摘要
提出了一种基于深度学习的电力线信道传输特性识别方法,通过人工智能方法完成对电力线信道传输特性的识别。传统的信道传输特性识别一般采用信道估计方法,该方法在噪声较大时估计效果不理想。所提方法采用去噪自编码器能有效对噪声进行抑制,可以在噪声较强的环境下实现信道传输特性的正确识别。在实际应用中,针对自编码器神经网络去噪后数据存在背景效应的问题,提出使用颜色调整方法进一步滤除干扰,提高了对去噪样本的识别成功率。基于深度学习框架建立了仿真模型,分析了9类基准传输样本的信道传输特性识别效果。仿真结果表明,该方法能够在不同的神经网络模型中以及多种噪声条件下,完成对电力线信道传输特性的识别,对进一步完善电力线通信质量保障具有重要意义。
A transmission characteristic recognition method of power line channel based on deep learning is proposed, which completes recognition of power line channel transmission characteristics with artificial intelligence method. Traditionally, channel estimation method is usually used to identify the transmission characteristics of channels. The estimation effect of this method is not ideal when noise is high. In the method proposed in this paper, de-noising self-encoder can suppress the noise effectively, and correct identification of channel transmission characteristics can be achieved in noisy environment. In practical application, aiming at the background effect of data after de-noising with self-encoder neural network, a color adjustment method is proposed to further filter interference, and the success rate of de-noising sample recognition is improved. Based on deep learning framework, a simulation model is established, and the recognition effect of channel transmission characteristics of nine kinds of benchmark transmission samples is analyzed. Simulation results show that the method can recognize the transmission characteristics of power line channel in different neural network models and under various noise conditions. It is of great significance to further improve quality assurance of power line communication.
作者
史建超
胡正伟
贺冬梅
谢志远
SHI Jianchao;HU Zhengwei;HE Dongmei;XIE Zhiyuan(School of Electrical&Elcctronic Engineering,North China Electric Power University,Baoding 071003,Hebei Province,China)
出处
《电网技术》
EI
CSCD
北大核心
2019年第12期4283-4290,共8页
Power System Technology
基金
河北省自然科学基金项目(E2019502186)
河北省科技计划项目(17211704D)~~
关键词
电力线通信
泛在电力物联网
深度学习
传输特性识别
人工智能
卷积神经网络
power line communication
ubiquitous power internet of things
deep learning
recognition of transmission characteristics
artificial intelligence
convolutional neural network