基于深度卷积神经网络的客户侧用电安全性评估模型研究  

Research on Customer-side Power Safety Evaluation Model Based on Deep Convolutional Neural Network

作　　者：王白根 鲍兴江 邵竹星 陆钦 胡中鲲 WANG Baigen;BAO Xingjiang;SHAO Zhuxing;LU Qin;HU Zhongkun(Anqing Electric Power Supply Company of State Grid Anhui Electric Power Company,Anqing 246000,China)

机构地区：[1]国网安徽省电力有限公司安庆供电公司,安徽安庆246000

出　　处：《微型电脑应用》2025年第2期128-131,共4页Microcomputer Applications

摘　　要：考虑到深度卷积神经网络优秀的性能,提出了基于深度卷积神经网络的客户侧用电安全性评估模型。根据客户侧的用电安全实际情况,构建了客户侧用电安全评级指标体系。利用深度卷积神经网络的数据处理性能以及计算性能,构建了基于深度卷积神经网络的客户侧用电安全性评价模型。实验结果表明,所提模型其性能指标表现良好,其F 1值达到了0.962,Recall值达到了0.962,AUC值到达了0.949。在实际的检测中,准确率达到了0.978,相较于深度神经网络、支持向量机模型和逻辑回归模型高出了0.027~0.066。实时预警正确率达到了0.957,实时预警误报率仅为0.037。因此,所提模型在用电安全评估中具备更强的高效性以及实用性。Considering the excellent performance of deep convolutional neural networks,a customer-side power safety evaluation model based on deep convolutional neural networks is proposed.Based on the actual situation of power consumption safety on the customer-side,this paper constructs the power consumption safety rating index system on the customer-side,uses the huge data processing and computing performance of deep convolutional neural network,a customer-side power safety evaluation model based on deep convolutional neural network is constructed.The experimental results show that the performance index of the proposed model is good,and its F1 value reaches 0.962,the Recall value reached 0.962,and the AUC value reached 0,949,the accuracy rate reached 0.978.In the actual detection,the accuracy reaches 0.978,which was 0.027~0.066 higher than that of the deep neural network,support vector machine model and logistic regression model.The correct rate of real-time early warning reached 0.957,and the false alarm rate of real-time early warning was only 0.037.Therefore,the proposed model is more efficient and practical in the power safety evaluation.

关 键 词：深度学习 深度卷积神经网络 粗糙集 评级指标体系 安全评估 用电安全 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]