Deep learning-based near-field effect correction method for Controlled Source Electromagnetic Method and application.

Area of Science:

• Geophysics
• Deep Learning Applications
• Electromagnetic Methods

Background:

• Near-field effects in Controlled Source Electromagnetic (CSEM) methods pose challenges in geophysical exploration.
• Accurate interpretation of CSEM data is crucial for understanding subsurface geological structures.

Purpose of the Study:

• To develop and validate a deep learning-based method for correcting near-field effects in CSEM data.
• To enhance the accuracy of geological structure interpretation by mitigating near-field influences.

Main Methods:

• Generation of diverse datasets using forward simulation for a layered geologic model.
• Construction of a deep learning network combining Long Short-Term Memory (LSTM) and Convolutional Neural Network (CNN) architectures.
• Experimental validation on both simulated and measured CSEM data, including noise resilience tests.

Main Results:

• The LSTM-CNN network demonstrated high accuracy, with trained data closely matching simulated data for theoretical datasets.
• Application to measured CSEM data effectively removed false high-resistance anomalies observed at lower frequencies.
• The method showed significant improvement in mitigating near-field effects and noise.

Conclusions:

• The proposed deep learning-based correction method effectively eliminates the influence of near-field effects in CSEM.
• This technique offers practical benefits for geophysical exploration by providing a more authentic representation of geological structures.