Generative adversarial network-based resistivity logging-while-drilling real-time image super-resolution reconstruction method, system and equipment and medium

Abstract

The invention discloses a resistivity-while-drilling logging real-time image super-resolution method, system and equipment based on a generative adversarial network, and a medium, and the method comprises the steps: firstly, reading resistivity-while-drilling storage data of a certain region, obtaining a high-resolution image from the storage data according to pixels, carrying out the sampling of the resistivity-while-drilling storage data at equal intervals, and carrying out the imaging, obtaining a corresponding low-resolution image, wherein the high-resolution image and the corresponding low-resolution image form a paired image data set; secondly, a generative adversarial network model is built, the generative adversarial network model comprises a generator and a discriminator, the generator takes a low-resolution image as input and outputs a high-resolution image through stacked SE-RRDB modules and up-sampling operation, and the discriminator evaluates the generated image through multilayer convolution and jump connection and judges the authenticity of the image; thirdly, training the generative adversarial network model by the image data set, optimizing the network, and storing the optimal generative adversarial network model; and finally, acquiring resistivity logging-while-drilling real-time data, processing the data, and inputting the processed data into the trained generative adversarial network model to obtain a real-time image after super-resolution. The method has the advantages of rapidness, high efficiency, good reconstruction effect and high practicability.

Description

Technical Field

[0001] The present invention relates to the technical field of oil well detection, and in particular to a method, system, equipment and medium for real-time super-resolution reconstruction of resistivity logging while drilling based on a generative adversarial network. Background Art

[0002] Imaging logging is an indispensable and important means for modern geological exploration and oil and gas development. Imaging logging uses instruments to scan the well wall, and generates high-resolution two-dimensional or three-dimensional images by collecting and processing various signal data such as resistivity, acoustic waves, and nuclear magnetic resonance of the well wall, so as to analyze the lithology, structural characteristics, fracture distribution, porosity and other information of the formation in detail, and provide accurate data support for geological research and oil and gas reservoir description. Among the images generated by imaging logging, resistivity images are very sensitive to electrical changes around the well wall, so resistivity images have the highest resolution. With the help of resistivity images, even very small features, such as natural or induced fractures, can be clearly identified.

[0003] Initially, logging images were scanned by cable-based instruments. Cable logging works by pressing the sensor against the borehole wall, allowing it to slide as the tool passes through the wellbore. However, this method is prone to sensor wear during drilling. LWD resistivity logging uses a rotating drill as a sensor platform, allowing the sensor to move along a spiral path, thereby reducing wear. By transmitting the data measured by the sensor to the surface in real time, a series of problems that cannot be handled by cable logging technology are effectively solved. The focus is on wellbore integrity, geosteering, and completion stage optimization. By identifying wellbore events and formation characteristics, high-resolution real-time images can significantly improve the effectiveness of these applications. High-resolution real-time images improve wellbore stability management by analyzing drilling-induced fractures, wellbore breakouts, etc. Geosteering uses high-resolution real-time data to accurately adjust the drilling direction to ensure that the wellbore is in the best position in the target formation. High-resolution data makes real-time bedding evaluation and structural analysis more accurate, thereby optimizing the drilling path and reducing risks during drilling. In the completion stage, completion decisions can be made faster by identifying key geological structures that can affect the communication or bonding quality between production or injection zones. In Trevizan WA, de Jesus C M. Application of GAN to Resolution Enhancement of LWD Real-Time Image Logs to Support Decision Making [J]. Petrophysics-The SPWLA Journal of Formation Evaluation and Reservoir Description, 2023, 64 (06): 890-899., Willian et al. used a generative adversarial network-based model to super-reconstruct low-resolution real-time images to obtain better quality super-resolution images, so as to timely identify geological and geomechanical structures and support the decision-making of completion operations. Although good results were achieved, the effect of super-resolution reconstruction of low-resolution images under different hyperparameter combinations was not measured. Summary of the invention

[0004] In order to solve the defects of the above-mentioned prior art, the purpose of the present invention is to provide a method, system, equipment and medium for super-resolution reconstruction of real-time images of resistivity logging while drilling based on a generative adversarial network; construct a high-resolution storage data image and a low-resolution real-time data image data set; use a deep learning algorithm to train the constructed paired image data set, and after the training is completed, perform super-resolution reconstruction of the low-resolution image to approach the resolution of the storage data, which can not only make up for the shortcomings of real-time data, but also optimize decision support and improve the risk management capability during the drilling process.

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

[0006] A method for real-time super-resolution reconstruction of resistivity logging while drilling based on a generative adversarial network specifically comprises the following steps:

[0007] Step 1: Read the resistivity storage data while drilling in different regions, sample the resistivity storage data at equal intervals and then image them to obtain corresponding low-resolution images; map the resistivity storage data while drilling to a pixel interval of 0-255, image them according to pixels, and obtain logging images as high-resolution images for storage; the high-resolution images and the corresponding low-resolution images are paired image data sets; perform data expansion on the paired image data sets to obtain an expanded image data set;

[0008] Step 2: Build a generative adversarial network model based on the expanded image dataset obtained in step 1;

[0009] Step 3: Use the expanded image data set obtained in step 1 to train the generative adversarial network model in step 2, optimize the network model, and save the optimal generative adversarial network model;

[0010] Step 4: Obtain the real-time data of resistivity logging while drilling in the target area, map the real-time data of resistivity logging while drilling in the target area to a pixel range of 0-255, image the pixels and input the generated adversarial network model trained in step 3 to obtain a real-time image after super resolution; and compare it with the high-resolution image stored in the resistivity logging while drilling to measure the super-resolution effect.

[0011] The specific method of building a generative adversarial network model in step 2 is:

[0012] The generative adversarial network model includes a generator and a discriminator. The generator takes the low-resolution image obtained in step 1 as input, and includes a convolutional layer, a SE-RRDB module, and an upsampling module. The basic features of the image are extracted through the convolutional layer in the generator; the basic features of the image are extracted and fused through the SE-RRDB module; the upsampling module adjusts the number of channels of the convolutional layer and samples the fused image features to the target output size; the low-resolution image is converted into a super-resolution image output;

[0013] The discriminator uses a U-Net structure with spectral normalization. It evaluates the input high-resolution images and super-resolution images through multi-layer convolution and jump connections to determine whether the high-resolution images and super-resolution images are similar. It then feeds the discrimination results back to the generator to guide the generator to generate more realistic images.

[0014] Set the loss function of the generative adversarial network model to adjust the network parameters of the generator and discriminator until the final generative adversarial network model is obtained. The loss function is:

[0015] L SR =c1L1+c p L p

[0016] L1=‖I out -I gt ‖1

[0017]

[0018] Among them, L SR is the loss function, L1 is the pixel loss function, c adv is the adversarial weight, L p is the perceptual loss function, c p is the perception weight. out is the image generated by the generator, I gt is a real high-resolution image; ‖‖1 represents the L1 norm; N is the number of feature maps, φ i () is the feature map of the i-th layer of the pre-trained VGG network.

[0019] The process of step 3 is as follows:

[0020] The paired image dataset in step 1 is divided into a training set and a test set. The training set is used to train the generative adversarial network model. The test set data is imported into the trained generative adversarial network model for testing. The low-resolution image is super-reconstructed. According to the super-resolution reconstruction result, the hyperparameters of the generative adversarial network model are adjusted. The peak signal-to-noise ratio and structural similarity of the image are counted to find the optimal hyperparameter setting. The network model corresponding to the optimal hyperparameter is the optimal generative adversarial network model, and the optimal generative adversarial network model is saved.

[0021] The hyperparameters include learning rate, batch size, and loss function weight.

[0022] The method for finding the optimal hyperparameter setting is: using different learning rate and batch size combinations to respectively verify the super-resolution reconstruction effect of the generative adversarial network under different training parameters, and selecting the learning rate and batch size corresponding to the average peak signal-to-noise ratio and the minimum structural similarity as the optimal parameters of the generative adversarial network model.

[0023] A real-time image super-resolution reconstruction system for resistivity logging while drilling based on a generative adversarial network, comprising:

[0024] The first acquisition module is used to acquire the resistivity storage data while drilling in different regions, map the resistivity storage data while drilling to a pixel interval of 0-255, image the resistivity storage data as a high-resolution image according to the pixel, image the resistivity storage data while drilling after equal-interval sampling, and obtain the corresponding low-resolution image; used in step 1 to realize the construction of the data set;

[0025] A generative adversarial network model building module is used to build a generative adversarial network model, which includes a generator and a discriminator; that is, it is used in step 2 to realize the construction of a generative adversarial network model;

[0026] The generator includes a convolutional layer, a SE-RRDB module and an upsampling module. The convolutional layer extracts the basic features of the image. The SE-RRDB module extracts and fuses the basic features of the image. The upsampling module adjusts the number of channels of the convolutional layer and samples the fused image features to the target output size. It is used in step 2 to transform the low-resolution image into a super-resolution image. The SE-RRDB module includes three RDB submodules and one SENet submodule connected in sequence. The SENet submodule includes a global average pooling layer, a fully connected layer, and an activation function connected in sequence, and improves the performance of the convolutional neural network by introducing a channel attention mechanism. The RDB submodule includes multiple convolutional layers connected in sequence, and the output of each convolutional layer is spliced ​​with the output of the previous convolutional layer.

[0027] The discriminator includes multiple convolutional layers, and skip connections are added between the convolutional layers to retain low-level features, helping the discriminator to better capture multi-scale information; it is used in step 2 to obtain the discrimination results for the input high-resolution image and super-resolution image, thereby improving the final effect of super-resolution reconstruction;

[0028] Feedback module, which sets the loss function of the generative adversarial network model to adjust the network parameters of the generator and discriminator until the final generative adversarial network model is obtained;

[0029] The network optimization module is used in step 3 to divide the paired image data set constructed in step 1 into a training set and a test set, use the training set to train the generative adversarial network model, import the test set data into the trained generative adversarial network model for testing, and perform super-resolution reconstruction on the low-resolution image; according to the super-resolution result, adjust the hyperparameters of the generative adversarial network model; count the peak signal-to-noise ratio and structural similarity of the image to find the optimal hyperparameter setting, the generative adversarial network model corresponding to the optimal hyperparameter is the optimal super-resolution model, and the optimal super-resolution model is saved;

[0030] The second acquisition module is used in step 4 to acquire the target area's resistivity logging while drilling data and scale it into a resistivity logging while drilling image to achieve acquisition of the target resistivity logging while drilling real-time image and resistivity logging while drilling storage image;

[0031] A super-resolution module is used to input the target resistivity logging while drilling real-time data into a super-resolution model to obtain a super-resolution image, which is used in step 4 to achieve super-resolution reconstruction of the low-resolution resistivity logging while drilling real-time image;

[0032] The evaluation module is used in step 4 to compare the real-time image obtained after super-resolution reconstruction with the processed stored data to measure the super-resolution effect.

[0033] A real-time image super-resolution device for resistivity logging while drilling based on a generative adversarial network, comprising:

[0034] A memory and a processor, wherein the memory stores a computer program, and is used to perform super-resolution reconstruction of real-time images according to the method for super-resolution reconstruction of resistivity logging while drilling based on a generative adversarial network described in steps 1 to 4 when executing the computer program, so as to improve image detail information and enhance image resolution and clarity.

[0035] A computer-readable storage medium stores a computer program. The computer-readable storage medium provides the function of storing and executing an image processing program. When the computer program is executed by a processor, the method for super-resolution reconstruction of resistivity logging while drilling real-time images based on a generative adversarial network described in steps 1 to 4 can be used to achieve efficient and rapid super-resolution reconstruction of resistivity logging while drilling real-time images.

[0036] Compared with the prior art, the beneficial effects of the present invention are:

[0037] Due to the slow communication speed of mud pulse telemetry technology and the increase in the amount of data in the measurement while drilling system, the gap between real-time data and stored data is further widened. Therefore, the present invention reconstructs the real-time data with super-resolution to approach the resolution of the stored data, which can not only make up for the deficiency of real-time data, but also optimize decision support and improve the risk management capability during drilling.

[0038] The present invention achieves more sophisticated and effective real-time image super-resolution processing in the generative adversarial network by introducing the attention mechanism and optimizing the training parameters. Compared with the prior art, the present invention has significantly improved the effects in detail reconstruction, training stability and practical application, can provide higher quality images, and provide more reliable data support for practical applications such as resistivity logging while drilling.

[0039] The present invention proposes a real-time image super-resolution reconstruction method based on a generative adversarial network. By introducing an attention mechanism into the generator, the improved generator can integrate more image information. The real-time image is super-resolution reconstructed using the optimal training parameters, and the real-time image super-resolution effect under different resolutions is tested, which makes the subsequent completion phase optimization with the help of the real-time image of the well logging while drilling more reliable. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 This is a flow chart of super-resolution reconstruction of the present invention.

[0041] Figure 2 It is a schematic diagram of the overall network structure of the present invention.

[0042] Figure 3 It is a schematic diagram of the generator structure of the present invention.

[0043] Figure 4 It is a structural schematic diagram of the SE-RRDB submodule of the present invention.

[0044] Figure 5 It is a structural diagram of the SENet submodule of the present invention.

[0045] Figure 6 Schematic diagram of the structure of the discriminator of the present invention.

[0046] Figure 7 This is the super-resolution reconstruction effect diagram. DETAILED DESCRIPTION

[0047] The present invention is further described in detail below with reference to the accompanying drawings and embodiments.

[0048] See also Figure 1 , a real-time image super-resolution reconstruction method of resistivity logging while drilling based on a generative adversarial network, comprising the following steps:

[0049] Step 1. Read the resistivity storage data while drilling in different regions, sample the resistivity storage data at equal intervals and then image them to obtain corresponding low-resolution images; map the resistivity storage data while drilling to a pixel interval of 0-255, image them according to pixels, and obtain logging images as high-resolution images for storage; the high-resolution images and the corresponding low-resolution images are paired image data sets; perform data expansion on the paired image data sets to obtain an expanded image data set;

[0050] In step 1, the resistivity storage data while drilling has 132 data within a depth of 1m, and 224 sectors horizontally. In order to convert the stored data into image display, the conductivity curve range must be read first, and mapped to a pixel interval of 0-255 according to the conductivity range. Finally, imaging is performed according to pixels to obtain a high-resolution data set with a size of 224×132. Due to the limitations of telemetry technology, the stored data cannot be fully transmitted to the ground. Therefore, the resistivity storage data while drilling is sampled at equal intervals, and the sampled data is imaged to obtain a low-resolution data set with a size of 56×34. A total of 500 pairs of high-resolution-low-resolution image pairs are constructed as a data set. To expand the data set, the data set is horizontally flipped and rotated to obtain 1000 pairs of image pairs.

[0051] Step 2. Build a generative adversarial network model based on the expanded image dataset obtained in step 1;

[0052] In this embodiment, the following methods are specifically included:

[0053] The network structure is as follows Figure 2 As shown in the figure, the generative adversarial network model includes a generator and a discriminator. The generator takes the low-resolution image obtained in step 1 as input, including a convolutional layer, a SE-RRDB module and an upsampling module. The LR image is a low-resolution image, the SR image is a super-resolution image, and the HR image is a real image. The basic features of the image are extracted through the convolutional layer in the generator; the basic features of the image are extracted and fused through the SE-RRDB module; the upsampling module adjusts the number of channels of the convolutional layer and samples the fused image features to the target output size; the discriminator inputs the generated super-resolution image and the original high-resolution image into the discriminator at the same time, and determines the probability that the image is a real image. Through the continuous confrontation between the generator and the discriminator, the super-resolution image outputted in the end becomes closer to the real image;

[0054] like Figure 3 As shown, the low-resolution input image first passes through a convolutional layer to extract primary features. After feature extraction, it passes through multiple stacked SE-RRDBs to extract image depth features. It is then sampled back to the original size through an upsampling module and outputs a high-resolution image through a convolutional layer.

[0055] like Figure 4 As shown, the RDB submodule includes a plurality of sequentially connected convolutional layers, and the output of each convolutional layer is concatenated with the output of the previous convolutional layer.

[0056] like Figure 5As shown in the figure, the SENet submodule is a module that improves the performance of convolutional neural networks by introducing a channel attention mechanism. The SENet module mainly consists of three parts: compression, excitation, and scaling. The compression operation compresses the features of each channel into a single value through global average pooling, thereby achieving global information aggregation of the features. The excitation operation recalibrates the features of each channel through a two-layer fully connected network. First, there is a dimensionality reduction operation, followed by a dimensionality increase operation. The scaling operation applies the weights generated by the excitation operation to each channel of the input feature map to achieve channel-level feature recalibration.

[0057] like Figure 6 As shown, the discriminator in step 2 includes multiple convolutional layers, and jump connections are added between the convolutional layers to retain low-level features and help the discriminator better capture multi-scale information; used in step 2, the input high-resolution image and super-resolution image are discriminated to obtain the final effect of super-resolution reconstruction;

[0058] The loss function of the generative adversarial network model is:

[0059] L SR =c1L1+c p L p

[0060] L1=‖I out -I gt ‖1

[0061]

[0062] Among them, L SR is the loss function, L1 is the pixel loss function, c adv is the adversarial weight, L p is the perceptual loss function, c p is the perception weight. out is the image generated by the generator, I gt is the real high-resolution image. ‖‖1 represents the L1 norm. N is the number of feature maps, φ i () is the feature map of the i-th layer of the pre-trained VGG network.

[0063] Step 3. Divide the constructed paired image dataset into a training set and a test set in a ratio of 9:1. Use the training set to train the generative adversarial network model, import the test set image dataset into the trained network for testing, and perform real-time image super-resolution; according to the super-resolution results, adjust the hyperparameters of the neural network, the hyperparameters include learning rate, batch size, and loss function weight; calculate the peak signal-to-noise ratio and structural similarity of the image, compare the image reconstruction effect of the generative adversarial network model under different hyperparameters, and find the optimal hyperparameter setting. The network model corresponding to the optimal hyperparameter is the generative adversarial network model with the best super-resolution effect, and save it;

[0064] Step 4. Obtain the real-time resistivity logging while drilling data of the target area, map the real-time resistivity logging while drilling data of the target area to a pixel range of 0-255, image it according to the pixels, and input it into the generative adversarial network model trained in step 3 to obtain a real-time image after super resolution. Figure 7 The first column is the depth channel, which records the measured value of the well depth during the logging process; the second column is the input logging image to be reconstructed; the third column is the reconstructed logging image; and the fourth column is the high-resolution logging image.

[0065] Table 1 selects two super-resolution methods based on the generative adversarial network model and an interpolation method and the method of this embodiment (SE-ESRGAN) to quantitatively evaluate the reconstructed image using peak signal-to-noise ratio and structural similarity; the peak signal-to-noise ratio is the degree of closeness between the super-resolution image and the high-resolution image at the pixel level; the structural similarity is the degree of closeness between the super-resolution image and the high-resolution image at the brightness, contrast, and structure level; by comparison in Table 1, it can be seen that the peak signal-to-noise ratio and structural similarity are both optimal, indicating that this method can reconstruct the real-time logging image with high quality.

[0066] Table 1

[0067]

Claims

1. A method, system, device and medium for real-time image super-resolution reconstruction of resistivity logging while drilling based on generative adversarial network, characterized in that: The specific steps include: Step 1: Read the resistivity storage data while drilling in different regions, sample the resistivity storage data at equal intervals and then image them to obtain corresponding low-resolution images; map the resistivity storage data while drilling to a pixel interval of 0-255, image them according to pixels, and obtain logging images as high-resolution images for storage; the high-resolution images and the corresponding low-resolution images are paired image data sets; perform data expansion on the paired image data sets to obtain an expanded image data set; Step 2: Based on the expanded image dataset obtained in step 1, build a generative adversarial network model; Step 3: Use the expanded image data set obtained in step 1 to train the generative adversarial network model in step 2, optimize the network model, and save the optimal generative adversarial network model; Step 4: Obtain the real-time data of resistivity logging while drilling in the target area, map the real-time data of resistivity logging while drilling in the target area to a pixel range of 0-255, image the pixels and input the generated adversarial network model trained in step 3 to obtain a real-time image after super resolution; and compare it with the high-resolution image stored in the resistivity logging while drilling to measure the super-resolution effect.

2. The method for real-time image super-resolution reconstruction of resistivity logging while drilling based on generative adversarial network according to claim 1 is characterized in that: The specific method of building a generative adversarial network model in step 2 is: The generative adversarial network model includes a generator and a discriminator. The generator takes the low-resolution image obtained in step 1 as input, and includes a convolutional layer, a SE-RRDB module, and an upsampling module. The basic features of the image are extracted through the convolutional layer in the generator; the basic features of the image are extracted and fused through the SE-RRDB module; the upsampling module adjusts the number of channels of the convolutional layer and samples the fused image features to the target output size; the low-resolution image is converted into a super-resolution image output; The discriminator uses a U-Net structure with spectral normalization. It evaluates the input high-resolution images and super-resolution images through multi-layer convolution and jump connections to determine whether the high-resolution images and super-resolution images are similar. It then feeds the discrimination results back to the generator to guide the generator to generate more realistic images. Set the loss function of the generative adversarial network model to adjust the network parameters of the generator and discriminator until the final generative adversarial network model is obtained. The loss function is: L SR =c1L1+c p L p L1=‖I out -I gt ‖1 Among them, L SR is the loss function, L1 is the pixel loss function, c adv is the adversarial weight, L p is the perceptual loss function, c p is the perception weight. out is the image generated by the generator, I gt is a real high-resolution image; ‖‖1 represents the L1 norm; N is the number of feature maps, φ i () is the feature map of the i-th layer of the pre-trained VGG network.

3. The method, system, device and medium for real-time image super-resolution reconstruction of resistivity logging while drilling based on generative adversarial network according to claim 1, characterized in that: The process of step 3 is as follows: The paired image dataset in step 1 is divided into a training set and a test set. The training set is used to train the generative adversarial network model. The test set data is imported into the trained generative adversarial network model for testing. The low-resolution image is super-reconstructed. According to the super-resolution reconstruction result, the hyperparameters of the generative adversarial network model are adjusted. The peak signal-to-noise ratio and structural similarity of the image are counted to find the optimal hyperparameter setting. The network model corresponding to the optimal hyperparameter is the optimal generative adversarial network model, and the optimal generative adversarial network model is saved.

4. The method for real-time super-resolution reconstruction of resistivity logging while drilling based on generative adversarial network according to claim 3 is characterized in that: The hyperparameters include learning rate, batch size, and loss function weight.

5. The method for real-time super-resolution reconstruction of resistivity logging while drilling based on generative adversarial network according to claim 1, characterized in that: The method for finding the optimal hyperparameter setting is: using different learning rate and batch size combinations to respectively verify the super-resolution reconstruction effect of the generative adversarial network under different training parameters, and selecting the learning rate and batch size corresponding to the average peak signal-to-noise ratio and the minimum structural similarity as the optimal parameters of the generative adversarial network model.

6. A real-time image super-resolution reconstruction system for resistivity logging while drilling based on generative adversarial networks, characterized in that: include: The first acquisition module is used to acquire the resistivity storage data while drilling in different regions, map the resistivity storage data while drilling to a pixel interval of 0-255, image the resistivity storage data as a high-resolution image according to the pixel, image the resistivity storage data while drilling after equal-interval sampling, and obtain the corresponding low-resolution image; used in step 1 to realize the construction of the data set; A generative adversarial network model building module is used to build a generative adversarial network model, which includes a generator and a discriminator; that is, it is used in step 2 to realize the construction of a generative adversarial network model; The generator includes a convolutional layer, a SE-RRDB module, and an upsampling module. The convolutional layer extracts the basic features of the image; the SE-RRDB module extracts and fuses the basic features of the image; The upsampling module adjusts the number of channels of the convolution layer and samples the fused image features to the target output size; it is used in step 2 to transform the low-resolution image into a super-resolution image; the SE-RRDB module includes three RDB submodules and one SENet submodule connected in sequence; the SENet submodule includes a global average pooling layer, a fully connected layer, and an activation function connected in sequence, and improves the performance of the convolutional neural network by introducing a channel attention mechanism; the RDB submodule includes multiple convolutional layers connected in sequence, and the output of each convolutional layer is spliced ​​with the output of the previous convolutional layer; The discriminator includes multiple convolutional layers, and skip connections are added between the convolutional layers to retain low-level features, helping the discriminator to better capture multi-scale information; it is used in step 2 to obtain the discrimination results for the input high-resolution image and super-resolution image, thereby improving the final effect of super-resolution reconstruction; Feedback module, which sets the loss function of the generative adversarial network model to adjust the network parameters of the generator and discriminator until the final generative adversarial network model is obtained; The network optimization module is used in step 3 to divide the paired image data set constructed in step 1 into a training set and a test set, use the training set to train the generative adversarial network model, import the test set data into the trained generative adversarial network model for testing, and perform super-resolution reconstruction on the low-resolution image; according to the super-resolution result, adjust the hyperparameters of the generative adversarial network model; count the peak signal-to-noise ratio and structural similarity of the image to find the optimal hyperparameter setting, the generative adversarial network model corresponding to the optimal hyperparameter is the optimal super-resolution model, and the optimal super-resolution model is saved; The second acquisition module is used in step 4 to acquire the target area's resistivity logging while drilling data and scale it into a resistivity logging while drilling image to achieve acquisition of the target resistivity logging while drilling real-time image and resistivity logging while drilling storage image; A super-resolution module is used to input the target resistivity logging while drilling real-time data into a super-resolution model to obtain a super-resolution image, which is used in step 4 to achieve super-resolution reconstruction of the low-resolution resistivity logging while drilling real-time image; The evaluation module is used in step 4 to compare the real-time image obtained after super-resolution reconstruction with the processed stored data to measure the super-resolution effect.

7. A real-time image super-resolution device for resistivity logging while drilling based on a generative adversarial network, characterized in that: include: A memory and a processor, wherein the memory stores a computer program, and is used to perform super-resolution reconstruction of real-time images according to the method for super-resolution reconstruction of resistivity logging while drilling based on a generative adversarial network described in steps 1 to 4 when executing the computer program, so as to improve image detail information and enhance image resolution and clarity.

8. A computer-readable storage medium storing a computer program, characterized in that: The computer-readable storage medium provides the function of storing and executing an image processing program. When the computer program is executed by the processor, it can realize efficient and rapid super-resolution reconstruction of real-time resistivity logging while drilling images based on the generative adversarial network-based real-time resistivity logging while drilling super-resolution reconstruction method described in steps 1 to 4.

Images