Denoising of Electromagnetic Data from Different Geological Blocks Using a Hybrid PSO-GWO Algorithm and CNN

Liu Zhong yuanDiquan Li^*Liu ye chengZhang xian

• Central South University, Changsha, China

The wide-field electromagnetic (WFEM) method, known for its strengths in deep detection, interference resistance, and adaptability to various surface conditions, has been widely used in geophysical exploration for resources such as shale gas and groundwater in complex geological environments. However, noise interference remains a significant challenge, especially in areas with intense human activity, such as mining regions and urban environments, where electromagnetic noise is complex, non-periodic, and pseudo-random. This interference severely degrades the signal-to-noise ratio (SNR), complicating the interpretation of subsurface structures.In shale gas exploration in the Sichuan Changning Block and groundwater exploration in the Shanxi Xinyuan Coal Mine area, diverse and challenging noise sources further complicate WFEM applications. Traditional denoising methods often struggle to effectively address such complex and non-periodic noise. To solve this problem, we propose an innovative denoising method for WFEM data that integrates Particle Swarm Optimization (PSO) and Grey Wolf Optimization (GWO) to optimize a Convolutional Neural Network (CNN). This method leverages CNN's ability to extract nonlinear features, enabling it to effectively identify and separate multiple noise types (e.g., Gaussian white noise, impulse noise, and attenuation noise), thereby enhancing denoising accuracy. The PSO-GWO hybrid algorithm optimizes CNN hyperparameters, improving the model's adaptability to different noise environments. A sample library was constructed, covering shale gas and groundwater exploration areas, with typical signal features extracted for noise identification and removal using the optimized CNN model. Simulation results demonstrate that this method significantly outperforms traditional techniques in terms of SNR enhancement and its ability to handle complex, non-periodic noise. The proposed denoising approach has been successfully applied in the Sichuan Changning Block and Shanxi Xinyuan Coal Mine area, showing significant improvements in WFEM data quality and the reliability of geological interpretations. This method offers a robust solution to noise challenges in geophysical exploration, enhancing interpretative accuracy across diverse geological blocks.