摘要
针对当前低压配电网剩余电流保护设备只能依靠剩余电流幅值作为保护机构动作依据,无法识别触电特征的问题,提出了基于小波分解降噪和长短期记忆(LSTM)神经网络的低压配电网生命体触电电流检测方法。首先将总剩余电流信号通过小波分解算法进行降噪,然后将降噪后的生命体触电电流波形作为输入,对LSTM神经网络进行训练,建立生命体触电电流检测模型。仿真实验表明:该方法在速度和准确率上与卷积神经网络(CNN)和反向传播(BP)神经网络相比有明显优势,能够满足剩余电流保护装置速动性的要求,并且稳定性好,生命体触电识别准确率高,对新一代的继电保护设备的研究与开发具有一定的参考价值。
Aiming at the problem that the residual current protection equipment in low-voltage distribution networks can only rely on residual current amplitude as the basis for the action of the protection mechanism and cannot identify the characteristics of electrocution,a method for detecting electrocution current of biological bodies based on wavelet decomposition denoising and long and short-term memory(LSTM)neural network is proposed.The total residual current signal is noise-reduced by wavelet decomposition,and the noise-reduced waveform with time-domain and waveform characteristics is used as the input to train the long and short-term memory neural network to establish the biological body electrocution current detection model.Simulation results show that the method has obvious advantages over convolutional neural netwoek(CNN)and back propagation(BP)neural network in terms of speed and can meet the speed requirements of residual current protection device,and stability is good.The accuracy of biological body electrocution recognition is high,which has certain reference value for research and development of new generation relaying protection equipment.
作者
赵启承
虞雁凌
ZHAO Qicheng;YU Yanling(Electric Power Scientific Research Institute,Zhejiang Electric Power Co,Hangzhou 310014,China;Zhejiang Yian Power Electronic Technology Co,Hangzhou 311300,China)
出处
《传感器与微系统》
CSCD
北大核心
2022年第1期142-145,共4页
Transducer and Microsystem Technologies
基金
国网浙江省电力有限公司科技项目(5211DS190034)。
关键词
低压配电网
生命体触电电流
小波分解
长短期记忆神经网络
触电识别
low-voltage distribution network
biological body electrocution current
wavelet decomposition
long and short-term memory(LSTM)neural networks
electrocution identification
