电力学报2025，Vol.40Issue(1)：33-40,8.DOI:10.13357/j.dlxb.2025.004
基于局部离群因子的配电网低电压异常数据检测
Low Voltage Anomaly Data Detection in Distribution Networks based on Local Outlier Factor
摘要
Abstract
In low-voltage distribution network systems,anomalies in power data frequently occur because of some factors such as equipment failures,network delays,and adverse weather conditions.To enhance the reli-ability and stability of abnormal data processing in low-voltage monitoring of distribution networks,a low-volt-age anomaly detection method based on the local outlier factor(LOF)is proposed.First,the low-voltage intel-ligent monitoring model is introduced.A data preprocessing model is established based on the residual U-Net(RU-Net)architecture+normalization method.To address the issue of missing data,the model employs a mask description method to illustrate data absence scenarios,utilizes inverse neural networks to adjust the loss function,enhances the reconstruction accuracy of data points,and achieves missing data imputation.Subse-quently,an anomaly detection method based on the LOF algorithm is applied to identify,detect,and eliminate outliers.In the simulation analysis,an anomaly detection model is constructed using Python,and model valida-tion is performed.The results demonstrate that,the proposed model can effectively screen out anomalous points.Finally,taking the 10 kV bus voltage monitoring data of a certain power supply company as an exam-ple,30%of the data is extracted for missing data testing,and the effectiveness of this anomaly detection meth-od is verified through simulation.The results show that the model can achieve data imputation and generate out-put graphs.When K=10,the model accurately identifies anomalous data points while maintaining stability and produces processed anomaly detection plots.The simulation experiments validate the effectiveness of the pro-posed method,providing technical support for improving data quality in distribution network monitoring.关键词
配电网/大数据挖掘/低电压监测/数据清洗/局部离群因子Key words
distribution network/big data mining/low voltage detection/data cleaning/local outlier factor(LOF)分类
信息技术与安全科学引用本文复制引用
李强,范李平,杜丰夷,黄宇,刘云飞,漆曾,张作轩,马辉..基于局部离群因子的配电网低电压异常数据检测[J].电力学报,2025,40(1):33-40,8.基金项目
国家自然科学基金项目(52377191) （52377191）
国网科技项目(GDJS2400846). （GDJS2400846）
