基于voting集成的智能电能表故障多分类方法

Multi-classification method of smart electricity meter fault based on voting integration

为提升智能电能表故障准确分类能力,助力维护人员迅速排除故障,提出基于投票法voting集成的智能电能表故障多分类方法。针对实际智能电能表故障数据进行编码预处理,基于皮尔逊系数法筛选智能电能表故障分类关键影响因素,结合合成少数类过采样技术(synthetic minority oversampling technique, SMOTE)算法解决数据类别不平衡问题,由此建立模型所需数据集,再通过投票法进行模型融合,结合粒子群PSO(particle swarm optimization)确定各基模型的权重,据此构建基于极限梯度提升树(extreme gradient boosting trees, XGBT)、K近邻(k-nearest neighbor, KNN)和朴素贝叶斯(naive bayes, NB)模型的智能电能表故障多分类方法。实测实验结果表明:所提出方法能有效实现智能电能表的故障快速准确分类,与现有方法相比,在智能电能表的故障分类精确率、召回率及F1-Score均有明显提升。

In order to improve the ability to accurately classify faults of smart electricity meters and help maintainers to quickly troubleshoot faults,this paper proposes a multi-classification method for smart electricity meter faults based on voting integration.This paper performs coding preprocessing for the actual fault data of smart electricity meters,screens the key influencing factors of fault classification of smart electricity meters based on the Pearson coefficient method,and combines the SMOTE algorithm to solve the problem of data category imbalance,thereby establishing the data set required for the model,and then,voting method is used for model fusion,combined with particle swarm optimization(PSO)to determine the weight of each base model.On this basis,a multi-classification method of smart electricity meter fault based on the XGBT+KNN+NB model is constructed.The actual test results show that the method proposed in this paper can effectively realize the rapid and accurate classification of the faults of the smart electricity meter.Compared with the existing methods,the fault classification accuracy,the recall rate and the F 1-Score of the smart electricity meter have been significantly improved.