防灾减灾学报

低信噪比水位信号对数据分析、异常提取造成了较大影响，为此，本文提出一种基于经验小波变换(EWT)、K均值聚类(Kmeans)与小波阈值(WT)的算法(简称“EWT-Kmeans-WT”算法)，以实现对低信噪比井水位信号的有效去噪处理。EWT-Kmeans-WT算法能够准确识别并分离水位信号中的有用成分，有效消除噪声干扰，并恢复低信噪比水位信号的正常动态变化特征。对秭归井水位的去噪研究表明，该算法在恢复水位固体潮响应与气压响应方面表现优异，其信噪比(4.667 6)、波形相似系数(0.174 3)与平滑度(0.053 8)均较高，而均方误差(2.42×10^-6)较小，去噪性能显著优于EWT、WT及EMD方法。此外，该算法还能较好地表现水位的全日波、半日波响应动态变化，固体潮的重构效果优于T＿Tide调和分析。因此，EWT-Kmeans-WT算法在低信噪比水位信号的去噪领域具有较好的应用前景。

The water level signal with low signal-to-noise ratio(low SNR) has a large negative impact on data analysis and seismic anomaly information extraction. For this reason,an algorithm based on Empirical Wavelet Transform(EWT),Kmeans clustering(Kmeans),and Wavelet Thresholding(WT) is proposed in this paper in order to achieve effective denoising of water level signal with low SNR. The EWT-Kmeans-WT algorithm can accurately separate and identify the useful signals in the water level signal, effectively eliminate the noise interference in the water level,and restore the normal dynamic signal components of the water level. The denoising study on the water level of wells at Zigui Seismic Station shows that the algorithm can better recover the dynamic changes,such as earth tides and barometric pressure effects,and its SNR(4.667 6),NCC(0.174 3),and RVR(0.053 8) are larger whereas the RMSE(2.42×10^-6) is smaller than that of the EWT,WT,and EMD methods. In addition,the dynamic changes of the full-day and half-day tidal wave response of the water level are well represented,and the reconstruction effect of the solid tide is more in line with the actual situation than that rebuilt by the T＿Tide software, which sheds a light on that the EWT-Kmeans-WT algorithm have good application prospects for denoising water level signals with low SNR.