基于多层全连接神经网络的漏电流容性分量补偿方法研究  

Capacitive Component Leakage Current Compensation Method Based on Multilayer Fully Connected Neural Network

作　　者：周星雨 吴桂初 吴自然 李泉坊 ZHOU Xingyu;WU Guichu;WU Ziran;LI Quanfang(School of Electrical and Electronic Engineering,Wenzhou University,Wenzhou 325000,China;Yueqing Industrial Research Institute,Wenzhou University,Wenzhou 325699,China;Zhejiang Juchuang Intelligent Technology Co.,Ltd.,Wenzhou 325000,China)

机构地区：[1]温州大学电气与电子工程学院,浙江温州325000 [2]温州大学乐清工业研究院,浙江乐清325699 [3]浙江聚创智能科技有限公司,浙江温州325000

出　　处：《电器与能效管理技术》2023年第3期54-61,共8页Electrical & Energy Management Technology

基　　金：温州市重大科技项目(ZG2020049);温州市瓯海区会领军人才专项《低压电器智能制造车间的研发与产业化》。

摘　　要：漏电流容性分量普遍存在于配电系统中,对漏电监控产生显著干扰,从而影响了电气火灾监控报警的准确性。搭建漏电流试验平台,模拟实际线路情况,设计数据采集系统实时显示,采集4种电压模式、4种负载模式在阻性漏电和容性漏电情况下的数据。提出利用多层全连接BP神经网络对漏电流阻性分量进行预测,消除漏电流容性分量。经测试,所提模型对漏电流阻性分量预测误差仅1.08%,能有效识别漏电流阻性分量,提高电气火灾预警的准确性。The capacitive component leakage current(CCLC)commonly exists in power distribution systems.However,it significantly interferes the leakage current measurement and reduces the accuracy of electrical fire monitoring.The leakage current test platform is built to simulate the actual circuit condition,and a data acquisition system is designed to display and acquire the data of four voltage modes and four load modes under the resistive leakage and capacitive leakage conditions.A multilayer fully connected BP neural network is used to eliminate the CCLC by predicting the resistive component leakage current(RCLC).The test results show that the prediction error of the RCLC reaches only 1.08%,which can prove that the methd can effectively identify the RCLC and improve the accuracy of electrical fire alarming.

关 键 词：漏电流容性分量 电气火灾监控 多层全连接神经网络 漏电流容性分量补偿 

分 类 号：TM714.3[电气工程—电力系统及自动化]