基于数据挖掘的电量异常数据智能识别方法研究  

Research on Intelligent Identification Method of Power Abnormal Data Based on Data Mining

作　　者：杨婧 石云辉 卢启芳 YANG Jing;SHI Yunhui;LU Qifang(Metrology Center,Guizhou Power Grid Co.,Ltd.,Guiyang 550000,China)

机构地区：[1]贵州电网有限责任公司计量中心,贵州贵阳550000

出　　处：《自动化仪表》2023年第11期64-68,共5页Process Automation Instrumentation

基　　金：贵州电网有限责任公司营销技改基金资助项目(0600002022030102JL00002)。

摘　　要：为了准确识别电量异常数据,设计了基于数据挖掘的电量异常数据智能识别方法。首先,设计电量异常数据智能识别方法的工作流程,构建电量异常数据识别指标体系。然后,预处理数据。采用主成分分析算法对识别指标进行线性组合降维处理,构建异常识别综合指标。最后,基于相关系数矩阵,利用数据挖掘算法构建主成分因子模型,确定指标权重。创新性地使用数据挖掘技术中的模糊C均值算法进行电量异常特征聚类,利用径向基神经网络构建异常识别模型,通过训练获得相似性识别阈值内的最优识别结果,以实现电量异常数据智能识别。仿真测试结果表明,该方法识别准确率高于70%,识别时间平均为402.32 ms。该方法可以准确识别电量异常数据,且识别效率较高,具有较高的实际应用价值。To accurately identify the abnormal power data,an intelligent identification method of abnormal power data based on data mining is designed.Firstly,the workflow of the intelligent identification method of power abnormal data is designed,and the identification index system of power abnormal data is constructed.Then,the data is preprocessed.The principal component analysis algorithm is used to reduce the dimensionality of the identification indexes by linear combination and construct the comprehensive indexes of anomaly identification.Finally,based on the correlation coefficient matrix,the data mining algorithm is used to construct the principal component factor model and determine the indicator weights.Innovatively,the fuzzy C-mean algorithm in data mining technology is used to cluster power abnormal features,and the radial basis neural network is used to construct the abnormal recognition model,and the optimal recognition results within the similarity recognition threshold are obtained through training,to realize the intelligent recognition of power abnormal data.Simulation test results show that the recognition accuracy of the method is higher than 70%,and the average recognition time is 402.32 ms.The method can accurately recognize the power abnormal data,and the recognition efficiency is high,which is of high practical application value.

关 键 词：数据挖掘 电量异常数据识别 主成分分析 模糊C均值算法 径向基神经网络 相似性识别 

分 类 号：TH-39[机械工程]