南京理工大学学报（自然科学版）2023，Vol.47Issue(6)：790-796,858,8.DOI:10.14177/j.cnki.32-1397n.2023.47.06.008
基于改进K-Means和DNN算法的电力数据异常检测
Power data anomaly detection based on improved K-Means and DNN algorithm
摘要
Abstract
To obtain high data identification efficiency,a power data anomaly detection technology based on improved K-Means and deep neural network(DNN)algorithm is proposed combining Spark architecture theory.Firstly,the improved K-means clustering algorithm is used to improve the selection of the initial clustering center and the sample weight,and the parallel operation of each clustering subclass is realized,thereby improving the identification efficiency of abnormal data.Secondly,a correction model of power anomaly data is established based on DNN to correct the identified anomaly data,and a feedback correlation layer is added to the training layer of DNN in view of the problems such as the difficulty of controlling the parameters of the existing clustering algorithm and the outlier detection algorithm and the lack of randomness of the algorithm,and the improved particle swarm algorithm is used to deeply optimize the weight space in the network model.Finally,the supervisory control and data acquisition(SCADA)monitoring system data of a provincial dispatching center is used as a sample to carry out study verification,and the simulation results show that the proposed method can efficiently deal with the anomalies of power grid operation big data.The successful repair rate of data maintains at over 91%.The average accuracy of prediction after data repair reaches 95.2%.关键词
电力大数据/异常检测/Spark架构/K-Means算法/深度神经网络Key words
power big data/anomaly detection/Spark framework/K-means algorithm/deep neural network分类
信息技术与安全科学引用本文复制引用
常荣,徐敏..基于改进K-Means和DNN算法的电力数据异常检测[J].南京理工大学学报（自然科学版）,2023,47(6):790-796,858,8.基金项目
南方电网公司科技项目(050400HY42210001) （050400HY42210001）
