Multi-physics field joint inversion method and device in geothermal exploration
By using a multiphysics joint inversion method, seismic, electromagnetic, and well logging data are obtained, and a multiphysics joint inversion network is constructed. This solves the problem of excessive reliance on rock physical relationships in traditional methods and enables high-resolution geothermal exploration data support.
Patent Information
- Application Number
- CN202511708086.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-20
- Publication Date
- 2026-05-01
AI Technical Summary
Existing electromagnetic-seismic joint inversion methods rely excessively on assumptions of consistent rock physical relationships or physical properties and structures, making it difficult to meet the needs of detailed exploration of deep geothermal resources.
A multiphysics joint inversion method is adopted to acquire seismic data, electromagnetic data and well logging data. By combining well point constraint loss, seismic constraint loss and electromagnetic constraint loss with gradient consistency loss, a multiphysics joint inversion network is constructed. Using a two-branch architecture of bilinear interpolation, Fourier coding, sinusoidal activation function and delayed injection residual learning, high-resolution resistivity wave impedance joint prediction is achieved.
Accurately constructing a multi-physics joint inversion network overcomes the problem of stringent implementation conditions of traditional methods, provides a data foundation for geothermal exploration, and supports reservoir location and target area selection.
Smart Images

Figure CN121956192A_ABST
Abstract
Description
Multiphysics Joint Inversion Method and Device in Geothermal Exploration Technical Field
[0001] This invention belongs to the field of geothermal exploration and relates to a multi-physics field joint inversion method and apparatus in geothermal exploration. Background Technology
[0002] In the field of geothermal exploration, the inversion of key geothermal parameters is a crucial technology throughout the entire process of geothermal exploration and development, playing a vital supporting role in geothermal resource evaluation and target area selection. Electromagnetic methods can effectively identify the resistivity distribution characteristics of subsurface targets, while seismic methods can accurately reflect the structural and interface features of subsurface geological units at high resolution. The combination of electromagnetic and seismic data is an effective means of achieving refined geothermal exploration.
[0003] Currently, there are two main types of electromagnetic-seismic joint inversion methods. The first is joint inversion based on rock physics models. This type of method constructs a physical model between velocity and resistivity based on laboratory measurements or well logging data, thereby establishing the relationship between electromagnetic and seismic data. However, the quality of this joint inversion depends on stable and reliable rock physics relationships and is difficult to adapt to complex lithological development zones with rapidly changing lithologies. The second type is joint inversion based on structural similarity. This type of method assumes that the spatial variations of the subsurface physical parameters to be inverted are consistent and uses a cross-gradient constraint method to couple the velocity and resistivity relationships. However, when the spatial distributions of velocity and resistivity of geological units are inconsistent, this joint inversion method struggles to adjust the boundaries of the resistivity and velocity models to ensure that the resistivity and velocity models have similar physical property structures.
[0004] It is evident that the above methods all have significant shortcomings, relying excessively on the assumption of good rock physical relationships or consistent physical property parameters in the work area; they are difficult to meet the current needs of fine-grained deep geothermal exploration. Summary of the Invention
[0005] This invention provides a multi-physics joint inversion method for geothermal exploration, which can accurately construct a multi-physics joint inversion network. It overcomes the problems of excessive reliance on assumptions of consistency in rock physical relationships or physical properties and stringent implementation conditions in traditional electromagnetic-seismic joint inversion, thus providing a data foundation for geothermal exploration. The multi-physics joint inversion method includes: acquiring seismic data, electromagnetic data, and well logging data from geothermal exploration; determining the loss function of the multi-physics joint inversion network based on well point constraint loss, seismic constraint loss, electromagnetic constraint loss, and gradient consistency loss of the multi-physics joint inversion network; the multi-physics joint inversion network includes a network input based on bilinear interpolation and Fourier coding, an implicit representation network based on a sinusoidal activation function, and a dual-branch architecture based on delayed injection and residual learning; and constructing the multi-physics joint inversion network based on its loss function, seismic data, electromagnetic data, and well logging data from geothermal exploration.
[0006] This invention provides a multiphysics joint inversion device for geothermal exploration, which can accurately construct a multiphysics joint inversion network. It overcomes the problems of excessive reliance on assumptions of consistency between rock physical relationships or physical properties and the stringent implementation conditions of traditional electromagnetic-seismic joint inversion, thus providing a data foundation for geothermal exploration. The device includes: a data acquisition module for acquiring seismic data, electromagnetic data, and well logging data from geothermal exploration; a loss function determination module for determining the loss function of the multiphysics joint inversion network based on well point constraint loss, seismic constraint loss, electromagnetic constraint loss, and gradient consistency loss of the multiphysics joint inversion network; the multiphysics joint inversion network includes a network input based on bilinear interpolation and Fourier coding, an implicit representation network based on a sinusoidal activation function, and a dual-branch architecture based on delayed injection and residual learning; and an inversion network construction module for constructing the multiphysics joint inversion network based on its loss function, seismic data, electromagnetic data, and well logging data from geothermal exploration.
[0007] This invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the multiphysics joint inversion method in geothermal exploration described above.
[0008] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned multiphysics joint inversion method for geothermal exploration.
[0009] This invention also provides a computer program product, which includes a computer program that, when executed by a processor, implements the multiphysics joint inversion method in geothermal exploration described above.
[0010] In this embodiment of the invention, seismic data, electromagnetic data, and well logging data from geothermal exploration are acquired; based on the well point constraint loss, seismic constraint loss, and electromagnetic constraint loss from geothermal exploration, and the gradient consistency loss of the multiphysics joint inversion network, the loss function of the multiphysics joint inversion network is determined; based on the loss function of the multiphysics joint inversion network, the seismic data, electromagnetic data, and well logging data from geothermal exploration, the multiphysics joint inversion network is constructed; this embodiment of the invention can accurately construct the multiphysics joint inversion network, overcoming the problems of excessive reliance on the assumption of consistency of rock physical relationships or physical property structure and the harsh implementation conditions of traditional electromagnetic-seismic joint inversion, which is beneficial for providing a data foundation for geothermal exploration. Attached Figure Description
[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings: Figure 1 is an example diagram of the multiphysics joint inversion method in geothermal exploration in an embodiment of the present invention; Figure 2 is a specific example diagram of determining the loss function of the multiphysics joint inversion network in an embodiment of the present invention; Figure 3 is a specific example diagram of training the multiphysics joint inversion network in an embodiment of the present invention; Figure 4 is a structural example diagram of the multiphysics joint inversion device in geothermal exploration in an embodiment of the present invention; Figure 5 is a specific example diagram of the structure of the multiphysics joint inversion device in geothermal exploration in an embodiment of the present invention; Figure 6 is a schematic diagram of the computer equipment structure in an embodiment of the present invention. Detailed Implementation
[0012] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0013] Figure 1 is an example diagram of the multiphysics joint inversion method in geothermal exploration according to an embodiment of the present invention. As shown in Figure 1, the multiphysics joint inversion method in geothermal exploration includes: Step 101, acquiring seismic data, electromagnetic data and well logging data in geothermal exploration; Step 102, determining the loss function of the multiphysics joint inversion network based on the well point constraint loss, seismic constraint loss and electromagnetic constraint loss in geothermal exploration, and the gradient consistency loss of the multiphysics joint inversion network; the multiphysics joint inversion network includes a network input based on bilinear interpolation and Fourier coding, an implicit representation network based on sinusoidal activation function, and a dual-branch architecture based on delayed injection and residual learning; Step 103, constructing the multiphysics joint inversion network based on the loss function of the multiphysics joint inversion network, the seismic data, electromagnetic data and well logging data in geothermal exploration.
[0014] In step 101, seismic data, electromagnetic data, and well logging data from geothermal exploration are acquired.
[0015] In a specific embodiment, to meet the requirements of electromagnetic-seismic joint inversion in geothermal exploration scenarios, three types of core data are acquired for multi-physics joint inversion: Seismic data acquisition: Seismic data that reflects the structural and interface characteristics of underground geological units is collected. This data needs to have high resolution to support the subsequent modeling of underground geological structures by the inversion network; Electromagnetic data acquisition: Electromagnetic data that reflects the resistivity distribution characteristics of underground targets is collected, providing a basis for the inversion network to build an underground resistivity model; Well logging data acquisition: Well logging data of the geothermal exploration area is collected. This data can serve as a key basis for well point constraints, used for subsequent construction of electromagnetic-seismic and resistivity wave impedance training and verification samples, and at the same time provides data support for the calculation of well point constraint loss in the loss function of the inversion network, ensuring that the inversion results conform to the actual geological conditions.
[0016] In this embodiment, after acquiring seismic data, electromagnetic data, and well logging data in geothermal exploration, the process may further include: preprocessing the seismic data, electromagnetic data, and well logging data, which may include: removing environmental noise interference signals from the seismic data, correcting electrical response deviations in the electromagnetic data, and supplementing missing physical property parameters at well points in the well logging data.
[0017] Figure 2 is a specific example of determining the loss function of the multiphysics joint inversion network in an embodiment of the present invention. As shown in Figure 2, the loss function of the multiphysics joint inversion network is determined based on the well point constraint loss, seismic constraint loss, and electromagnetic constraint loss in geothermal exploration, and the gradient consistency loss of the multiphysics joint inversion network. This may include: Step 201, determining the weight coefficients of the well point constraint loss, seismic constraint loss, electromagnetic constraint loss, and gradient consistency loss; Step 202, combining the well point constraint loss, seismic constraint loss, electromagnetic constraint loss, and gradient consistency loss with their corresponding weight coefficients to determine the loss function of the multiphysics joint inversion network.
[0018] In this embodiment, the wellpoint constraint loss is determined based on the actual physical property parameters at the wellpoint in the geothermal exploration logging data; the seismic constraint loss is determined based on the seismic response characteristics corresponding to the geothermal exploration seismic data; the electromagnetic constraint loss is determined based on the electrical response characteristics corresponding to the geothermal exploration electromagnetic data; and the gradient consistency loss is determined based on the spatial distribution gradient of physical property parameters in the multi-physics joint inversion network.
[0019] In this embodiment, the loss function of the multiphysics joint inversion network can be determined in the following way:
[0020] in, The loss function of the multiphysics joint inversion network is... For well point constraint loss, For earthquake constraint loss, For electromagnetic confinement loss, For gradient consistency loss, , , , These are the weighting coefficients for wellpoint constraint loss, seismic constraint loss, electromagnetic constraint loss, and gradient consistency loss.
[0021] In a specific embodiment, the weighting coefficients of each loss are first determined: Considering the electromagnetic-seismic joint inversion requirements in geothermal exploration scenarios, and taking into account the goal of the inversion network to output high-resolution multiphysics information with clear structure and rich details, as well as the reliability and inversion contribution of well logging data, seismic data, and electromagnetic data, wellpoint constraint losses are set. The weighting coefficient is Seismic constraint loss The weighting coefficient is Electromagnetic confinement loss The weighting coefficient is Gradient consistency loss The weighting coefficient is Each weighting coefficient is used to balance the impact of different constraint losses on the total loss, ensuring that the inversion results closely match the actual well logging physical parameters, as well as the seismic and electrical response characteristics, while also ensuring a reasonable spatial distribution gradient of physical parameters. Then, a loss function is constructed: based on the physical meaning of each loss, where... Determined based on the actual physical property parameters at the well points in geothermal exploration logging data. Based on the seismic response characteristics corresponding to geothermal exploration seismic data, it is determined that... Determined based on the electrical response characteristics corresponding to geothermal exploration electromagnetic data. Determination of spatial distribution gradient of physical parameters based on multiphysics joint inversion network; , , , Each with corresponding weight coefficients , , , After multiplying and summing, the loss function of the multiphysics joint inversion network is obtained. This loss function is used to guide the inversion network, which includes a dual-branch architecture. It contains an implicit inversion network based on sinusoidal activation and a guided inversion structure based on delayed injection and residual learning. This improves the network's high-frequency representation ability and complex region modeling ability, and finally realizes the inversion of the high-resolution resistivity wave impedance joint prediction model.
[0022] In step 102, the loss function of the multiphysics joint inversion network is determined based on the well point constraint loss, seismic constraint loss, and electromagnetic constraint loss in geothermal exploration, and the gradient consistency loss of the multiphysics joint inversion network. The multiphysics joint inversion network includes a network input based on bilinear interpolation and Fourier coding, an implicit representation network based on sinusoidal activation function, and a dual-branch architecture based on delayed injection and residual learning.
[0023] In a specific embodiment, to address the requirements of electromagnetic-seismic joint inversion in geothermal exploration scenarios, the multiphysics joint inversion network's function and structure are designed as follows: Network input based on bilinear interpolation and Fourier coding: Conventional sinusoidal activation-based implicit representation networks (SIREN) only use coordinates as input, lacking awareness of structural information and tending to learn smoother results, exhibiting spectral bias. To address this issue, a guided feature fusion mechanism based on bilinear interpolation is proposed. This module combines spatial location to sample neighborhood information from the guiding map. Simultaneously, Fourier position encoding is performed before inputting the network, achieving explicit guidance of the structural context and enhancing structural information awareness.
[0024] Implicit Representation Network Based on Sine Activation Function: The sine (sin) activation function possesses inherent periodicity and high-frequency modeling capabilities, making it suitable for solving partial differential equations containing derivative terms, and even more suitable for tasks where the network output must satisfy partial differential equation constraints. The network uses sine (sin) as the activation function to fit complex continuous fields and their derivatives.
[0025] A dual-branch architecture based on delayed injection and residual learning is proposed to avoid overfitting the network to the guiding information. Electromagnetic and seismic data are used as guiding information. The extracted features are concatenated with the location-encoded coordinates to construct a dual-branch network structure of guiding information and coordinates, thus decoupling spatial structure modeling from guiding feature fusion. The inversion network first independently learns the spatial structure information contained in the coordinates, and then delays the introduction of guiding features as residual correction terms to improve the modeling ability for complex regions.
[0026] In this embodiment, the construction of the multiphysics joint inversion network based on the loss function of the multiphysics joint inversion network, seismic data, electromagnetic data, and well logging data from geothermal exploration can include: inputting bilinear interpolation into the network to guide feature fusion based on the seismic data, electromagnetic data, and well logging data from geothermal exploration to obtain a guide map; sampling domain information from the guide map based on the seismic data, electromagnetic data, and well logging data from geothermal exploration, combined with spatial location, and performing Fourier encoding on the domain information to obtain an implicit representation network with a sinusoidal activation function; and using a delayed injection and residual learning architecture based on the implicit representation network with the sinusoidal activation function to obtain the multiphysics joint inversion network.
[0027] In a specific embodiment, to address the electromagnetic-seismic joint inversion requirements in geothermal exploration scenarios, and based on a determined loss function, the network is constructed as follows: Network architecture construction: A dual-branch architecture multi-physics joint inversion network is constructed, using subsurface real physical property benchmarks provided by well logging data, such as actual resistivity, wave impedance values, and continuous physical property profiles at well points; as implicit constraints, ensuring that the architecture's functionality matches actual geological laws, specifically including: combining spatial location with network input based on bilinear interpolation and Fourier coding, sampling neighborhood information from electromagnetic and seismic data guidance maps. Simultaneously, Fourier location coding is performed before inputting into the network to achieve explicit guidance of structural context, enhancing structural information perception.
[0028] Implicit Representation Network Based on Sine Activation Function
[0029] The sine activation function possesses inherent periodicity and high-frequency modeling capabilities, making it suitable for solving partial differential equations containing derivative terms, and even more suitable for tasks where the network output must satisfy partial differential equation constraints. The network uses a sine wave as the activation function to fit complex continuous fields and their derivatives.
[0030] A dual-branch architecture based on delayed injection and residual learning
[0031] Electromagnetic and seismic data are used as guiding information. By concatenating the extracted features with the location-encoded coordinates, a dual-branch network structure of guiding information and coordinates is constructed, decoupling spatial structure modeling from guiding feature fusion. The inversion network first independently learns the spatial structure information contained in the coordinates, and then introduces the guiding features as a delayed residual correction term to improve the modeling ability for complex regions.
[0032] Data and loss function integration: The acquired geothermal exploration seismic and electromagnetic data are used as guiding information inputs for the two branches of the network. The loss function is embedded in the network training process, with the goal of minimizing the loss function. Each constraint term in the loss function constrains the network output to ensure that the network learning process not only fits the real physical property parameters of well logging, but also matches the geological characteristics reflected by the seismic and electromagnetic data. At the same time, it ensures that the spatial distribution gradient of physical property parameters is reasonable. Finally, the construction of the multi-physics joint inversion network is completed, laying the foundation for subsequent training and inversion.
[0033] Figure 3 is a specific example of training a multiphysics joint inversion network in an embodiment of the present invention. As shown in Figure 3, after constructing the multiphysics joint inversion network based on the loss function of the multiphysics joint inversion network, seismic data, electromagnetic data, and well logging data in geothermal exploration, the process may further include: Step 301: Constructing training samples and validation samples for the multiphysics joint inversion network based on the seismic data, electromagnetic data, and well logging data in geothermal exploration; Step 302: Training the multiphysics joint inversion network based on the training samples; Step 303: Determining the prediction accuracy of the trained multiphysics joint inversion network based on the validation samples; Step 304: If the prediction accuracy is higher than the preset accuracy, determining the joint prediction model of resistivity and wave impedance for geothermal exploration based on the trained multiphysics joint inversion network; Step 305: If the prediction accuracy is lower than the preset accuracy, repeating the training and validation process until the prediction accuracy meets the preset accuracy.
[0034] In a specific embodiment, the following operations are performed based on the constructed dual-branch architecture multiphysics joint inversion network and loss function: Constructing training and validation samples: Based on seismic data (reflecting the structural and interface characteristics of underground geological units), electromagnetic data (reflecting the resistivity distribution characteristics of underground targets), and well logging data (including real physical property parameters at well points) obtained from geothermal exploration, training and validation samples for the multiphysics joint inversion network are constructed. Well logging data is used to ensure the authenticity of the physical property parameters in the samples, providing a reliable data foundation for subsequent network training and accuracy verification; Training the multiphysics joint inversion network: The constructed training samples are input into the multiphysics joint inversion network, using the loss function... Minimize the target network training, through (Based on well logging data) (Based on earthquake data) (Based on electromagnetic data) (Based on the spatial distribution gradient of physical property parameters) The network output is constrained respectively, and the spatial information learning and guided feature fusion are decoupled by the dual-branch structure to improve the network's high-frequency expression ability and complex region modeling ability; Determine the network prediction accuracy: Input the validation sample into the trained multi-physics joint inversion network to obtain the resistivity and wave impedance prediction results of the network output. By comparing the prediction results with the real physical property parameters in the validation sample, the prediction accuracy of the network is evaluated and determined; Output the joint prediction model: If the prediction accuracy determined in step 303 is higher than the preset accuracy, the network training is judged to be up to standard. The trained multi-physics joint inversion network is used as the core computing carrier, and the seismic data and electromagnetic data of the geothermal exploration target area are input. The network is then used for dual-branch collaborative computing; This prediction model is the physical property distribution characterization result generated by the network after processing the target area data. It can realize the collaborative characterization of electrical properties and structural properties, providing data support for geothermal reservoir location and target area selection. It can be directly used for geothermal reservoir range delineation, geothermal target area screening and development risk assessment.
[0035] Repeated training and validation: If the prediction accuracy is lower than the preset accuracy, adjust the network parameters, such as the loss weight coefficients or the layer structure parameters of the dual-branch network, return to step 302 to retrain the network using the training samples, and perform the accuracy validation in step 303 again until the network prediction accuracy meets the preset accuracy requirements.
[0036] This invention also provides a multi-physics field joint inversion device for geothermal exploration, as described in the following embodiments. Since the principle behind this device is similar to that of the multi-physics field joint inversion method in geothermal exploration, its implementation can be found in the implementation of the multi-physics field joint inversion method in geothermal exploration; repeated details will not be elaborated further.
[0037] Figure 4 is a structural example diagram of the multiphysics joint inversion device in geothermal exploration according to an embodiment of the present invention. As shown in Figure 4, the device includes: a data acquisition module 401, used to acquire seismic data, electromagnetic data and well logging data in geothermal exploration; a loss function determination module 402, used to determine the loss function of the multiphysics joint inversion network based on the well point constraint loss, seismic constraint loss and electromagnetic constraint loss in geothermal exploration, and the gradient consistency loss of the multiphysics joint inversion network; the multiphysics joint inversion network includes a network input based on bilinear interpolation and Fourier coding, an implicit representation network based on sinusoidal activation function, and a dual-branch architecture based on delayed injection and residual learning; and an inversion network construction module 403, used to construct the multiphysics joint inversion network based on the loss function of the multiphysics joint inversion network, the seismic data, electromagnetic data and well logging data in geothermal exploration.
[0038] In one embodiment, the data acquisition module 401 is specifically used to: preprocess seismic data, electromagnetic data and well logging data, including: removing environmental noise interference signals from seismic data, correcting electrical response deviations in electromagnetic data, and supplementing missing physical property parameters at well points in well logging data.
[0039] In one embodiment, the loss function determination module is specifically used to: determine the weight coefficients of well point constraint loss, seismic constraint loss, electromagnetic constraint loss and gradient consistency loss; and determine the loss function of the multiphysics joint inversion network by combining the well point constraint loss, seismic constraint loss, electromagnetic constraint loss and gradient consistency loss and their corresponding weight coefficients.
[0040] In one embodiment, the wellpoint constraint loss is determined based on the actual physical property parameters at the wellpoint in the geothermal exploration logging data; the seismic constraint loss is determined based on the seismic response characteristics corresponding to the geothermal exploration seismic data; the electromagnetic constraint loss is determined based on the electrical response characteristics corresponding to the geothermal exploration electromagnetic data; and the gradient consistency loss is determined based on the spatial distribution gradient of physical property parameters in the multi-physics joint inversion network.
[0041] In one embodiment, the loss function determination module is specifically used to determine the loss function of the multiphysics joint inversion network in the following manner:
[0042] in, The loss function of the multiphysics joint inversion network is... For well point constraint loss, For earthquake constraint loss, For electromagnetic confinement loss, For gradient consistency loss, , , , These are the weighting coefficients for wellpoint constraint loss, seismic constraint loss, electromagnetic constraint loss, and gradient consistency loss.
[0043] In one embodiment, the inversion network construction module 403 is specifically used for: inputting network guidance feature fusion using bilinear interpolation based on seismic data, electromagnetic data, and well logging data from geothermal exploration to obtain a guidance map; sampling domain information from the guidance map based on the seismic data, electromagnetic data, and well logging data from geothermal exploration, and performing Fourier encoding on the domain information to obtain an implicit representation network of a sinusoidal activation function; and using a delayed injection and residual learning architecture based on the implicit representation network of the sinusoidal activation function to obtain a multiphysics joint inversion network.
[0044] Figure 5 is a specific example of the structure of the multiphysics joint inversion device for geothermal exploration in an embodiment of the present invention. As shown in Figure 5, in one embodiment, the multiphysics joint inversion device for geothermal exploration shown in Figure 4 may further include: a training module 501.
[0045] In one embodiment, the training module 501 is specifically used for: constructing training samples and validation samples for a multiphysics joint inversion network based on seismic data, electromagnetic data, and well logging data from geothermal exploration; training the multiphysics joint inversion network based on the training samples; determining the prediction accuracy of the trained multiphysics joint inversion network based on the validation samples; if the prediction accuracy is higher than a preset accuracy, determining a joint prediction model for resistivity and wave impedance of geothermal exploration based on the trained multiphysics joint inversion network; if the prediction accuracy is lower than a preset accuracy, repeating the training and validation process until the prediction accuracy meets the preset accuracy.
[0046] Based on the aforementioned inventive concept, as shown in Figure 6, the present invention also proposes a computer device 600, including a memory 610, a processor 620, and a computer program 630 stored in the memory 610 and executable on the processor 620. When the processor 620 executes the computer program 630, it implements the aforementioned multiphysics joint inversion method in geothermal exploration.
[0047] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned multiphysics joint inversion method for geothermal exploration.
[0048] This invention also provides a computer program product, which includes a computer program that, when executed by a processor, implements the multiphysics joint inversion method in geothermal exploration described above.
[0049] In this embodiment of the invention, seismic data, electromagnetic data, and well logging data from geothermal exploration are acquired; based on the well point constraint loss, seismic constraint loss, and electromagnetic constraint loss from geothermal exploration, and the gradient consistency loss of the multiphysics joint inversion network, the loss function of the multiphysics joint inversion network is determined; based on the loss function of the multiphysics joint inversion network, the seismic data, electromagnetic data, and well logging data from geothermal exploration, the multiphysics joint inversion network is constructed; this embodiment of the invention can accurately construct the multiphysics joint inversion network, overcoming the problems of excessive reliance on the assumption of consistency of rock physical relationships or physical property structure and the harsh implementation conditions of traditional electromagnetic-seismic joint inversion, which is beneficial for providing a data foundation for geothermal exploration.
[0050] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0051] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in one or more blocks of the flowchart illustrations and / or one or more blocks of the block diagrams.
[0052] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means that implement the functions specified in one or more flowcharts and / or one or more block diagrams.
[0053] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, such that the instructions, which execute on the computer or other programmable apparatus, provide steps for implementing the functions specified in one or more flowcharts and / or one or more block diagrams.
[0054] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A multi-physics field joint inversion method for geothermal exploration, characterized in that, include: Seismic data, electromagnetic data, and well logging data from geothermal exploration are acquired. Based on the wellpoint constraint loss, seismic constraint loss, and electromagnetic constraint loss from geothermal exploration, and the gradient consistency loss of the multiphysics joint inversion network, the loss function of the multiphysics joint inversion network is determined. The multiphysics joint inversion network includes a network input based on bilinear interpolation and Fourier coding, an implicit representation network based on a sinusoidal activation function, and a dual-branch architecture based on delayed injection and residual learning. The multiphysics joint inversion network is constructed based on its loss function, the seismic data, electromagnetic data, and well logging data from geothermal exploration.
2. The method as described in claim 1, characterized in that, Based on the well point constraint loss, seismic constraint loss, and electromagnetic constraint loss in geothermal exploration, and the gradient consistency loss of the multiphysics joint inversion network, the loss function of the multiphysics joint inversion network is determined, including: determining the weight coefficients of the well point constraint loss, seismic constraint loss, electromagnetic constraint loss, and gradient consistency loss; and combining the well point constraint loss, seismic constraint loss, electromagnetic constraint loss, and gradient consistency loss with their corresponding weight coefficients to determine the loss function of the multiphysics joint inversion network.
3. The method as described in claim 2, characterized in that, The wellpoint constraint loss is determined based on the actual physical property parameters at the wellpoint in the geothermal exploration logging data; the seismic constraint loss is determined based on the seismic response characteristics corresponding to the geothermal exploration seismic data; the electromagnetic constraint loss is determined based on the electrical response characteristics corresponding to the geothermal exploration electromagnetic data; and the gradient consistency loss is determined based on the spatial distribution gradient of physical property parameters in the multi-physics joint inversion network.
4. The method as described in claim 2, characterized in that, The loss function of the multiphysics joint inversion network is determined as follows: in, The loss function of the multiphysics joint inversion network is... For well point constraint loss, For earthquake constraint loss, For electromagnetic confinement loss, For gradient consistency loss, 、 、 、 These are the weighting coefficients for wellpoint constraint loss, seismic constraint loss, electromagnetic constraint loss, and gradient consistency loss.
5. The method as described in claim 1, characterized in that, After constructing the multiphysics joint inversion network based on its loss function, seismic data, electromagnetic data, and well logging data from geothermal exploration, the process includes: constructing training and validation samples for the multiphysics joint inversion network based on the seismic, electromagnetic, and well logging data from geothermal exploration; training the multiphysics joint inversion network using the training samples; determining the prediction accuracy of the trained multiphysics joint inversion network using the validation samples; if the prediction accuracy is higher than the preset accuracy, determining the joint prediction model for resistivity and wave impedance of geothermal exploration based on the trained multiphysics joint inversion network; if the prediction accuracy is lower than the preset accuracy, repeating the training and validation process until the prediction accuracy meets the preset accuracy.
6. The method as described in claim 1, characterized in that, After acquiring seismic, electromagnetic, and well logging data in geothermal exploration, the process also includes: preprocessing the seismic, electromagnetic, and well logging data, including: removing environmental noise interference signals from the seismic data, correcting electrical response deviations in the electromagnetic data, and supplementing missing physical property parameters at well points in the well logging data.
7. The method as described in claim 1, characterized in that, Based on the loss function of the multiphysics joint inversion network and seismic, electromagnetic, and well logging data from geothermal exploration, a multiphysics joint inversion network is constructed. This includes: inputting bilinear interpolation into the network using seismic, electromagnetic, and well logging data from geothermal exploration to guide feature fusion, obtaining a guiding map; sampling domain information from the guiding map using the seismic, electromagnetic, and well logging data from geothermal exploration, combined with spatial location, and performing Fourier encoding on the domain information to obtain an implicit representation network with a sinusoidal activation function; and using a delayed injection and residual learning architecture based on the implicit representation network with the sinusoidal activation function to obtain the multiphysics joint inversion network.
8. A multi-physics field joint inversion device for geothermal exploration, characterized in that, include: The data acquisition module is used to acquire seismic data, electromagnetic data, and well logging data in geothermal exploration. The loss function determination module is used to determine the loss function of the multiphysics joint inversion network based on the well point constraint loss, seismic constraint loss, and electromagnetic constraint loss in geothermal exploration, and the gradient consistency loss of the multiphysics joint inversion network. The multiphysics joint inversion network includes a network input based on bilinear interpolation and Fourier coding, an implicit representation network based on sinusoidal activation function, and a two-branch architecture based on delayed injection and residual learning. The inversion network construction module is used to construct the multiphysics joint inversion network based on the loss function of the multiphysics joint inversion network, seismic data, electromagnetic data, and well logging data in geothermal exploration.
9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method according to any one of claims 1-7.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method of any one of claims 1-7.
11. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the method of any one of claims 1-7.
Citation Information
Cited By
Time-lapse magnetotelluric inversion method and system that fuse seismic constraints and physical guidance
CN122239183A
