Regional chain type three-dimensional reconstruction method and device for rock reservoir engineering scale seepage model

Abstract

The invention provides a rock reservoir engineering scale seepage model regional chain type three-dimensional reconstruction method and device, and the method comprises the steps: carrying out the feature extraction and fusion of a pore scale and test piece scale seepage feature channel map based on a multi-scale attention mechanism, and obtaining a dual-channel feature fusion model; constructing and training a regional chain-type generative adversarial network on the dual-channel feature fusion model, and generating a chain-type structure block taking the dual-channel feature fusion model as a basic unit through the trained generative adversarial network; and under the condition that the seepage channel feature matching degree of the plurality of generated chain type structure blocks at the test piece scale and the pore scale is not less than a preset matching degree threshold value, obtaining a rock reservoir three-dimensional seepage model at the engineering scale based on the plurality of generated chain type structure blocks. According to the method, the scale limitation of the rock reservoir seepage model is broken through, and the three-dimensional model of the seepage channel with the similar model scale and the engineering scale of the rock reservoir is established.

Description

Technical Field

[0001] The present invention relates to the technical field of deep learning, and in particular, to a method and device for region-chain three-dimensional reconstruction of a seepage model at the engineering scale of a rock reservoir. Background Art

[0002] Based on CO 2 In recent years, the number of engineering projects for enhancing the exploitation of deep-earth reservoir resources through geological sequestration has gradually increased. However, due to the high-permeability environment of deep reservoirs and the "black box" state of the internal seepage channel structure of deep rock reservoirs, the migration and evolution mechanisms of various resource fluids (such as oil and geothermal energy) within deep rock reservoirs are not clearly understood, greatly reducing the evaluation accuracy of the resource reserves and exploitation volume of rock reservoirs.

[0003] Currently, the modeling methods for rock reservoir seepage models mainly include three types of modeling techniques: theoretical-based reservoir seepage models (three-dimensional percolation models and equivalent medium reservoir seepage models), numerical simulation-based reservoir seepage models (fracture reservoir fluid models and lattice-capillary-suspension seepage models), and physical characteristic-based reservoir seepage models (reservoir seismic rock seepage models and tight sandstone reservoir physical seepage models).

[0004] Among them, the construction technology of theoretical-based reservoir seepage models, due to using equivalent bodies for characterization, ignores the interaction effects of different pores, cracks, and fluids, and there are certain differences between its theoretical parameters and actual situations; the construction technology of numerical simulation-based reservoir seepage models, due to using simplified reservoir structures, ignores the internal fluid flow characteristics and reservoir structure evolution characteristics of reservoirs; the construction technology of physical characteristic-based reservoir seepage models overly relies on reservoir physical quantity characteristics (such as wave velocity, etc.), and mostly uses average seepage property parameters to calculate reservoir seepage characteristics, resulting in insufficient characterization of the heterogeneous structure characteristics and seepage property distribution of rock reservoirs.

[0005] That is to say, the existing modeling technologies all have deficiencies such as model simplification, low construction accuracy, limited application scope, and non-visualization of seepage channels. Summary of the Invention

[0006] The present invention provides a method and device for region-chain three-dimensional reconstruction of a seepage model at the engineering scale of a rock reservoir, aiming to solve the defect of low construction accuracy of the existing three-dimensional modeling technology for model reservoirs at the engineering scale in the prior art, and to realize a method for region-chain three-dimensional reconstruction of a seepage model at the engineering scale of a rock reservoir with a wide application scope and high construction accuracy.

[0007] The present invention provides a method for region-chain three-dimensional reconstruction of a seepage model at the engineering scale of a rock reservoir, including: Obtaining a pore-scale seepage characteristic channel map and a specimen-scale seepage characteristic channel map of a rock specimen in a target area respectively; Feature extraction and fusion based on a multi-scale attention mechanism are performed on the pore-scale seepage characteristic channel map and the specimen-scale seepage characteristic channel map to obtain a dual-channel feature fusion model; On the dual-channel feature fusion model, a region-chain generative adversarial network is constructed and trained, and a chain-structured block based on the dual-channel feature fusion model is generated through the trained generative adversarial network; When the matching degrees of the generated multiple chain-structured blocks with the seepage channel characteristics at the specimen scale and the pore scale are not less than a preset matching degree threshold, a three-dimensional seepage model of a rock reservoir at the engineering scale is obtained based on the generated multiple chain-structured blocks.

[0008] According to a method for regional chain three-dimensional reconstruction of a seepage model of a rock reservoir at the engineering scale provided by the present invention, the step of performing feature extraction and fusion based on a multi-scale attention mechanism on the pore-scale seepage characteristic channel map and the specimen-scale seepage characteristic channel map to obtain a dual-channel feature fusion model specifically includes: Local feature extraction and global feature extraction are respectively performed on the pore-scale seepage characteristic channel map and the specimen-scale seepage characteristic channel map; The pore-scale seepage characteristic channel map and the specimen-scale seepage characteristic channel map are respectively subjected to feature fusion of local features and global features to obtain a pore-scale fusion feature and a specimen-scale fusion feature; The pore-scale fusion feature and the specimen-scale fusion feature are subjected to feature fusion to obtain a dual-channel feature fusion model.

[0009] According to a method for regional chain three-dimensional reconstruction of a seepage model of a rock reservoir at the engineering scale provided by the present invention, before the step of constructing and training a region-chain generative adversarial network on the dual-channel feature fusion model, it further includes: Digital labeling is performed on the dual-channel feature fusion model, and the digital labeling is used to label the connected seepage channels and the non-connected seepage channels of the dual-channel feature fusion model.

[0010] According to a method for regional chain three-dimensional reconstruction of a seepage model of a rock reservoir at the engineering scale provided by the present invention, the step of constructing and training a region-chain generative adversarial network on the dual-channel feature fusion model specifically includes: The dual-channel feature fusion model is used as the input of the generative adversarial network for the generative block channel region chain; A loss function of the generator of the generative adversarial network is constructed based on generation and content loss; A loss function of the discriminator of the generative adversarial network is constructed based on the mean square error of the probability density functions of the input features and the fusion features.

[0011] A method for regional chain three-dimensional reconstruction of a seepage model at the engineering scale of a rock reservoir according to the present invention further includes: Obtain the seepage channel generation characteristics of the generated multiple chain structure blocks at the specimen scale and the pore scale; Calculate the first error between the seepage channel generation characteristics at the specimen scale and the seepage channel characteristics; Calculate the second error between the seepage channel generation characteristics at the pore scale and the seepage channel characteristics; Use the linear weighted result of the first error and the second error as the seepage channel feature matching degree of the chain structure block at the specimen scale and the pore scale.

[0012] A method for regional chain three-dimensional reconstruction of a seepage model at the engineering scale of a rock reservoir according to the present invention, the step of generating chain structure blocks based on the dual-channel feature fusion model by the trained generative adversarial network specifically includes: Adopt a probability function based on the Weibull function, use the dual-channel feature fusion model as the input of the generative adversarial network, and repeatedly generate multiple chain structure blocks based on the dual-channel feature fusion model.

[0013] The present invention also provides a device for regional chain three-dimensional reconstruction of a seepage model at the engineering scale of a rock reservoir, including: An acquisition module for respectively acquiring a pore-scale seepage feature channel map and a specimen-scale seepage feature channel map of a target area rock specimen; An extraction module for performing feature extraction and fusion based on a multi-scale attention mechanism on the pore-scale seepage feature channel map and the specimen-scale seepage feature channel map to obtain a dual-channel feature fusion model; A generation module for constructing and training a regional chain generative adversarial network on the dual-channel feature fusion model, and generating chain structure blocks based on the dual-channel feature fusion model by the trained generative adversarial network; A reconstruction module for obtaining a three-dimensional seepage model of a rock reservoir at the engineering scale based on the generated multiple chain structure blocks when the seepage channel feature matching degree of the generated multiple chain structure blocks at the specimen scale and the pore scale is greater than a preset matching degree threshold.

[0014] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, and when the processor executes the program, it implements the method for regional chain three-dimensional reconstruction of a seepage model at the engineering scale of a rock reservoir as described in any one of the above.

[0015] The present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the method for regional chain three-dimensional reconstruction of the rock reservoir engineering scale seepage model as described in any one of the above.

[0016] The present invention also provides a computer program product, including a computer program. When the computer program is executed by a processor, it implements the method for regional chain three-dimensional reconstruction of the rock reservoir engineering scale seepage model as described in any one of the above.

[0017] The method and device for regional chain three-dimensional reconstruction of the rock reservoir engineering scale seepage model provided by the present invention break through the scale limitation of the rock reservoir seepage model by constructing and training a regional chain generative adversarial network on a dual-channel feature fusion model, establish a three-dimensional model of the seepage channel with a scale similar to the rock reservoir and the engineering scale, and realize a method for regional chain three-dimensional reconstruction of the rock reservoir engineering scale seepage model with higher accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0019] Figure 1 is a schematic flow chart of the method for regional chain three-dimensional reconstruction of the rock reservoir engineering scale seepage model provided by the present invention; Figure 2 In (a) is a schematic diagram of the three-dimensional pore-scale pore-throat network structure in the method for regional chain three-dimensional reconstruction of the rock reservoir engineering scale seepage model provided by the present invention; Figure 2 In (b) is the pore-scale three-dimensional dominant seepage channel and isolated pore model in the method for regional chain three-dimensional reconstruction of the rock reservoir engineering scale seepage model provided by the present invention; Figure 2 In (c) is the specimen-scale three-dimensional dominant-inferior seepage channel model in the method for regional chain three-dimensional reconstruction of the rock reservoir engineering scale seepage model provided by the present invention; Figure 2 In (d) is the engineering-scale rock seepage channel model in the method for regional chain three-dimensional reconstruction of the rock reservoir engineering scale seepage model provided by the present invention; Figure 3 In (a) is a schematic diagram of the two-dimensional pore-scale pore-throat network structure in the method for regional chain three-dimensional reconstruction of the rock reservoir engineering scale seepage model provided by the present invention; Figure 3 In (b) is the pore-scale two-dimensional dominant seepage channel and isolated pore model in the regional chain three-dimensional reconstruction method of the rock reservoir engineering scale seepage model provided by the present invention; Figure 3 In (c) is the specimen-scale two-dimensional dominant-inferior seepage channel model in the regional chain three-dimensional reconstruction method of the rock reservoir engineering scale seepage model provided by the present invention; Figure 4 is the structural schematic diagram of the regional chain three-dimensional reconstruction device of the rock reservoir engineering scale seepage model provided by the present invention; Figure 5 is the structural schematic diagram of the electronic device provided by the present invention. Specific Embodiments

[0020] To make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions in the present invention will be clearly and completely described below with reference to the accompanying drawings in the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present invention without creative efforts shall fall within the protection scope of the present invention.

[0021] The following will be combined with Figures 1 to 3 to introduce the regional chain three-dimensional reconstruction method of the rock reservoir engineering scale seepage model of the present invention. As Figure 1 shown, it includes: Step 101, respectively obtain the pore-scale seepage characteristic channel map and the specimen-scale seepage characteristic channel map of the rock specimen in the target area; The target area is the area that needs to conduct geological research such as the geological reserves and exploitation volume of rock reservoirs. According to the geological exploration data and research requirements of the target area, drill cores in the target area and select several drill cores that can represent the formation characteristics of the target area.

[0022] Process the obtained several drill cores into rock specimens with conventional standard sizes under true triaxial conditions. Optionally, in this embodiment, the rock specimens are processed into those with a length, width and height of 100 mm.

[0023] On this basis, use an X-uCT scanning device (X-ray computed tomography device) to perform high-resolution imaging on the pore-scale rock seepage process of the rock specimen according to the set resolution to obtain the pore-scale seepage characteristic channel map. In this embodiment, the imaging resolution of the X-uCT scanning device is 2 μm .

[0024] Use an NMR imaging device (Nuclear Magnetic Resonance Imaging) to scan and image the seepage process of a rock specimen at a set resolution to obtain a seepage characteristic channel map at the specimen scale. In this embodiment, the imaging resolution of the NMR imaging device is 1 mm.

[0025] It can be understood that both the seepage characteristic channel map at the pore scale and the seepage characteristic channel map at the specimen scale are three-dimensional images characterized by an image sequence composed of multiple two-dimensional images of the rock specimen, and can be used to characterize the seepage channel characteristics of the formation to which the rock specimen belongs.

[0026] Step 102, perform feature extraction and fusion based on a multi-scale attention mechanism on the seepage characteristic channel map at the pore scale and the seepage characteristic channel map at the specimen scale to obtain a dual-channel feature fusion model; Optionally, perform feature extraction on the seepage characteristic channel map at the pore scale to obtain the seepage channel characteristics of the rock specimen at the pore scale.

[0027] Optionally, perform feature extraction on the seepage characteristic channel map at the specimen scale to obtain the seepage channel characteristics of the rock specimen at the specimen scale.

[0028] Perform feature fusion on the seepage channel characteristics of the rock specimen at the pore scale and the specimen scale to obtain a dual-channel feature fusion model, enabling it to simultaneously characterize the seepage channel characteristics of the formation to which the rock specimen belongs at the pore scale and the specimen scale.

[0029] Optionally, use the multi-scale attention mechanism for feature extraction. Among them, multi-scale representation obtains the seepage channel characteristics of the rock at the specimen scale and the pore scale respectively, and the attention mechanism is reflected in the multi-scale feature fusion process. Based on the importance of the features, the importance of the seepage channel characteristics at the specimen scale and the seepage channel characteristics at the pore scale in the dual-channel feature fusion model is automatically adjusted.

[0030] Step 103, construct and train a regional chained generative adversarial network on the dual-channel feature fusion model, and generate a chained structural block based on the dual-channel feature fusion model through the trained generative adversarial network; Construct a data set based on the dual-channel feature fusion model, construct and train a generative adversarial network on this data set, so that the trained generative adversarial network can be used to generate a chained structural block, and the generated chained structural block has seepage channel characteristics similar to those of the dual-channel feature fusion model, that is, it has seepage channel characteristics similar to those of the rock specimen at both the specimen scale and the pore scale.

[0031] On this basis, the trained generative adversarial network can be used to take the dual-channel feature fusion model as the input, and generate multiple chain structure blocks with similar seepage channel characteristics to it at both the specimen scale and the pore scale, which serve as the basic units of the three-dimensional model at the engineering scale.

[0032] Step 104, when the matching degree of the seepage channel characteristics of the generated multiple chain structure blocks at the specimen scale and the pore scale is not less than the preset matching degree threshold, an engineering-scale three-dimensional seepage model of the rock reservoir is obtained based on the generated multiple chain structure blocks.

[0033] Taking the generated chain structure blocks as the basic units as the input, the regional chain three-dimensional reconstruction technology is used to propose a seepage model with a similar scale to the target area rock reservoir, that is, to establish an engineering-scale rock seepage model.

[0034] Specifically, multiple chain structure blocks are repeatedly generated. When the seepage channel characteristics represented by the multiple chain structure blocks are similar to the real seepage channel characteristics of the target area represented by the rock specimen, it is considered that the three-dimensional model composed of the multiple chain structure blocks has seepage channel characteristics similar to the target area. Based on the multiple chain structure blocks at this time, the three-dimensional seepage model of the rock reservoir in the target area at the engineering scale can be rendered.

[0035] In a specific embodiment, as Figure 2 shown, different formation depths of the engineering-scale model of the rock reservoir are set. In this embodiment, four types of formations are set, each with a thickness of 5 meters. Multiple chain structure blocks are generated for each layer of the model to obtain a preliminary engineering-scale model. The seepage channel characteristics of the obtained engineering-scale model are matched with the seepage channel characteristics of the rock specimen, and their characteristic matching degree is calculated. When the characteristic matching degree is greater than the preset matching degree threshold, it is considered that the seepage channel characteristics of the constructed preliminary engineering-scale model are similar to the seepage channel characteristics of the target area.

[0036] On this basis, in the order of the formations (the digital labels from top to bottom are No.1, No.2, No.3, No.4) the engineering-scale seepage model of the rock reservoir is output as a sequence of 2D matrix image data in the format of ".jpg", where 100 images are output for each formation of the rock reservoir, and each Figure 2 2D matrix image uses the value of "NaN" to represent the non-rock reservoir area. Based on the obtained sequence of 2D matrix image data of the engineering-scale seepage model of the rock reservoir, the Slice function in the MATLAB software is used for visual display to obtain the three-dimensional seepage model of the rock reservoir.

[0037] The present invention constructs and trains a regional chain generative adversarial network on a dual-channel feature fusion model, breaks through the scale limitation of the seepage model of rock reservoirs, establishes a three-dimensional model of seepage channels with similar model scales and engineering scales of rock reservoirs, and realizes a regional chain three-dimensional reconstruction method for the seepage model of rock reservoirs at the engineering scale with higher accuracy.

[0038] In the regional chain three-dimensional reconstruction method for the seepage model of rock reservoirs at the engineering scale of the present invention, the step of performing feature extraction and fusion based on a multi-scale attention mechanism on the seepage feature channel map at the pore scale and the seepage feature channel map at the specimen scale to obtain a dual-channel feature fusion model specifically includes: Perform local feature extraction and global feature extraction on the seepage feature channel map at the pore scale and the seepage feature channel map at the specimen scale respectively; Perform feature fusion of local features and global features on the seepage feature channel map at the pore scale and the seepage feature channel map at the specimen scale respectively to obtain the fused feature at the pore scale and the fused feature at the specimen scale; Perform feature fusion on the fused feature at the pore scale and the fused feature at the specimen scale to obtain a dual-channel feature fusion model.

[0039] In order to accurately obtain the seepage channel characteristics of rock specimens, in this embodiment, a fine fusion model (Multiscale-FCF) based on a multi-scale feature fusion mechanism is constructed.

[0040] Specifically, point convolution at the pore scale and specimen scale and weighted image matrix generalization operations are used to obtain local and global feature data of the seepage channels of rock reservoirs, and their corresponding expressions are as follows respectively: ; In the formula, X i represents the input image, represents the local point convolution function, represents the global point convolution function, is BathNorm function, is a 1×1 convolution, is ReLU activation function, is the average pooling operation function.

[0041] By the above method, taking the seepage feature channel map at the pore scale and the seepage feature channel map at the specimen scale as inputs respectively, the local features and global features of each map can be extracted respectively.

[0042] On this basis, first perform feature fusion on the local features and global features of the seepage feature channel map at the pore scale and the seepage feature channel map at the specimen scale respectively: ; In the formula, represents an image matrix scaling function for scaling an image to the same scale, is the symbol for multiplying matrix elements, represents a scaling function.

[0043] It can be understood that for the pore-scale seepage characteristic channel map / specimen-scale seepage characteristic channel map, both are three-dimensional images composed of multiple two-dimensional image sequences. Therefore, X in the formula represents the feature fusion result of the previous image, characterizes the feature fusion result of the current image. After performing feature fusion on the first image in the two-dimensional image sequence respectively through the above formula, and then iteratively updating it with the feature fusion result of the next image, the finally obtained is the fusion result of the local features and global features of the three-dimensional image, that is, the fusion features at the pore scale and the fusion features at the specimen scale.

[0044] On this basis, it is also necessary to perform feature fusion on the fusion features at the pore scale and the fusion features at the specimen scale to obtain a dual-channel feature fusion model:

[0045] In the formula, is a feature fusion function, represents the fusion feature at the pore scale, represents the fusion feature at the specimen scale, is a sigmoid activation function (with a value range of 0 - 1, automatically taking values during the training process), is the symbol for adding matrix elements.

[0046] Through the above method, the sigmoid activation function is used to achieve feature fusion guided by the attention mechanism, and the fusion features at the pore scale and the fusion features at the specimen scale are fused into dual-channel features for use as the training basis of the generative adversarial network.

[0047] In the method for regional chain three-dimensional reconstruction of the seepage model at the engineering scale of rock reservoirs of the present invention, before the step of constructing and training a regional chain generative adversarial network on the dual-channel feature fusion model, it further includes: Performing digital labeling on the dual-channel feature fusion model, and the digital labeling is used to label the connected seepage channels and non-connected seepage channels of the dual-channel feature fusion model.

[0048] Further, in order to better divide the seepage dominant and inferior channels of the engineering-scale rock reservoir by the generated chain-structured blocks, in this embodiment, digital labels are assigned to the connected seepage channels and the unconnected seepage channels in the dual-channel feature fusion model to achieve fine digital characterization.

[0049] Among them, the connected seepage channels represent the seepage dominant channels, and the unconnected seepage channels represent the seepage inferior channels.

[0050] Optionally, the 8-neighborhood convolution operator shown in the following formula is used to extract the seepage channel connectivity and non-connectivity features of the dual-channel feature fusion model, and digital labels are assigned to them in sequence: ; In the formula, represents the 8-neighborhood convolution operator, is the digital characterization function of the seepage channel connectivity and non-connectivity features, i is the serial number of the connected phase, p is the number of connected phases, j is the serial number of the unconnected phase and the serial number starts from p +1, q is the number of unconnected phases.

[0051] Through the above method, a dual-channel feature fusion model with digital characterization is obtained. As shown in Figure 3 , Figure 3 in which (a) is a two-dimensional pore-scale pore-throat network structure diagram, Figure 3 in which (b) is a pore-scale dominant seepage channel and isolated pore model with digital characterization, Figure 3 in which (c) is a digital characterization of the specimen-scale dominant-inferior seepage channel model.

[0052] In the method for regional chain three-dimensional reconstruction of the seepage model in the rock reservoir engineering scale of the present invention, the step of constructing and training a regional chain generative adversarial network on the dual-channel feature fusion model specifically includes: Taking the dual-channel feature fusion model as the input of the generative adversarial network for the generative block channel region chain; Constructing the loss function of the generator of the generative adversarial network based on the generation and content losses; Constructing the loss function of the discriminator of the generative adversarial network based on the mean square error of the probability density functions of the input features and the fusion features.

[0053] Optionally, in this embodiment, taking the dual-channel feature fusion model with digital characterization as the input parameter of the block channel region chain, a multi-channel feature generation-adversarial network (Multi-Channel GAN) is established as the constructed generative adversarial network to realize the generation of chain-structured blocks.

[0054] Specifically, both the generator and discriminator of the Multi-Channel GAN adopt a dual-channel network structure design for processing at the specimen scale and pore scale. Among them, the loss function of the generator consists of a generation loss and a content loss, and the loss function of the discriminator consists of the mean square error of the probability density functions of the input features and the fused features, and its expression is as follows: ; ; In the formula, represents the generator of the Multi-Channel GAN, represents the discriminator of the Multi-Channel GAN, is the weighted average coefficient (the value range is [0, 1]); is the predicted feature fusion image, is the real feature fusion image.

[0055] represents the loss function of the generator, represents the generation loss, represents the content loss; represents the loss function of the discriminator, represents the input loss, represents the feature fusion loss; is the probability density function, is the image feature input function, is the weighted average coefficient (the value range is [0, 1]).

[0056] The Multi-Channel GAN can be constructed in the above manner. On this basis, by adjusting the hyperparameters of the number of training times (100 times), learning rate (0.95), and the number of model layers (32), the optimized Multi-Channel GAN model is obtained as the trained generative adversarial network.

[0057] In the method for regional chain three-dimensional reconstruction of the seepage model at the engineering scale of the rock reservoir of the present invention, it further includes: Obtaining the seepage channel generation characteristics of the generated multiple chain structure blocks at the specimen scale and pore scale; Calculating the first error between the seepage channel generation characteristics at the specimen scale and the seepage channel characteristics; Calculating the second error between the seepage channel generation characteristics at the pore scale and the seepage channel characteristics; Taking the linear weighted result of the first error and the second error as the seepage channel feature matching degree of the chain structure block at the specimen scale and pore scale.

[0058] Optionally, in this embodiment, a probability function based on the Weibull function is used to generate a chain structure block with the fusion model of the rock seepage channel characteristics at the pore scale and the specimen scale as the basic unit. Taking the chain structure block basic unit as the input, the regional chain 3D reconstruction technology is used to establish a similar-scale seepage model of the rock reservoir, and its seepage channel characteristics at the specimen scale and the pore scale are calculated as the seepage channel generation characteristics of the generated multiple chain structure blocks at the specimen scale and the pore scale.

[0059] Furthermore, for the generated model, it is also necessary to determine the matching degree between its seepage channel characteristics and the true seepage channel characteristics of the target area characterized by the rock specimen. For this purpose, in this embodiment, the seepage channel characteristic matching degree is calculated by the following method PCMR :

[0060] In the formula, represents the predicted value of the pore scale of the rock reservoir, represents the true value of the seepage channel characteristics of the pore scale of the rock reservoir; represents the predicted value of the specimen scale of the rock reservoir, represents the true value of the seepage channel characteristics of the specimen scale of the rock reservoir; is the weighted average coefficient.

[0061] That is to say, calculate the difference between the predicted value and the true value of the pore scale of the rock reservoir, and take the ratio of it to the true value of the pore scale of the rock reservoir as the first error; calculate the difference between the predicted value and the true value of the specimen scale of the rock reservoir, and take the ratio of it to the true value of the specimen scale of the rock reservoir as the second error. Take the linear weighted result of the first error and the second error as the characteristic matching degree PCMR .

[0062] Corresponding to the above PCMR calculation method, in this embodiment, the preset matching degree threshold is defined as 90%, that is, when the PCMR of the three-dimensional model composed of the generated multiple chain structure blocks is not less than 90%, it is considered that the seepage channel characteristics of the three-dimensional model composed of the generated multiple chain structure blocks are similar to the seepage channel characteristics of the target area. Therefore, the three-dimensional seepage model of the rock reservoir at the engineering scale of the target area can be rendered.

[0063] In addition, if the PCMRIf it is less than 90%, continue to optimize and train the Multi-Channel GAN model, and use the retrained Multi-Channel GAN model to generate chain structure blocks until the three-dimensional model composed of multiple generated chain structure blocks PCMR is not less than 90%.

[0064] In the method for regional chain three-dimensional reconstruction of the seepage model at the engineering scale of a rock reservoir according to the present invention, the step of generating chain structure blocks based on the dual-channel feature fusion model by the trained generative adversarial network specifically includes: Adopt a probability function based on the Weibull function, use the dual-channel feature fusion model as the input of the generative adversarial network, and repeatedly generate multiple chain structure blocks based on the dual-channel feature fusion model.

[0065] In this embodiment, a probability function based on the Weibull function is used as the probability density function to realize the generation of multiple chain structure blocks for each formation. The regional chain structure unit characteristics distribution of the seepage channel is established through the Weibull function, so as to better simulate the seepage channel characteristics at the engineering scale.

[0066] Next, a device for regional chain three-dimensional reconstruction of the seepage model at the engineering scale of a rock reservoir provided by the present invention will be described. The device for regional chain three-dimensional reconstruction of the seepage model at the engineering scale of a rock reservoir described below can be mutually corresponding and referred to the method for regional chain three-dimensional reconstruction of the seepage model at the engineering scale of a rock reservoir described above.

[0067] As Figure 4 shown, the device for regional chain three-dimensional reconstruction of the seepage model at the engineering scale of a rock reservoir includes an acquisition module 401, an extraction module 402, a generation module 403, and a reconstruction module 404; The acquisition module 401 is configured to respectively acquire a pore-scale seepage feature channel map and a specimen-scale seepage feature channel map of a rock specimen in a target area; The target area is an area where geological research such as the resource reserves and extraction volume of a rock reservoir needs to be carried out. According to the geological exploration data and research requirements of the target area, core drilling is carried out in the target area, and several drilling cores capable of characterizing the formation characteristics of the target area are selected.

[0068] The several obtained drilling cores are processed and prepared into rock specimens with conventional standard sizes under true triaxial conditions. Optionally, in this embodiment, the rock specimens are processed into rock specimens with a length, width, and height of 100 mm.

[0069] On this basis, an X-uCT scanning device (X-ray computed tomography device) is used to perform high-resolution imaging on the rock seepage process at the pore scale of the rock specimen according to the set resolution, and a seepage characteristic channel map at the pore scale is obtained. In this embodiment, the imaging resolution of the X-uCT scanning device is 2 μm .

[0070] An NMR imaging device (Nuclear Magnetic Resonance Imaging) is used to perform scanning imaging on the seepage process of the rock specimen according to the set resolution, and a seepage characteristic channel map at the specimen scale is obtained. In this embodiment, the imaging resolution of the NMR imaging device is 1 mm.

[0071] It can be understood that both the seepage characteristic channel map at the pore scale and the seepage characteristic channel map at the specimen scale are three-dimensional images characterized by an image sequence composed of multiple two-dimensional images of the rock specimen, and can be used to characterize the seepage channel characteristics of the formation to which the rock specimen belongs.

[0072] An extraction module 402 is used to perform feature extraction and fusion on the seepage characteristic channel map at the pore scale and the seepage characteristic channel map at the specimen scale based on a multi-scale attention mechanism to obtain a dual-channel feature fusion model; Optionally, feature extraction is performed on the seepage characteristic channel map at the pore scale to obtain the seepage channel characteristics of the rock specimen at the pore scale.

[0073] Optionally, feature extraction is performed on the seepage characteristic channel map at the specimen scale to obtain the seepage channel characteristics of the rock specimen at the specimen scale.

[0074] Feature fusion is performed on the seepage channel characteristics of the rock specimen at the pore scale and the specimen scale to obtain a dual-channel feature fusion model, enabling it to simultaneously characterize the seepage channel characteristics of the formation to which the rock specimen belongs at the pore scale and the specimen scale.

[0075] Optionally, the multi-scale attention mechanism is used for feature extraction. Among them, multi-scale representations obtain the rock seepage channel characteristics at the specimen scale and the pore scale respectively, and the attention mechanism is reflected in the multi-scale feature fusion process. Based on the importance of the features, the importance of the seepage channel characteristics at the specimen scale and the seepage channel characteristics at the pore scale in the dual-channel feature fusion model is automatically adjusted.

[0076] A generation module 403 is used to construct and train a regional chained generative adversarial network on the dual-channel feature fusion model, and generate a chained structural block based on the dual-channel feature fusion model through the trained generative adversarial network; Construct a dataset based on a dual-channel feature fusion model, construct and train a generative adversarial network on this dataset, so that the trained generative adversarial network can be used to generate chain structure blocks, and the generated chain structure blocks have seepage channel features similar to those of the dual-channel feature fusion model, that is, they have seepage channel features similar to those of rock specimens at both the specimen scale and the pore scale.

[0077] On this basis, the trained generative adversarial network can be used with the dual-channel feature fusion model as the input to generate multiple chain structure blocks that have similar seepage channel features to it at both the specimen scale and the pore scale, as the basic units of the three-dimensional model at the engineering scale.

[0078] The reconstruction module 404 is used to obtain a three-dimensional seepage model of the rock reservoir at the engineering scale based on the generated multiple chain structure blocks when the matching degree of the seepage channel features of the generated multiple chain structure blocks at the specimen scale and the pore scale is greater than a preset matching degree threshold.

[0079] Taking the generated chain structure blocks as the basic units as the input, adopt the regional chain three-dimensional reconstruction technology to propose a seepage model with a similar scale to the target area rock reservoir, that is, establish a rock seepage model at the engineering scale.

[0080] Specifically, repeatedly generate multiple chain structure blocks. When the seepage channel features represented by the multiple chain structure blocks are similar to the true seepage channel features of the target area represented by the rock specimen, it is considered that the three-dimensional model composed of the multiple chain structure blocks has seepage channel features similar to those of the target area. Based on the multiple chain structure blocks at this time, a three-dimensional seepage model of the rock reservoir at the engineering scale of the target area can be rendered.

[0081] In a specific embodiment, as Figure 2 shown, set different formation depths of the rock reservoir engineering scale model. In this embodiment, four types of formations are set, each layer has a thickness of 5 meters. Generate multiple chain structure blocks for each layer of the model to obtain a preliminary engineering scale model. Match the seepage channel features of the obtained engineering scale model with the seepage channel features of the rock specimen, and calculate the feature matching degree. When the feature matching degree is greater than the preset matching degree threshold, it is considered that the seepage channel features of the constructed preliminary engineering scale model are similar to the seepage channel features of the target area.

[0082] On this basis, in the order of the formations (the digital labels from top to bottom are No.1, No.2, No.3, No.4) output the three-dimensional seepage model of the rock reservoir at the engineering scale as a sequence of 2D matrix images in the format of ".jpg", where 100 images are output for each formation of the rock reservoir, and each Figure 2In the D matrix image, the "NaN" value is used to represent the non-rock reservoir area. Based on the acquired sequence of 2D matrix image data of the rock reservoir engineering-scale seepage model, the Slice function in MATLAB software is used for visual display to obtain the three-dimensional seepage model of the rock reservoir.

[0083] In the present invention, by constructing and training a regional chain generative adversarial network on a dual-channel feature fusion model, the scale limitation of the rock reservoir seepage model is broken through, a three-dimensional model similar to the model scale and engineering-scale seepage channels of the rock reservoir is established, and a method for regional chain three-dimensional reconstruction of the rock reservoir engineering-scale seepage model with higher accuracy is realized.

[0084] Figure 5 An example of the physical structure diagram of an electronic device is shown as Figure 5 As shown, the electronic device may include: a processor 510, a communication interface 520, a memory 530, and a communication bus 540. Among them, the processor 510, the communication interface 520, and the memory 530 complete mutual communication through the communication bus 540. The processor 510 can call the logical instructions in the memory 530 to execute the method for regional chain three-dimensional reconstruction of the rock reservoir engineering-scale seepage model. The method includes: respectively obtaining the pore-scale seepage feature channel map and the specimen-scale seepage feature channel map of the rock specimen in the target area; performing feature extraction and fusion based on a multi-scale attention mechanism on the pore-scale seepage feature channel map and the specimen-scale seepage feature channel map to obtain a dual-channel feature fusion model; constructing and training a regional chain generative adversarial network on the dual-channel feature fusion model, and generating a chain structure block based on the dual-channel feature fusion model through the trained generative adversarial network; when the matching degree of the seepage channel features between the generated multiple chain structure blocks at the specimen scale and the pore scale is not less than a preset matching degree threshold, obtaining a three-dimensional seepage model of the rock reservoir at the engineering scale based on the generated multiple chain structure blocks.

[0085] In addition, when the logical instructions in the aforementioned memory 530 are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs.

[0086] On the other hand, the present invention also provides a computer program product. The computer program product includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the method for regional chain three-dimensional reconstruction of the seepage model at the engineering scale of a rock reservoir provided by the above-mentioned various methods. The method includes: respectively obtaining the seepage feature channel map at the pore scale and the seepage feature channel map at the specimen scale of the rock specimen in the target area; performing feature extraction and fusion based on a multi-scale attention mechanism on the seepage feature channel map at the pore scale and the seepage feature channel map at the specimen scale to obtain a dual-channel feature fusion model; constructing and training a regional chain generative adversarial network on the dual-channel feature fusion model, and generating a chain structure block based on the dual-channel feature fusion model through the trained generative adversarial network; when the matching degree of the seepage channel features between the generated multiple chain structure blocks at the specimen scale and the pore scale is not less than a preset matching degree threshold, obtaining a three-dimensional seepage model of the rock reservoir at the engineering scale based on the generated multiple chain structure blocks.

[0087] In another aspect, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the method for region chain-based three-dimensional reconstruction of the seepage model at the engineering scale of a rock reservoir provided by the above-mentioned various methods. The method includes: respectively obtaining the pore-scale seepage characteristic channel map and the specimen-scale seepage characteristic channel map of the rock specimen in the target area; performing feature extraction and fusion based on a multi-scale attention mechanism on the pore-scale seepage characteristic channel map and the specimen-scale seepage characteristic channel map to obtain a dual-channel feature fusion model; constructing and training a region chain-based generative adversarial network on the dual-channel feature fusion model, and generating a chain structure block based on the dual-channel feature fusion model through the trained generative adversarial network; when the matching degree of the seepage channel features between the generated multiple chain structure blocks at the specimen scale and the pore scale is not less than a preset matching degree threshold, obtaining a three-dimensional seepage model of the rock reservoir at the engineering scale based on the generated multiple chain structure blocks.

[0088] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without creative labor.

[0089] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0090] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A regional chain three-dimensional reconstruction method for a rock reservoir engineering scale seepage model, characterized in that: include: Obtaining the pore-scale seepage characteristic channel map and the specimen-scale seepage characteristic channel map of the rock specimen in the target area respectively; Performing feature extraction and fusion based on a multi-scale attention mechanism on the pore-scale seepage characteristic channel map and the specimen-scale seepage characteristic channel map to obtain a dual-channel feature fusion model; On the dual-channel feature fusion model, construct and train a regional chain-type generative adversarial network, and generate a chain structure block with the dual-channel feature fusion model as a basic unit through the trained generative adversarial network; When the matching degree of the seepage channel characteristics at the specimen scale and the pore scale of the generated multiple chain structure blocks is not less than a preset matching degree threshold, a three-dimensional seepage model of the rock reservoir at an engineering scale is obtained based on the generated multiple chain structure blocks.

2. The method for regional chain three-dimensional reconstruction of rock reservoir engineering scale seepage model according to claim 1, characterized in that: The step of extracting and fusing the pore-scale seepage characteristic channel map and the specimen-scale seepage characteristic channel map based on a multi-scale attention mechanism to obtain a dual-channel feature fusion model specifically includes: Performing local feature extraction and global feature extraction on the pore-scale seepage characteristic channel map and the specimen-scale seepage characteristic channel map respectively; The local features and global features of the pore-scale seepage characteristic channel map and the specimen-scale seepage characteristic channel map are respectively fused to obtain pore-scale fusion features and specimen-scale fusion features; The pore-scale fusion features and the specimen-scale fusion features are fused to obtain a dual-channel feature fusion model.

3. The method for regional chain three-dimensional reconstruction of rock reservoir engineering scale seepage model according to claim 2, characterized in that: Before the step of constructing and training a regional chain-type generative adversarial network on the dual-channel feature fusion model, the method further includes: The dual-channel feature fusion model is digitally labeled, and the digital labels are used to mark the connected seepage channels and the non-connected seepage channels of the dual-channel feature fusion model.

4. The method for regional chain three-dimensional reconstruction of a rock reservoir engineering scale seepage model according to any one of claims 1 to 3, characterized in that: The step of constructing and training a regional chain-type generative adversarial network on the dual-channel feature fusion model specifically includes: Using the dual-channel feature fusion model as input to the generator block channel region chain of the generative adversarial network; Constructing a loss function of the generator of the generative adversarial network based on generation and content losses; The loss function of the discriminator of the generative adversarial network is constructed based on the mean square error of the probability density function of the input features and the fused features.

5. The method for regional chain three-dimensional reconstruction of a rock reservoir engineering scale seepage model according to any one of claims 1 to 3, characterized in that: Also includes: Obtaining the seepage channel generation characteristics of the generated plurality of chain structure blocks at the specimen scale and the pore scale; Calculating the first error between the seepage channel generation characteristics and the seepage channel characteristics at the specimen scale; Calculating the second error between the seepage channel generation characteristics and the seepage channel characteristics at the pore scale; The linear weighted result of the first error and the second error is used as the matching degree of the seepage channel characteristics of the chain structure block at the specimen scale and the pore scale.

6. The method for regional chain three-dimensional reconstruction of a rock reservoir engineering scale seepage model according to any one of claims 1 to 3, characterized in that: The step of generating a chain structure block based on the dual-channel feature fusion model by the training-completed generative adversarial network specifically includes: A probability function based on the Weibull function is adopted, the dual-channel feature fusion model is used as the input of the generative adversarial network, and a plurality of chain structure blocks with the dual-channel feature fusion model as the basic unit are repeatedly generated.

7. A regional chain three-dimensional reconstruction device for a rock reservoir engineering scale seepage model, characterized in that: include: An acquisition module, used to respectively acquire a pore-scale seepage characteristic channel map and a specimen-scale seepage characteristic channel map of a rock specimen in a target area; An extraction module is used to extract and fuse the pore-scale seepage characteristic channel map and the specimen-scale seepage characteristic channel map based on a multi-scale attention mechanism to obtain a dual-channel feature fusion model; A generation module, used to construct and train a regional chain-type generative adversarial network on the dual-channel feature fusion model, and generate a chain structure block with the dual-channel feature fusion model as a basic unit through the trained generative adversarial network; A reconstruction module is used to obtain a three-dimensional seepage model of a rock reservoir at an engineering scale based on the generated multiple chain structure blocks when the matching degree of seepage channel characteristics at the specimen scale and the pore scale is greater than a preset matching degree threshold.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method for regional chain three-dimensional reconstruction of the rock reservoir engineering scale seepage model as claimed in any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for regional chain three-dimensional reconstruction of a rock reservoir engineering-scale seepage model as claimed in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for regional chain three-dimensional reconstruction of a rock reservoir engineering-scale seepage model as claimed in any one of claims 1 to 6 is implemented.

Images