光伏电站电力听诊系统降噪方法*

在光伏电站电力听诊系统中,为降低光伏电站采集声音的噪声,对光伏电站常见的各种噪声进行时频谱分析,基于各类噪声特点和经典的U型网络结构设计了降噪网络,为提升降噪效果,将能量信息加入到复数卷积网络中,同时融合基于Transformer的专家子系统进一步提升效果。最后,通过构建数据集和多组对比实验证明了所设计网络对常见混叠信号可以大幅提高声音信号的信噪比,最高可达6.14dB。另外,通过多组烧蚀实验证明了文章中所述各项优化点的有效性。

Research on noise reduction method of electric auscultation system in photovoltaic power station

In order to reduce the noise of collected sound of photovoltaic power station in the photovoltaic power plant power listening system, the common noise of photovoltaic power plant is analyzed in time domain and spectrum. Besides,the noise reduction network is proposed based on various noise characteristics and U-net classical network. To enhance the noise reduction effect, energy information is added to the complex convolutional network and fused with the Transformer-based expert subsystem. Finally, the constructed dataset and multiple sets of comparison experiments demonstrate that the designed network can significantly improve the signal-to-noise ratio of the sound signal up to 6.14 dB for the common mixed signals. in addition, the effectiveness of the optimization points described in the paper is demonstrated by multiple sets of ablation experiments.