<abstract xmlns="http://www.w3.org/1999/xhtml">

<p>The exploration potential of oil and gas is substantial, with considerable untapped resources available for future exploration and development. This paper investigates the application of time-frequency electromagnetic (EM) methods within the domain of computational electromagnetics, enhanced by deep learning techniques, for the profound exploration of oil and gas reservoirs. To identify zones favorable for hydrocarbon presence, it captures electric and magnetic field responses across time and frequency domains. From these responses, it extracts crucial parameters such as resistivity and excitation polarizability. The methodology initiates with a forward simulation to establish the most productive construction parameters for the time-frequency EM approach. Subsequently, data collected are processed and analyzed to qualitatively evaluate each physical parameter. This is followed by quantitative constraints on resistivity and polarizability, employing the adaptive differential evolution algorithm to execute a comprehensive parameter inversion using the horizontal E-field (Ex) data derived from the time-frequency EM method. This inversion facilitates the determination of resistivity and polarizability distributions at various depths, culminating in the generation of corresponding depth cross-sections. The inversion fitting and example analyses yielded that the apparent resistivity profiles of low-resistance backslope formations and backslope formations with high and low resistance interlayers form relative anomaly magnitudes of about 13% and 7.8%, respectively. In the real measurement from the depression to the uplift zone, the thickness of the giant inclined low-resistance conductive layer gradually decreases. The points of the extreme values and minima gradually move from the low-frequency band to the high-frequency band, which is identical to the simulation law. The anomaly amplitude is between [-0.4 Ω·m, 0.6 Ω·m], which can be accurately imaged. This study innovatively combines deep learning technology and electromagnetism, which provides a basis for the next decision on oil and gas exploration and target evaluation preference.</p>
</abstract>

The exploration potential of oil and gas is substantial, with considerable untapped resources available for future exploration and development. This paper investigates the application of time-frequency electromagnetic (EM) methods within the domain of computational electromagnetics, enhanced by deep learning techniques, for the profound exploration of oil and gas reservoirs. To identify zones favorable for hydrocarbon presence, it captures electric and magnetic field responses across time and frequency domains. From these responses, it extracts crucial parameters such as resistivity and excitation polarizability. The methodology initiates with a forward simulation to establish the most productive construction parameters for the time-frequency EM approach. Subsequently, data collected are processed and analyzed to qualitatively evaluate each physical parameter. This is followed by quantitative constraints on resistivity and polarizability, employing the adaptive differential evolution algorithm to execute a comprehensive parameter inversion using the horizontal E-field (Ex) data derived from the time-frequency EM method. This inversion facilitates the determination of resistivity and polarizability distributions at various depths, culminating in the generation of corresponding depth cross-sections. The inversion fitting and example analyses yielded that the apparent resistivity profiles of low-resistance backslope formations and backslope formations with high and low resistance interlayers form relative anomaly magnitudes of about 13% and 7.8%, respectively. In the real measurement from the depression to the uplift zone, the thickness of the giant inclined low-resistance conductive layer gradually decreases. The points of the extreme values and minima gradually move from the low-frequency band to the high-frequency band, which is identical to the simulation law. The anomaly amplitude is between [-0.4 Ω·m, 0.6 Ω·m], which can be accurately imaged. This study innovatively combines deep learning technology and electromagnetism, which provides a basis for the next decision on oil and gas exploration and target evaluation preference.

Deep Learning-based Computational Electromagnetics for Deep Exploration of Oil and Gas Reservoirs

Applied Mathematics and Nonlinear Sciences

This work is licensed under the Creative Commons Attribution 4.0 International License.

The exploration potential of oil and gas is substantial, with considerable untapped resources available for future exploration and development. This paper...

Deep Learning-based Computational Electromagnetics for Deep Exploration of Oil and Gas Reservoirs

Data publikacji: 05 lip 2024

Zakres stron: -

Otrzymano: 05 mar 2024

Przyjęty: 26 maj 2024

DOI: https://doi.org/10.2478/amns-2024-1642

Słowa kluczowe
Time-frequency electromagnetic method, Deep learning, Adaptive differential evolutionary algorithm, Forward and reverse, Deep exploration of oil and gas reservoirs

© 2024 Na Li et al., published by Sciendo

This work is licensed under the Creative Commons Attribution 4.0 International License.

Deep Learning-based Computational Electromagnetics for Deep Exploration of Oil and Gas Reservoirs

Data publikacji: 05 lip 2024

Zakres stron: -

Otrzymano: 05 mar 2024

Przyjęty: 26 maj 2024

DOI: https://doi.org/10.2478/amns-2024-1642

Słowa kluczoweTime-frequency electromagnetic method, Deep learning, Adaptive differential evolutionary algorithm, Forward and reverse, Deep exploration of oil and gas reservoirs

© 2024 Na Li et al., published by Sciendo

This work is licensed under the Creative Commons Attribution 4.0 International License.

Słowa kluczowe
Time-frequency electromagnetic method, Deep learning, Adaptive differential evolutionary algorithm, Forward and reverse, Deep exploration of oil and gas reservoirs