基于电压波形的开关电源故障实时监测方法  

Real-Time Monitoring Method of Switching Power Supply Fault Based on Voltage Waveform

作　　者：黄福全 晋龙兴 卢正飞 孔德深 张安龙 段俊欢 叶欣 HUANG Fuquan;JIN Longxing;LU Zhengfei;KONG Deshen;ZHANG Anlong;DUAN Junhuan;YE Xin(Shenzhen Power Supply Bureau,Shenzhen Guangdong 518000,China;CYG SUNRI Co.,Ltd.,Shenzhen Guangdong 518057,China)

机构地区：[1]深圳供电局有限公司,广东深圳518000 [2]长园深瑞继保自动化有限公司,广东深圳518057

出　　处：《电子器件》2025年第1期148-154,共7页Chinese Journal of Electron Devices

基　　金：深圳供电局有限责任公司科技项目(090000KK52210169)。

摘　　要：开关电源在电网二次设备中广泛应用,开关电源故障将导致保护装置拒动,存在大面积停电的风险。针对电源故障,研究基于电压波形图像特征的开关电源故障预警方法,实现了主站实时在线监测各厂站二次设备开关电源故障预警结果。该方法选择原始电压数据作为初始特征,解决特征数据难以获取的问题;使用深度学习卷积神经网络模型,加强系统的故障预警能力;融合电压波形图像分析以及电压波形偏离度计算两种算法,避免对开关电源故障的漏检。实验表明,这一方式相比传统预警方法,不仅可以实现实时在线预警,而且可以提前发现轻微退化的开关电源。The switching power supply is widely used in the secondary equipment of the power grid.The failure of the switching power supply will cause the protection device to refuse to operate,and there is a risk of large-scale power failure.Targeting at power failure,the switching power failure warning method based on voltage waveform image features is studied,and the real-time online monitoring of switching power failure warning results of secondary equipment in each power station is realized.The original voltage data are selected as the initial feature to solve the problem that the feature data are difficult to obtain.The deep learning convolution neural network mod-el is used to strengthen the fault early warning capability of the system.The two algorithms of voltage waveform image analysis and volt-age waveform deviation calculation are integrated to avoid missing detection of switching power supply fault.The experiment shows that the proposed method can not only realize real-time online warning,but also detect the slightly degraded switching power supply in ad-vance compared with traditional warning methods.

关 键 词：开关电源 故障预警 图像特征 深度学习 电压波形 

分 类 号：TM910.7[电气工程—电力电子与电力传动] TP183[自动化与计算机技术—控制理论与控制工程]