基于VMD的大地电磁信号去噪研究

Research on denoising of magnetotelluric signal based on VMD

大地电磁信号是解释地质构造的重要信息载体,其受长周期和随机噪声影响严重,导致地质构造的反演结果出现严重的偏差。为了解决该问题,基于变分模态分解(Variational Mode Decomposition,VMD)提出了一种综合性的大地电磁信号去噪算法。对原始电磁信号进行多分辨VMD处理去除长周期噪声,采用小波包阈值去噪法去除信号的随机噪声,使用信号重构得到去噪处理后的大地电磁信号。使用此方法对工程实测大地电磁信号进行处理,结果表明,此方法能够对大地电磁信号的长周期噪声和随机噪声进行抑制,并且极大限度地保存了信号的有效分量,提高了时域信号的周期性,全频分段的视电阻率曲线得到了明显优化。

The magnetotelluric signal is an important information carrier to explain geological structure.It is seriously affected by long period and random noise,which leads to serious deviation of the inversion result of geological structure.In order to solve this problem,a comprehensive denoising algorithm for MT signals was proposed based on Variational Mode Decomposition(VMD).Multi-resolution VMD processing was performed on the original electromagnetic signal to remove the long-period noise.The wavelet packet threshold denoising method was used to remove the random noise of the signal.The signal reconstruction was used to obtain the denoised electromagnetic signal.It was used this method to process the measured earth electromagnetic signals of the project.The results showed that this method could suppress the long-period noise and random noise of the magnetotelluric signal and greatly save the effective component of the signal,improve the periodicity of the time-domain signal and the full-frequency segmented apparent resistivity curve obviously optimized.