、电压等电气参数的量测值与实际值不同的节点,是配电网态势感知、运行和管理的基础.为此,提出一种基于终端数据挖掘的异常检测方法.首先,通过挖掘节点电压、功率间的潮流约束关系,构建了一种潮流映射框架下的异常检测体系,将不符合约束关系的节点判断为异常.其次,构建对冲反向传播-残差神经网络拟合潮流约束关系,在配电网拓扑和参数信息缺失的情况下实现快速准确的电压计算.再次,针对配电网拓扑变动导致方法失效问题,引入模型回溯概念和生成树算法进行识别,将变动信息反馈给主模型,实现检测方法在动态拓扑下的应用.最后,结合实际案例与仿真实验验证了所提方法的有效性.
Abstract
Abnormal node detection is used to find the nodes where measured value of electrical parameters such as power and voltage are different from actual value resulted from the anomalies like electricity theft and equipment failure.It is the basis of distribution network situational awareness,operation and management.In this regard,an abnormal detection method is proposed based on terminal data mining.Firstly,by mining power flow constraint relationship between node voltage and power,an anomaly detection system under power flow mapping framework is constructed.And nodes that do not conform to the constraint relationship are judged as anomalies.Secondly,hedge backpropagation and residual neural network is constructed to fit the power flow constraint relationship,which can realize fast and accurate voltage calculation when topology and parameter information of distribution network are missing.Thirdly,the concept of model backtracking and spanning tree algorithm are introduced to identify the common topological structure changes in distribution network.The change information is fed back to the main model to realize the application of the detection in dynamic topology.Finally,the validity of the proposed method is verified by using practical cases and simulation experiments.
关键词
异常节点检测/潮流映射/终端数据驱动/配电网拓扑结构Key words
abnormal node detection/power flow mapping/terminal data driven/distribution network topology引用本文复制引用
基金项目
国家自然科学基金青年科学基金资助项目(52307214)
出版年
2024
NETL
NSTL
万方数据