Controllable source electromagnetic apparent resistivity curve correction method and system based on deep learning

Abstract

The present invention discloses a method and system for correcting a controlled source electromagnetic apparent resistivity curve based on deep learning. The method comprises: generating N apparent resistivity curves using wave impedance recursion theory, recorded as noise-free apparent resistivity curves; adding random noise to the noise-free apparent resistivity curves according to the electromagnetic noise influence law to obtain noise-containing apparent resistivity curves; constructing a ResNet-Transformer network architecture, using the noise-containing and corresponding noise-free apparent resistivity curves as training input and output, respectively, to obtain an apparent resistivity curve correction model through training; inputting the noise-affected controlled source electromagnetic apparent resistivity curves collected in real time into the apparent resistivity curve correction model to correct the input apparent resistivity curves and obtain a smooth apparent resistivity curve. The present invention can effectively correct noise-affected controlled source electromagnetic apparent resistivity curves, providing reliable data support for subsequent inversion interpretation.

Description

Technical Field

[0001] The present invention belongs to the technical field of controlled source electromagnetic data processing, and specifically relates to a controlled source electromagnetic apparent resistivity curve correction method and system based on deep learning. Background Art

[0002] Controlled-source electromagnetic (CSEM), a key geophysical exploration tool, plays an irreplaceable role in deep resource exploration. However, increasing interference from surface electromagnetic noise can lead to disordered and distorted apparent resistivity curves, making it difficult to accurately reflect the deep subsurface electrical structure.

[0003] Traditional apparent resistivity curve correction methods, such as polynomial fitting, rely on pre-set model assumptions and are difficult to capture sudden changes in the apparent resistivity curve caused by non-stationary noise. Furthermore, filtering methods, which require manual selection of thresholds or frequency bands, have limited ability to distinguish between noise and signals that overlap in the frequency domain. Summary of the Invention

[0004] In response to technical problems such as traditional apparent resistivity curve correction methods only pursuing curve smoothness, which may violate the physical constraints of underground electrical structures, the present invention provides a controllable source electromagnetic apparent resistivity curve correction method and system based on deep learning, which can achieve high-precision correction of apparent resistivity curves under noise interference.

[0005] In order to achieve the above technical objectives, the present invention adopts the following technical solutions:

[0006] A controlled source electromagnetic apparent resistivity curve correction method based on deep learning, comprising:

[0007] S1, generates N apparent resistivity curves through wave impedance recursion theory, which are recorded as the apparent resistivity curves without noise;

[0008] S2, adding random noise to the apparent resistivity curve without noise according to the influence law of electromagnetic noise to obtain the apparent resistivity curve with noise;

[0009] S3, building a ResNet-Transformer network architecture, taking the apparent resistivity curve containing noise as the training input and the corresponding apparent resistivity curve without noise as the training output, and training to obtain the apparent resistivity curve correction model;

[0010] S4, inputting the controlled source electromagnetic apparent resistivity curve affected by noise, which is collected in real time, into the apparent resistivity curve correction model, correcting the input apparent resistivity curve through the model, and outputting a smooth apparent resistivity curve.

[0011] Furthermore, various apparent resistivity curves are generated through wave impedance recursion theory, including:

[0012] S1.1. Establish a multi-layer geoelectric structure model and define the range of resistivity and thickness of each layer:

[0013] ;

[0014] Where k represents the layer index of the multi-layer geoelectric structure, and represent the resistivity and thickness of the kth layer of geoelectric structure respectively; and represent the minimum and maximum resistivity of the kth layer, and Represent the minimum and maximum thickness of the kth layer respectively;

[0015] S1.2, for each layer of geoelectric structure, within the defined range of resistivity and thickness, randomly sample the resistivity and thickness, and introduce resistivity flipping mechanisms with different probabilities to simulate various geological anomalies;

[0016] S1.3, based on the electromagnetic field diffusion theory, the equivalent wave impedance of the surface of each layer of the geoelectric structure is recursively deduced from the base layer upwards:

[0017] ;

[0018] Where, represents the equivalent wave impedance of the surface of the k-th geoelectric structure, It represents the equivalent wave impedance of the surface of the k+1th layer of geoelectric structure, represents the eigenwave impedance of the k-th layer geoelectric structure, represents the propagation constant of the k-th layer geoelectric structure, represents the thickness of the kth layer of geoelectric structure; the eigenwave impedance and propagation constant of each layer of geoelectric structure are related to the resistivity and emission frequency of the geoelectric structure of that layer;

[0019] S1.4, generate the apparent resistivity of the multi-layer geoelectric structure based on the equivalent wave impedance of the first layer of geoelectric structure surface :

[0020] ;

[0021] Where, is the angular frequency, is the magnetic permeability, is the emission frequency of the controlled source electromagnetic method, and ;

[0022] S1.5, based on the apparent resistivity obtained at different emission frequencies using the controlled source electromagnetic method, a line is constructed based on the emission frequency. Apparent resistivity curve with is the independent variable.

[0023] Furthermore, the eigenwave impedance and propagation constant of each layer of geoelectric structure are calculated as follows:

[0024] ;

[0025] Where i is the imaginary unit.

[0026] Furthermore, each apparent resistivity curve generated by S1 through the wave impedance recursion theory is filtered and then used as the data source of S2 and / or S3.

[0027] Furthermore, the adding of random noise to the noise-free apparent resistivity curve according to the electromagnetic noise influence law includes:

[0028] S2.1, let any i-th noise-free apparent resistivity curve be Expressed as:

[0029] ;

[0030] Where, Represents the apparent resistivity curve The apparent resistivity value in They represent the controlled source electromagnetic method Transmitting frequency;

[0031] S2.2, add random noise generated as follows :

[0032] ;

[0033] Where A represents the intensity of random noise, Indicates the generation of [-1,1] A sequence of random values;

[0034] S2.3, Add to The i-th apparent resistivity curve containing noise in the logarithmic domain is obtained :

[0035] ;

[0036] S2.4, Final Calculation , which is the i-th noisy apparent resistivity curve containing noise in the final conventional number domain.

[0037] Furthermore, the constructed ResNet-Transformer network architecture includes, from input to output, the following: ResNet network layer, Transformer network layer, and regression layer;

[0038] The ResNet network layer first performs one-dimensional convolution, one-dimensional pooling, and ReLU activation on the input noisy apparent resistivity sequence; then continues to perform one-dimensional convolution, one-dimensional pooling, and ReLU activation on the output of the first ReLU activation; then performs one-dimensional convolution on the output of the first ReLU activation, and then adds it to the output of the second ReLU activation as the output of the ResNet network layer;

[0039] The Transformer network layer first performs position encoding on the output of the ResNet network layer, then processes it using a self-attention mechanism, and then outputs it after passing through a dropout layer and a fully connected layer;

[0040] The regression layer regresses the output of the Transformer network layer to obtain a noise-free apparent resistivity sequence.

[0041] Furthermore, when training the apparent resistivity curve correction model in S3, the apparent resistivity curve to be input into and output from the model is first converted into the logarithmic domain and then used as the training input and output of the model;

[0042] When S4 uses the trained apparent resistivity curve correction model for correction, the apparent resistivity value of the noise-affected controlled source electromagnetic apparent resistivity curve acquired in real time is converted to a logarithmic domain value and then input into the apparent resistivity curve correction model; and the output data of the apparent resistivity curve correction model is converted from the logarithmic domain to the conventional domain, which is the final smoothed apparent resistivity curve.

[0043] A system based on the above-mentioned controlled source electromagnetic apparent resistivity curve correction method comprises:

[0044] The smoothed apparent resistivity curve generation module is used to: generate N apparent resistivity curves by wave impedance recursion theory, which are recorded as noise-free apparent resistivity curves;

[0045] The noisy apparent resistivity curve generation module is used to add random noise to the noise-free apparent resistivity curve according to the influence law of electromagnetic noise to obtain the noise-containing apparent resistivity curve;

[0046] The apparent resistivity curve correction model training module is used to: build a ResNet-Transformer network architecture, use the noisy apparent resistivity curve as training input, and the corresponding noise-free apparent resistivity curve as training output, to train and obtain the apparent resistivity curve correction model;

[0047] The apparent resistivity curve correction model is used to correct the real-time input controlled source electromagnetic apparent resistivity curve affected by noise and output a smooth apparent resistivity curve.

[0048] Compared with the prior art, the present invention has the following technical effects:

[0049] The present invention provides a deep learning-based controlled source electromagnetic apparent resistivity curve correction method. On the one hand, ResNet extracts local features of the input data (such as noise distortion, curve mutations, etc.) through the convolutional layer in the residual block, and retains input details through jump connections to avoid the gradient vanishing problem, thereby effectively capturing local distortion features under noise interference. On the other hand, the self-attention mechanism of Transformer can model the global association between any two points in the sequence, ensuring the overall smoothness and physical consistency of the corrected curve. Moreover, ResNet-Transformer combines the deep learning capabilities of ResNet with the advantages of the self-attention mechanism of Transformer, which can achieve coordinated optimization of local noise suppression and overall curve trends. Thirdly, the wave impedance recursion theory generates training data that conforms to geological laws, increasing the reliability of apparent resistivity curve correction training.

[0050] This paper uses the ResNet-Transformer framework in deep learning to learn the nonlinear mapping relationship between noisy and smooth apparent resistivity curves in controlled source electromagnetic methods, and then applies it to the correction of measured apparent resistivity curves. Compared with traditional correction methods, the correction reliability and efficiency of the present method are higher. Compared with traditional deep learning networks such as CNN or LSTM networks, the ResNet-Transformer framework constructed in this paper is more suitable for the task of controlling source electromagnetic apparent resistivity correction, and the improvement effect is better. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a flow chart of the controlled source electromagnetic apparent resistivity curve correction method described in an embodiment of the present invention.

[0052] Figure 2 This is a comparison diagram of the apparent resistivity curves before and after adding random noise in an embodiment of the present invention.

[0053] Figure 3 This is a graph of apparent resistivity generated by the wave impedance recursion theory according to an embodiment of the present invention, where (a) to (i) represent different apparent resistivity curves.

[0054] Figure 4 This is a diagram of the ResNet-Transformer network structure of an embodiment of the present invention.

[0055] Figure 5 This is a ResNet-Transformer training curve diagram of an embodiment of the present invention, where (a) is the root mean square error RMSE and (b) is the loss.

[0056] Figure 6 This is a comparison diagram of the controlled source electromagnetic apparent resistivity curve measured before and after correction according to an embodiment of the present invention, where (a) to (i) represent different apparent resistivity curves, respectively. DETAILED DESCRIPTION

[0057] The following is a detailed description of an embodiment of the present invention. This embodiment is based on the technical solution of the present invention, provides a detailed implementation method and a specific operation process, and further explains the technical solution of the present invention.

[0058] Example 1

[0059] This embodiment provides a controllable source electromagnetic apparent resistivity curve correction method based on deep learning, referring to Figure 1 As shown, the following steps are included.

[0060] S1, generate N apparent resistivity curves through wave impedance recursion theory, which are recorded as noise-free apparent resistivity curves.

[0061] S1.1. Establish a multi-layer geoelectric structure model and define the range of resistivity and thickness of each layer:

[0062] ;

[0063] Where k represents the layer index of the multi-layer geoelectric structure, and represent the resistivity and thickness of the kth layer of geoelectric structure respectively; and represent the minimum and maximum resistivity of the kth layer, and represent the minimum and maximum thickness of the kth layer respectively.

[0064] S1.2, for each layer of geoelectric structure, randomly sample the resistivity and thickness within the defined range of resistivity and layer thickness, and introduce resistivity flipping mechanisms with different probabilities to simulate various geological anomalies.

[0065] S1.3, based on the electromagnetic field diffusion theory, the equivalent wave impedance of the surface of each layer of the geoelectric structure is recursively deduced from the base layer upwards:

[0066] ;

[0067] Where, represents the equivalent wave impedance of the surface of the k-th geoelectric structure, It represents the equivalent wave impedance of the surface of the k+1th layer of geoelectric structure, represents the eigenwave impedance of the k-th geoelectric structure, represents the propagation constant of the k-th layer geoelectric structure, Represents the thickness of the kth layer of geoelectric structure. Among them, the eigenwave impedance and propagation constant of each layer of geoelectric structure are related to the resistivity and transmission frequency of the geoelectric structure of this layer, and the calculation formula is:

[0068] ;

[0069] Where i is the imaginary unit, is the angular frequency, is the magnetic permeability, is the emission frequency of the controlled source electromagnetic method, and .

[0070] S1.4, generate the apparent resistivity of the multi-layer geoelectric structure based on the equivalent wave impedance of the first layer of geoelectric structure surface :

[0071] ;

[0072] S1.5, based on the apparent resistivity obtained at different emission frequencies using the controlled source electromagnetic method, a line is constructed based on the emission frequency. Apparent resistivity curve with is the independent variable.

[0073] In a more preferred embodiment, each apparent resistivity curve generated by the wave impedance recursion theory is converted to the logarithmic domain and filtered, and then used as the data source for the subsequent steps S2 and S3. Figure 2 The figure shows the apparent resistivity curves generated by the wave impedance recursion theory in this embodiment, including 50 apparent resistivity curves in total.

[0074] S2, according to the influence law of electromagnetic noise, random noise is added to the apparent resistivity curve without noise to obtain the apparent resistivity curve with noise.

[0075] S2.1, let any i-th noise-free apparent resistivity curve be Expressed as:

[0076] ;

[0077] Where, Represents the apparent resistivity curve The apparent resistivity value in They represent the controlled source electromagnetic method Transmitting frequency; in this embodiment, .

[0078] S2.2, add random noise generated as follows :

[0079] ;

[0080] Where A represents the intensity of random noise, Indicates the generation of [-1,1] A sequence of random values;

[0081] S2.3, Add to The i-th apparent resistivity curve containing noise in the logarithmic domain is obtained :

[0082] ;

[0083] S2.4, Final Calculation , which is the i-th noisy apparent resistivity curve containing noise in the final conventional number domain.

[0084] like Figure 3 Shown is a comparison of the apparent resistivity curves before and after adding random noise in this embodiment.

[0085] S3, build a ResNet-Transformer network architecture, use the noisy apparent resistivity curve as the training input, and the corresponding noise-free apparent resistivity curve as the training output, and train to obtain the apparent resistivity curve correction model.

[0086] The ResNet-Transformer network architecture constructed in the embodiment of the present invention is as follows: Figure 4 As shown, from input to output, it includes: ResNet network layer, Transformer network layer and regression layer;

[0087] The ResNet network layer first performs one-dimensional convolution, one-dimensional pooling, and ReLU activation on the input noisy apparent resistivity sequence; then continues to perform one-dimensional convolution, one-dimensional pooling, and ReLU activation on the output of the first ReLU activation; then performs one-dimensional convolution on the output of the first ReLU activation, and then adds it to the output of the second ReLU activation as the output of the ResNet network layer;

[0088] The Transformer network layer first performs position encoding on the output of the ResNet network layer, then processes it using a self-attention mechanism, and then outputs it after passing through a dropout layer and a fully connected layer;

[0089] The regression layer regresses the output of the Transformer network layer to obtain a noise-free apparent resistivity sequence.

[0090] In this embodiment, 4 / 5 of the training data is used as a training set to train the apparent resistivity curve correction model, and the other 1 / 5 of the training data is used as a validation set to validate the trained apparent resistivity curve correction model.

[0091] In a more preferred embodiment, when training the apparent resistivity curve correction model in step S3, the apparent resistivity curve to be input into and output from the model is first converted into a logarithmic domain and then used as the training input and output of the model.

[0092] Figure 5 Shown is a ResNet-Transformer training curve diagram of an embodiment of the present invention.

[0093] S4, inputting the controlled source electromagnetic apparent resistivity curve affected by noise, which is collected in real time, into the apparent resistivity curve correction model, correcting the input apparent resistivity curve through the model, and outputting a smooth apparent resistivity curve.

[0094] In a more preferred embodiment, when the trained apparent resistivity curve correction model is used for correction in step S4, the apparent resistivity value of the noise-affected controlled source electromagnetic apparent resistivity curve acquired in real time is converted to a logarithmic domain value and then input into the apparent resistivity curve correction model; and the output data of the apparent resistivity curve correction model is converted from the logarithmic domain to the conventional domain, i.e., the final smoothed apparent resistivity curve.

[0095] S4.1: Assume that the multiple collected controlled source electromagnetic apparent resistivity curves containing noise are:

[0096] ;

[0097] Where, represents the nth apparent resistivity curve, represents the apparent resistivity value in the nth apparent resistivity curve, Indicates the main frequency value of the controlled source electromagnetic method.

[0098] S4.2: Convert the controlled source electromagnetic apparent resistivity curve containing noise as follows:

[0099] ;

[0100] S4.3: Input the above-mentioned converted noisy apparent resistivity curves into the trained apparent resistivity curve correction model in sequence, and the model outputs the corrected apparent resistivity curve .

[0101] Step 4.4: Restore the model-corrected apparent resistivity curve to , and obtain the final controlled source electromagnetic apparent resistivity correction curve.

[0102] like Figure 6The figure shows a comparison of the measured controlled source electromagnetic apparent resistivity curves before and after correction. After correction by the method of the embodiment of the present invention, the apparent resistivity curves tend to be smooth, indicating that the method of the present invention can effectively improve the disorder of the apparent resistivity curve caused by noise, and significantly improve the data quality.

[0103] Example 2

[0104] This embodiment provides a system based on the controlled source electromagnetic apparent resistivity curve correction method described in Example 1, including:

[0105] The smoothed apparent resistivity curve generation module is used to: generate N apparent resistivity curves by wave impedance recursion theory, which are recorded as noise-free apparent resistivity curves;

[0106] The noisy apparent resistivity curve generation module is used to add random noise to the noise-free apparent resistivity curve according to the influence law of electromagnetic noise to obtain the noise-containing apparent resistivity curve;

[0107] The apparent resistivity curve correction model training module is used to: build a ResNet-Transformer network architecture, use the noisy apparent resistivity curve as training input, and the corresponding noise-free apparent resistivity curve as training output, to train and obtain the apparent resistivity curve correction model;

[0108] The apparent resistivity curve correction model is used to correct the real-time input controlled source electromagnetic apparent resistivity curve affected by noise and output a smooth apparent resistivity curve.

[0109] It should be understood that the functional unit modules in various embodiments of the present invention can be concentrated in one processing unit, or each unit module can exist physically separately, or two or more unit modules can be integrated into one unit module, and can be implemented in the form of hardware or software.

[0110] The above embodiments are preferred embodiments of the present application. Ordinary technicians in this field can also make various changes or improvements on this basis. Without departing from the overall concept of the present application, these changes or improvements should fall within the scope of protection required by the present application.

Claims

1. A controlled source electromagnetic apparent resistivity curve correction method based on deep learning, characterized in that: include: S1, generates N apparent resistivity curves through wave impedance recursion theory, which are recorded as the apparent resistivity curves without noise; S2, adding random noise to the apparent resistivity curve without noise according to the influence law of electromagnetic noise to obtain the apparent resistivity curve with noise; S3, building a ResNet-Transformer network architecture, taking the apparent resistivity curve containing noise as the training input and the corresponding apparent resistivity curve without noise as the training output, and training to obtain the apparent resistivity curve correction model; S4, inputting the controlled source electromagnetic apparent resistivity curve affected by noise, which is collected in real time, into the apparent resistivity curve correction model, correcting the input apparent resistivity curve through the model, and outputting a smooth apparent resistivity curve.

2. The controlled source electromagnetic apparent resistivity curve correction method according to claim 1, characterized in that: The apparent resistivity curves are generated by wave impedance recursion theory, including: S1.

1. Establish a multi-layer geoelectric structure model and define the range of resistivity and thickness of each layer: ； Where k represents the layer index of the multi-layer geoelectric structure, and represent the resistivity and thickness of the kth layer of geoelectric structure respectively; and represent the minimum and maximum resistivity of the kth layer, and Represent the minimum and maximum thickness of the kth layer respectively; S1.2, for each layer of geoelectric structure, within the defined range of resistivity and thickness, randomly sample the resistivity and thickness, and introduce resistivity flipping mechanisms with different probabilities to simulate various geological anomalies; S1.3, based on the electromagnetic field diffusion theory, the equivalent wave impedance of the surface of each layer of the geoelectric structure is recursively deduced from the base layer upwards: ； Where, represents the equivalent wave impedance of the surface of the k-th geoelectric structure, It represents the equivalent wave impedance of the surface of the k+1th layer of geoelectric structure, represents the eigenwave impedance of the k-th geoelectric structure, represents the propagation constant of the k-th layer geoelectric structure, represents the thickness of the kth layer of geoelectric structure; the eigenwave impedance and propagation constant of each layer of geoelectric structure are related to the resistivity and emission frequency of the geoelectric structure of that layer; S1.4, generate the apparent resistivity of the multi-layer geoelectric structure based on the equivalent wave impedance of the first layer of geoelectric structure surface : ； Where, is the angular frequency, is the magnetic permeability, is the emission frequency of the controlled source electromagnetic method, and ; S1.5, based on the apparent resistivity obtained at different emission frequencies using the controlled source electromagnetic method, a line is constructed based on the emission frequency. Apparent resistivity curve with is the independent variable.

3. The controlled source electromagnetic apparent resistivity curve correction method according to claim 2, characterized in that: The eigenwave impedance and propagation constant of each layer of geoelectric structure are calculated as follows: ； Where i is the imaginary unit.

4. The controlled source electromagnetic apparent resistivity curve correction method according to claim 1, characterized in that: Each apparent resistivity curve generated by S1 through wave impedance recursion theory is filtered and then used as the data source of S2 and / or S3.

5. The controlled source electromagnetic apparent resistivity curve correction method according to claim 1, characterized in that: The adding of random noise to the noise-free apparent resistivity curve according to the electromagnetic noise influence law includes: S2.1, let any i-th noise-free apparent resistivity curve be Expressed as: ； Where, Represents the apparent resistivity curve The apparent resistivity value in They represent the controlled source electromagnetic method Transmitting frequency; S2.2, add random noise generated as follows : ； Where A represents the intensity of random noise, Indicates the generation of [-1,1] A sequence of random values; S2.3, Add to The i-th apparent resistivity curve containing noise in the logarithmic domain is obtained : ； S2.4, Final Calculation , which is the i-th noisy apparent resistivity curve containing noise in the final conventional number domain.

6. The controlled source electromagnetic apparent resistivity curve correction method according to claim 1, characterized in that: The constructed ResNet-Transformer network architecture includes, from input to output, the following layers: ResNet network layer, Transformer network layer, and regression layer; The ResNet network layer first performs one-dimensional convolution, one-dimensional pooling, and ReLU activation on the input noisy apparent resistivity sequence; then continues to perform one-dimensional convolution, one-dimensional pooling, and ReLU activation on the output of the first ReLU activation; then performs one-dimensional convolution on the output of the first ReLU activation, and then adds it to the output of the second ReLU activation as the output of the ResNet network layer; The Transformer network layer first performs position encoding on the output of the ResNet network layer, then processes it using a self-attention mechanism, and then outputs it after passing through a dropout layer and a fully connected layer; The regression layer regresses the output of the Transformer network layer to obtain a noise-free apparent resistivity sequence.

7. The controlled source electromagnetic apparent resistivity curve correction method according to claim 1, characterized in that: When training the apparent resistivity curve correction model in S3, the apparent resistivity curve that needs to be input and output into the model is first converted into the logarithmic domain and then used as the training input and output of the model; When S4 uses the trained apparent resistivity curve correction model for correction, the apparent resistivity value of the noise-affected controlled source electromagnetic apparent resistivity curve acquired in real time is converted to a logarithmic domain value and then input into the apparent resistivity curve correction model; and the output data of the apparent resistivity curve correction model is converted from the logarithmic domain to the conventional domain, which is the final smoothed apparent resistivity curve.

8. A system based on the controlled source electromagnetic apparent resistivity curve correction method according to any one of claims 1 to 7, characterized in that: include: The smoothed apparent resistivity curve generation module is used to: generate N apparent resistivity curves by wave impedance recursion theory, which are recorded as noise-free apparent resistivity curves; The noisy apparent resistivity curve generation module is used to add random noise to the noise-free apparent resistivity curve according to the influence law of electromagnetic noise to obtain the noise-containing apparent resistivity curve; The apparent resistivity curve correction model training module is used to: build a ResNet-Transformer network architecture, use the noisy apparent resistivity curve as training input, and the corresponding noise-free apparent resistivity curve as training output, to train and obtain the apparent resistivity curve correction model; The apparent resistivity curve correction model is used to correct the real-time input controlled source electromagnetic apparent resistivity curve affected by noise and output a smooth apparent resistivity curve.

Images