阵列式观测大地电磁数据高精度去噪方法研究

Magnetotelluric (MT) is a passive geophysical detection method, and has great advantages for deep exploration. But Natural fluctuation electromagnetic(EM) signal is quite weak and has a broad frequency band, serious noise especially in areas of intense human activity results in failure of inversion and interpretation for some bands or even the whole frequency band of field data, research on MT de-noising method is important for both expanding the area for applying MT method and strengthening the scientific understanding of geological knowledge of places with high disturbance. For this purpose, the project adopts principal component analysis (PCA) to mine the inherent information within array MT data and furthermore establishes relationship model of array MT signals and synthesis signal with reference data to study the factors affecting the accuracy of the synthetic signal. Subsequently, using wavelet to analysis noise, multi-parameter including variance ratio, correlation coefficient and short-term energy ratio et al. with threshold attained by training the identified noise are employed to recognize noise. Noisy data are replaced by synthetic signal to achieve the goal of removing noise with high-precision. Especially, for long period magnetotelluric sounding(LMT), a de-noising method is developed based on geomagnetic data. Conclusively, this study provides new ideas and basic research data for extracting MT signals from area with strong noise in high-precision

大地电磁（MT）作为一种天然场源的探测方法，应用于深部探测具有明显优势。但由于大地电磁采集的是天然电磁场，信号振幅十分微弱、频带宽，易受人文电磁噪声干扰，导致某些频段甚至全频段数据无法参与反演解释工作。项目针对强干扰地区难以获得高信噪比大地电磁资料的技术瓶颈，研究基于阵列式观测大地电磁数据的高精度去噪方法，以多点多参考道数据为基准，利用小波变换分析噪声特征，结合方差比、相关系数、短时能量比等多种参数精确识别噪声；通过主成分分析方法挖掘阵列式天然电磁场的内在信息，利用人工神经网络建立天然电磁场信号内在关系模型；再结合参考数据合成信号，替换噪声数据得到重构信号，实现高精度去除大地电磁噪声的目的；特别针对长周期大地电磁数据，利用地磁台网构成磁场的阵列式观测数据体；为高精度去除大地电磁噪声提供新的思路和基础研究资料。

项目利用多点同步观测的电场和磁场数据，进行多点或阵列式处理，可以有效压制大地电磁人文噪声，具体包括：（1）利用参考点与本地测点原始时间序列的方差比，可以快速有效“定位”噪声；（2）利用同步测点之间电磁场的频率域站间传递函数和时间域单位脉冲响应函数，结合干净的参考点数据，可以高精度合成本地测点电场与磁场信号，用合成数据替换噪声段数据，实现对大地电磁数据的高精度去噪；（3）用地磁台站数据作为远参考，可以提高长周期大地电磁数据的信噪比；（4）对全国地磁与地电数据进行了实际处理，获得了周期超过十万秒的大地电磁阻抗，验证了去噪方法的效果和实用性。