Knowledge-driven multi-modal single-well geology intelligent interpretation method, device and storage medium
By acquiring and processing multimodal well data, generating neighboring well similarity indices and feature information links, and training a generative adversarial model, the problem of insufficient information utilization in well logging geological interpretation is solved, and higher-precision multimodal geological interpretation is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHINA UNIV OF PETROLEUM (BEIJING)
- Filing Date
- 2024-06-07
- Publication Date
- 2026-05-15
AI Technical Summary
Existing well logging geological interpretation methods do not fully utilize the correlation between prior information and well logging information, resulting in low interpretation accuracy, especially in the later stages of oil and gas reservoir development, where they cannot meet actual needs.
By acquiring historical multimodal well-to-well data and single-well logging data of the target area, a set of neighboring well similarity indices is generated. Combined with feature information data and noise data, a target generative adversarial model is trained to generate multimodal geological interpretation data.
It improves the accuracy and comprehensiveness of well logging geological interpretation, makes full use of multi-source heterogeneous information, and generates multi-modal geological interpretation data with diverse types and high correlation.
Smart Images
Figure CN118606724B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of geological exploration technology, specifically to a knowledge-driven multimodal single-well geological intelligent interpretation method, equipment, and storage medium. Background Technology
[0002] Well logging interpretation, also known as integrated well logging interpretation, is essentially about determining the relationship between well logging information and geological information. It involves using the correct methods to process well logging information into geological information, which can provide effective geological basis for subsequent oil and gas development.
[0003] In the process of realizing this invention, the inventors discovered that prior information such as geological tectonic background, sedimentary background, previous geological modeling results, previous seismic inversion results, and fault interpretation can fully reflect various geological information. However, existing well logging geological interpretation usually only considers the well logging information of individual oil wells without analyzing related prior information, resulting in insufficient utilization of prior information, wasting information resources, and leading to low accuracy of well logging geological interpretation. In addition, because the current well logging geological interpretation is relatively singular in dimension (for example, interpreting lithology first, then porosity, then permeability, and finally hydrocarbon saturation), and does not consider the correlation of various types of well logging geological information, the accuracy of well logging geological interpretation is further reduced, especially in the middle and late stages of oil and gas reservoir development, where the accuracy of well logging geological interpretation often fails to meet the requirements of actual development. Summary of the Invention
[0004] The purpose of this invention is to provide a knowledge-driven multimodal single-well geological intelligent interpretation method, device, and storage medium to alleviate the technical problem of low accuracy in current well logging geological interpretation.
[0005] To achieve the above objectives, one aspect of the present invention provides a knowledge-driven multimodal single-well geological intelligent interpretation method, comprising:
[0006] Acquire historical multimodal well-to-well data and single-well logging data within the target area;
[0007] Based on the historical multimodal well data and the single-well logging data, a set of neighboring well similarity indices is obtained;
[0008] Feature information data is generated based on the adjacent well similarity index set, the historical multimodal well data, and the single-well logging data;
[0009] Based on the feature information data and noise data, a feature information link is generated;
[0010] The target generation adversarial model is trained based on the aforementioned feature information links, so that the trained target generation adversarial model can generate multimodal geological interpretation data corresponding to the target area.
[0011] Optionally, the step of obtaining the neighboring well similarity index set based on the historical multimodal well data and the single-well logging data includes:
[0012] Based on the single-well logging data, determine the waveform similarity of the logging curves of any two oil wells;
[0013] The sedimentary microfacies coefficient, the waveform similarity of the well logging curve, the fault well spacing, and the straight-line distance between wells are calculated and processed to obtain the adjacent well similarity index set. The historical multimodal well data includes the sedimentary microfacies coefficient, the fault well spacing, and the straight-line distance between wells.
[0014] Optionally, the step of generating feature information data based on the neighboring well similarity index set, the historical multimodal well data, and the single-well logging data includes:
[0015] Based on the neighboring well similarity index set and the historical multimodal well data, multimodal soft data is generated;
[0016] The multimodal soft data and the single-well logging data are combined and processed to obtain multimodal data;
[0017] The multimodal data is embedded to obtain feature information data, wherein the feature information data includes multiple feature information sequences.
[0018] Optionally, the step of embedding the multimodal data to obtain feature information data includes:
[0019] Geological feature label data is embedded into the multimodal data to obtain feature information data, wherein the geological feature label data includes at least one of porosity, permeability, fracture information, lithology information and oil saturation.
[0020] Optionally, the step of generating feature information links based on the feature information data and noise data includes:
[0021] The feature information sequences that meet the preset similarity conditions in the feature information data are linked to obtain sample feature links, and
[0022] The noise data is processed using a normal distribution to obtain multiple noise feature vectors;
[0023] The plurality of noise feature vectors are input into the target generative adversarial model, so that the generative model in the target generative adversarial model generates noise feature links based on the plurality of noise feature vectors, wherein the feature information links include the sample feature links and the noise feature links.
[0024] Optionally, the step of training the target generative adversarial model based on the feature information link, so that the trained target generative adversarial model generates multimodal geological interpretation data corresponding to the target area, includes:
[0025] The sample feature links and the noise feature links are input into the target generative adversarial model, so that the discriminant model in the target generative adversarial model outputs the feature link discrimination result;
[0026] The target generation adversarial model is trained based on the feature link discrimination results, so that the trained target generation adversarial model outputs multimodal geological interpretation data corresponding to the target area.
[0027] Optionally, the step of training the target generative adversarial model based on the feature link discrimination result, so that the trained target generative adversarial model outputs multimodal geological interpretation data corresponding to the target region, includes:
[0028] Based on the feature link discrimination results, the loss functions for the sample feature links and the noise feature links are determined;
[0029] The generative model and the discriminative model are trained according to the loss function so that the trained generative model generates multimodal geological interpretation data corresponding to the target area.
[0030] Optionally, the step of training the generative model and the discriminative model according to the loss function, so that the trained generative model generates multimodal geological interpretation data corresponding to the target area, includes:
[0031] The generative model and the discriminative model are trained according to the loss function, so that the trained discriminative model outputs a discrimination result that meets the preset discrimination conditions.
[0032] The trained generative model generates target data that satisfies the preset discrimination conditions, wherein the multimodal geological interpretation data includes the target data.
[0033] On the other hand, the present invention provides a knowledge-driven multimodal single-well geological intelligent interpretation device, configured to perform the knowledge-driven multimodal single-well geological intelligent interpretation method according to any one of the above claims.
[0034] Furthermore, another aspect of the present invention provides a machine-readable storage medium storing instructions that, when executed by a processor, cause the processor to implement the knowledge-driven multimodal single-well geological intelligent interpretation method according to any of the preceding claims.
[0035] The above technical solution combines historical multimodal well data obtained a priori, conventional well logging information, and neighboring well similarity index sets characterizing the correlation between adjacent oil wells to generate feature information data with multi-source heterogeneity. Feature information links are then generated based on the feature information data, and a target generative adversarial network is trained on these feature information links. This enables the target generative adversarial network to automatically generate diverse and highly correlated multimodal geological interpretation data. Therefore, using multimodal geological interpretation data as the basis for well logging geological interpretation can improve the comprehensiveness of well logging geological interpretation, thereby effectively improving the accuracy of well logging geological interpretation.
[0036] Other features and advantages of the embodiments of the present invention will be described in detail in the following detailed description section. Attached Figure Description
[0037] The accompanying drawings are provided to further illustrate embodiments of the present invention and form part of the specification. They are used together with the following detailed description to explain the embodiments of the present invention, but do not constitute a limitation thereof. In the drawings:
[0038] Figure 1 This is a flowchart illustrating the knowledge-driven multimodal single-well geological intelligent interpretation method provided in this embodiment of the invention.
[0039] Figure 2a This is a schematic diagram of the fault well spacing provided in an embodiment of the present invention.
[0040] Figure 2b This is a schematic diagram of the straight-line distance between wells provided in an embodiment of the present invention.
[0041] Figure 3 This is a schematic diagram of a scenario for the knowledge-driven multimodal single-well geological intelligent interpretation method provided in an embodiment of the present invention.
[0042] Figure 4 This is a schematic diagram of the structure of the knowledge-driven multimodal single-well geological intelligent interpretation device provided in an embodiment of the present invention. Detailed Implementation
[0043] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the scope of the present invention.
[0044] One aspect of this invention provides a knowledge-driven, multimodal, intelligent geological interpretation method for single wells. For example... Figure 1 As shown, Figure 1 This is a flowchart illustrating the knowledge-driven multimodal single-well geological intelligent interpretation method provided in this embodiment of the invention. The specific process can be as follows:
[0045] S101. Acquire historical multimodal well-to-well data and single-well logging data within the target area.
[0046] The target area is the area currently being studied in well logging geology. The historical multimodal inter-well data are multi-source heterogeneous data extracted from prior information such as tectonic background, sedimentary background, previous geological modeling results, previous seismic inversion results, and fault interpretation. The single-well logging data includes conventional logging parameters.
[0047] Specifically, the oil and gas reservoir data reflected by prior information is typical multimodal data (characterized by multi-source heterogeneity). For example, three-dimensional geological models and structural models are three-dimensional image data, wellbore imaging logging is two-dimensional image data, study area text reports are textual information, and sedimentary microfacies interpretation is two-dimensional / three-dimensional image data. This prior information is diversified and highly correlated. If this prior information can be fully utilized, it can comprehensively reflect various well logging geological interpretations of the study area, providing data support for subsequent oil and gas reservoir development. In addition, conventional logging parameters can reflect the conventional properties of rock formations and are indispensable data for subsequent well logging geological interpretation. For example, conventional logging curves can reflect different lithologies and stratigraphic characteristics, and specific lithologies and stratigraphic positions can be determined based on the curves.
[0048] S102. Based on historical multimodal well data and single-well logging data, obtain the set of neighboring well similarity indices.
[0049] Among them, the neighboring well similarity index set is used to characterize the degree of similarity of common information between any two oil wells in the target area. The neighboring well similarity index set contains multiple neighboring well similarity indices.
[0050] Optionally, in this embodiment of the invention, historical multimodal well data includes sedimentary microfacies coefficients (used to determine whether two wells are on the same sedimentary facies; if the sedimentary microfacies coefficient is 0, it indicates that the two wells are not on the same sedimentary facies; if the sedimentary microfacies coefficient is 1, it indicates that the two wells are on the same sedimentary facies), fault well spacing, and straight-line distance between wells within the target area. Single-well logging data includes logging curves (e.g., natural gamma curve, spontaneous potential curve, wellbore diameter curve, deep lateral resistivity curve, shallow lateral resistivity curve, compensated neutron curve, density curve, and compensated sonic curve). Specifically, firstly, the waveform similarity of logging curves between any two wells is determined based on the single-well logging data. Then, the sedimentary microfacies coefficients, logging curve waveform similarity, fault well spacing, and straight-line distance between wells are substituted into Formula 1 for calculation to obtain the adjacent well similarity index set I = [I1, I2, ..., I...]. n ].
[0051]
[0052] Where S represents the sedimentary microfacies coefficient, C represents the similarity of logging curve waveforms, and F represents the fault well spacing (i.e., the sum of the vertical distances from the two oil wells to the fault line, such as...). Figure 2a As shown, Figure 2a This is a schematic diagram of the fault well spacing provided in an embodiment of the present invention. The vertical distance from the first well 21 to the fault line 20 is f1, and the vertical distance from the second well 22 to the fault line 20 is f2. Therefore, the fault well spacing F = f1 + f2, and D represents the straight-line distance between wells (e.g., ...). Figure 2b As shown, Figure 2b This is a schematic diagram of the straight-line distance between wells provided in an embodiment of the present invention. The straight-line distance D between wells is the straight-line distance from the first well 21 to the second well 22.
[0053] Specifically, the formula for calculating the waveform similarity C of the well logging curve is Formula 2:
[0054]
[0055] Where x1 and x2 represent the values of the conventional logging curves of the two oil wells at the same depth, and n1 and n2 represent the total number of logging curves of the two oil wells used in the calculation.
[0056] S103. Based on the neighboring well similarity index set, historical multimodal well data, and single-well logging data, generate feature information data.
[0057] Specifically, in this embodiment of the invention, firstly, the adjacent well similarity index set I, sedimentary microfacies coefficient S, fault well spacing F, and inter-well straight-line distance D are composed of multimodal soft data. Then, the multimodal soft data and each well logging curve are composite processed to obtain a multi-attribute composite vector: multimodal data x i(i = 1, 2, ..., n, where n is the number of samples), and finally, for the multimodal data x i Embedding processing is performed to obtain feature information data composed of multiple feature information sequences. Specifically, the embedding process includes: embedding the geological feature label data y... i Multimodal data x is embedded using Multimodel Embedding algorithm. i The feature information data D = {x} is obtained. i y i}, where the geological feature label data y i This includes at least one of porosity, permeability, fracture information, lithological information, and oil saturation.
[0058] Because current well logging geological interpretation is relatively singular in its dimensions (e.g., interpreting lithology first, then porosity, then permeability, and finally oil saturation), and does not consider the correlation between various types of well logging geological information, the accuracy of well logging geological interpretation is low. However, in this embodiment of the invention, by embedding geological feature label data representing well logging geological information such as porosity, permeability, fracture information, lithology information, and oil saturation into multimodal data, the generated feature information data contains rich prior information and integrates the correlation between various types of well logging geological information. Therefore, it can effectively expand the dimensions of well logging geological interpretation and improve the accuracy of well logging geological interpretation.
[0059] S104. Generate feature information links based on feature information data and noise data.
[0060] In this context, feature information links represent the feature information of both feature information data and noise data. Specifically, in this embodiment of the invention, firstly, the feature information sequences in the feature information data that meet preset similarity conditions are linked to obtain sample feature links, and then the noise data is processed according to a normal distribution to obtain multiple noise feature vectors. These multiple noise feature vectors are then input into a target generative adversarial model (GAP) so that the generative model in the target GAP generates noise feature links based on the multiple noise feature vectors. The sample feature links and noise feature links are used as feature information links, where the sample feature links are the sample data (real data), and the noise feature links are the generated data.
[0061] Optionally, the judgment method for satisfying the preset similarity condition includes: calculating the ratio (or difference) of data in any feature information sequence. If the ratio (or difference) is within a preset numerical range, it indicates that the preset similarity condition is satisfied. For example, the judgment method for satisfying the preset similarity condition is: if the difference of data in any feature information sequence is within a preset numerical range, it indicates that the preset similarity condition is satisfied. The preset numerical range is [-20, 20]. The porosity represented by the geological feature label data y1 in the first feature information sequence is 40%, and the porosity represented by the geological feature label data y2 in the second feature information sequence is 55%. Since the difference between the two is 15 (or -15), which is within the preset numerical range, it indicates that the first feature information sequence and the second feature information sequence satisfy the preset similarity condition.
[0062] Specifically, after identifying all feature information sequences that meet the preset similarity conditions, corresponding nodes are generated to link the feature information sequences through these nodes. For example... Figure 3 As shown, multiple feature information sequences 31 that meet preset similarity conditions are linked through node 32 to generate sample feature links 33. The data form of sample feature links 33 is: X = w1m i +w2I i +w3H i …+w k H i ′, where w1, w2, w3…w k m represents the data weight value. i For multimodal soft data (including sedimentary microfacies coefficients S, logging curve waveform similarity C, fault well spacing F, and inter-well straight-line distance D), I i H is the similarity index between adjacent wells. i …H i ′ represents several types of hard data.
[0063] Furthermore, in this embodiment of the invention, noisy data is randomly extracted from the latent space (a mathematical space representing the abstract features of the data), and the noisy data is normally distributed to generate a vector sequence {t1, t2…t}. i} (i.e., i noise feature vectors), and then input the noise feature vectors 34 into the target generative adversarial model so that the generative model 301 in the target generative adversarial model generates noise feature links 35 based on these noise feature vectors.
[0064] S105. Train the target generation adversarial model based on feature information linking, so that the trained target generation adversarial model can generate multimodal geological interpretation data corresponding to the target area.
[0065] Among them, multimodal geological interpretation data contains comprehensive and diverse geological information within the target area, which is used to provide data support for well logging geological interpretation.
[0066] Specifically, the target generative adversarial model consists of a generative model and a discriminative model. The generative model is responsible for capturing the distribution of sample data, while the discriminative model is generally a binary classifier that distinguishes whether the input data comes from the generative model or real sample data. After optimizing and training the target generative adversarial model, the generative model can estimate the distribution of sample data and generate data that conforms to this distribution.
[0067] In this embodiment of the invention, the unsupervised learning characteristics of the target generative adversarial model are utilized to enable the target generative adversarial model to learn sample feature links and noise feature links. The target generative adversarial model is then optimized and trained. During training, one side (discriminator model or generator model) is fixed while the parameters of the other model are updated. This process is repeated iteratively until the generator model is able to estimate the distribution of sample data, thereby enabling it to estimate the distribution of sample feature links and generate more multimodal geological interpretation data that conforms to this distribution.
[0068] Specifically, such as Figure 3 As shown, in this embodiment of the invention, sample feature links 33 and noise feature links 35 are input into the target generative adversarial model, so that the discriminant model 302 in the target generative adversarial model outputs the feature link discrimination result (i.e., the probability estimate obtained by classifying the sample feature links and noise feature links; if it is judged to be sample data, it is recorded as 1; if it is judged to be generated data, it is recorded as 0). Then, the target generative adversarial model is trained based on the feature link discrimination result, so that the trained target generative adversarial model outputs multimodal geological interpretation data corresponding to the target area.
[0069] Optionally, the above training process includes: determining the loss function (the probability of generated data being judged as sample data) for sample feature links and noise feature links based on the feature link discrimination results; and then training the generation model and the discrimination model according to the loss function. The training methods include model updating and data weight optimization. Specifically, the core of model updating lies in determining the objective function of model optimization according to the principle of minimizing loss, so as to achieve the optimization of model parameters. The expression of the objective function is shown in Equation 3.
[0070]
[0071] Where G is the generative model, D is the discriminative model, and V is the set of linked nodes (V = (v1, v2, ..., v...). V E represents the set of edge information for the linked nodes. S represents the stored set of multimodal data sequences of geological attributes (S = (s1, s2, ..., s...)). V )).
[0072] Furthermore, data weight optimization includes: using Bayesian optimization methods, and based on the principle of minimizing the model prediction capability evaluation index MSE, optimizing multimodal soft data and the adjacent well similarity index I. i and several types of hard data H i …H i Information weights (w1, w2, w3…w) k ).
[0073] After completing the model update and data weight optimization, the generative model is able to generate data that is more similar to the sample data, and the discriminative model is unable to accurately distinguish between the sample data and the generated data. Next, the discriminative model outputs a discrimination result that meets the preset discrimination conditions, and the generative model generates target data that meets the preset discrimination conditions. This target data is used as multimodal geological interpretation data. The preset discrimination conditions include: whether the target data generated by the generative model makes the discrimination result output by the discriminative model 1 / 2 (i.e., the similarity between the target data and the sample data is extremely high, making it impossible for the discriminator to distinguish them accurately). If so, it means that the discrimination result and the target data meet the preset discrimination conditions, and this target data is used as multimodal geological interpretation data. The multimodal geological interpretation data contains rich and highly correlated well logging geological information such as porosity, permeability, fracture information, lithological information, and oil saturation.
[0074] Existing well logging geological interpretation technologies typically only consider well logging information from individual wells without analyzing related prior information. This results in insufficient utilization of prior information, wasting information resources and leading to low accuracy in well logging geological interpretation. Furthermore, the lack of consideration for the correlation between various types of well logging geological information in current well logging geological interpretations results in a relatively singular dimension of interpretation, further reducing the accuracy of well logging geological interpretation.
[0075] The knowledge-driven multimodal single-well geological intelligent interpretation method provided by this invention combines prior-obtained historical multimodal well data, conventional logging information, and neighboring well similarity index sets characterizing the correlation between adjacent oil wells to generate feature information data with multi-source heterogeneity. Based on the feature information data, feature information links are generated, and a target generative adversarial network is trained on these feature information links. This enables the target generative adversarial network to automatically generate diverse and highly correlated multimodal geological interpretation data. Using multimodal geological interpretation data as the basis for logging geological interpretation can improve the comprehensiveness of logging geological interpretation, thereby effectively improving the accuracy of logging geological interpretation.
[0076] Furthermore, another aspect of the present invention provides a knowledge-driven multimodal single-well geological intelligent interpretation device; please refer to [link to relevant documentation]. Figure 4 , Figure 4This application provides a specific description of a knowledge-driven multimodal single-well geological intelligent interpretation device, which includes: an inter-well data acquisition unit 41, a neighboring well similarity index set acquisition unit 42, a feature information data generation unit 43, a feature information link generation unit 44, and a multimodal geological interpretation data generation unit 45.
[0077] (1) Inter-well data acquisition unit 41
[0078] The well data acquisition unit 41 is used to acquire historical multimodal well data and single-well logging data within the target area.
[0079] (2) Unit 42 for obtaining adjacent well similarity index set
[0080] The adjacent well similarity index set acquisition unit 42 is used to obtain the adjacent well similarity index set based on historical multimodal inter-well data and single-well logging data.
[0081] Specifically, the adjacent well similarity index set acquisition unit 42 is used for:
[0082] Based on single-well logging data, determine the similarity of logging curve waveforms between any two oil wells;
[0083] The sedimentary microfacies coefficient, logging curve waveform similarity, fault well spacing, and inter-well straight-line distance are calculated and processed to obtain the adjacent well similarity index set. Among them, the historical multimodal inter-well data includes sedimentary microfacies coefficient, fault well spacing, and inter-well straight-line distance.
[0084] (3) Feature information data generation unit 43
[0085] The feature information data generation unit 43 is used to generate feature information data based on the neighboring well similarity index set, historical multimodal well data and single-well logging data.
[0086] Specifically, the feature information data generation unit 43 is used for:
[0087] Multimodal soft data is generated based on the neighboring well similarity index set and historical multimodal inter-well data;
[0088] Multimodal soft data and single-well logging data are combined and processed to obtain multimodal data;
[0089] Multimodal data is embedded to obtain feature information data, which includes multiple feature information sequences.
[0090] Specifically, the feature information data generation unit 43 is also used for:
[0091] Geological feature label data is embedded into multimodal data to obtain feature information data. The geological feature label data includes at least one of porosity, permeability, fracture information, lithology information, and oil saturation.
[0092] (4) Feature information link generation unit 44
[0093] The feature information link generation unit 44 is used to generate feature information links based on feature information data and noise data.
[0094] Specifically, the feature information linking generation unit 44 is used for:
[0095] The feature information sequences that meet the preset similarity conditions in the feature information data are linked to obtain sample feature links, and
[0096] The noise data is processed using a normal distribution to obtain multiple noise feature vectors;
[0097] Multiple noise feature vectors are input into the target generative adversarial model so that the generative model in the target generative adversarial model generates noise feature links based on the multiple noise feature vectors. The feature information links include sample feature links and noise feature links.
[0098] (5) Multimodal geological interpretation data generation unit 45
[0099] The multimodal geological interpretation data generation unit 45 is used to train the target generation adversarial model based on feature information links, so that the trained target generation adversarial model can generate multimodal geological interpretation data corresponding to the target area.
[0100] The multimodal geological interpretation data generation unit 45 is specifically used for:
[0101] Input the sample feature link and the noise feature link into the target generative adversarial model so that the discriminative model in the target generative adversarial model outputs the feature link discrimination result;
[0102] The target generation adversarial model is trained based on the feature link discrimination results, so that the trained target generation adversarial model outputs multimodal geological interpretation data corresponding to the target area.
[0103] Specifically, the multimodal geological interpretation data generation unit 45 is also used for:
[0104] Based on the feature link discrimination results, the loss functions for sample feature links and noise feature links are determined;
[0105] The generative and discriminative models are trained based on the loss function so that the trained generative model can generate multimodal geological interpretation data corresponding to the target area.
[0106] Furthermore, the multimodal geological interpretation data generation unit 45 is also used for:
[0107] The generative and discriminative models are trained based on the loss function so that the trained discriminative model outputs a discrimination result that meets the preset discrimination conditions.
[0108] The trained generative model generates target data that meets preset discrimination conditions, including target data in multimodal geological interpretation data.
[0109] The knowledge-driven multimodal single-well geological intelligent interpretation device provided by this invention combines historical multimodal well data obtained a priori, conventional logging information, and neighboring well similarity index sets characterizing the correlation between adjacent oil wells to generate feature information data with multi-source heterogeneity. Based on the feature information data, feature information links are generated, and a target generative adversarial network is trained on these feature information links. This enables the target generative adversarial network to automatically generate diverse and highly correlated multimodal geological interpretation data. Using multimodal geological interpretation data as the basis for logging geological interpretation can improve the comprehensiveness of logging geological interpretation, thereby effectively improving the accuracy of logging geological interpretation.
[0110] The knowledge-driven multimodal single-well geological intelligent interpretation device includes a processor and a memory. The aforementioned inter-well data acquisition unit, adjacent well similarity index set acquisition unit, feature information data generation unit, feature information link generation unit, and multimodal geological interpretation data generation unit are all stored as program units in the memory. The processor executes the aforementioned program units stored in the memory to realize the corresponding functions.
[0111] The processor contains a kernel, which retrieves the corresponding program units from memory. One or more kernels can be configured, and adjusting kernel parameters can improve the accuracy of well logging geological interpretation.
[0112] The memory may include non-permanent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.
[0113] This invention provides a storage medium storing a program that, when executed by a processor, implements the knowledge-driven multimodal single-well geological intelligent interpretation method.
[0114] This invention provides a processor for running a program, wherein the program executes the knowledge-driven multimodal single-well geological intelligent interpretation method during runtime.
[0115] This invention provides a device including a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, it performs the following steps: acquiring historical multimodal well-to-well data and single-well logging data within a target area; obtaining a set of neighboring well similarity indices based on the historical multimodal well-to-well data and single-well logging data; generating feature information data based on the neighboring well similarity index set, historical multimodal well-to-well data, and single-well logging data; generating feature information links based on the feature information data and noise data; and training a target generative adversarial model based on the feature information links, so that the trained target generative adversarial model generates multimodal geological interpretation data corresponding to the target area. The device described herein can be a server, PC, PAD, mobile phone, etc.
[0116] This application also provides a computer program product, which, when executed on a data processing device, is suitable for executing an initialization program having the following method steps: acquiring historical multimodal well-to-well data and single-well logging data within a target area; obtaining a set of neighboring well similarity indices based on the historical multimodal well-to-well data and single-well logging data; generating feature information data based on the neighboring well similarity indices, historical multimodal well-to-well data, and single-well logging data; generating feature information links based on the feature information data and noise data; and training a target generative adversarial model based on the feature information links, so that the trained target generative adversarial model generates multimodal geological interpretation data corresponding to the target area.
[0117] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0118] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0119] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0120] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0121] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.
[0122] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.
[0123] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.
[0124] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.
[0125] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.
Claims
1. A knowledge-driven, multimodal, single-well geological intelligent interpretation method, characterized in that, include: Acquire historical multimodal well-to-well data and single-well logging data within the target area; Based on the historical multimodal well data and the single-well logging data, a set of neighboring well similarity indices is obtained; Feature information data is generated based on the adjacent well similarity index set, the historical multimodal well data, and the single-well logging data; Based on the feature information data and noise data, a feature information link is generated; The target generation adversarial model is trained based on the aforementioned feature information link, so that the trained target generation adversarial model generates multimodal geological interpretation data corresponding to the target area; The step of obtaining the neighboring well similarity index set based on the historical multimodal well data and the single-well logging data includes: Based on the single-well logging data, determine the waveform similarity of the logging curves of any two oil wells; The sedimentary microfacies coefficient, the waveform similarity of the well logging curve, the fault well spacing, and the straight-line distance between wells are calculated and processed to obtain the adjacent well similarity index set. The historical multimodal well data includes the sedimentary microfacies coefficient, the fault well spacing, and the straight-line distance between wells. The step of generating feature information data based on the adjacent well similarity index set, the historical multimodal well data, and the single-well logging data includes: Based on the neighboring well similarity index set and the historical multimodal well data, multimodal soft data is generated; The multimodal soft data and the single-well logging data are combined and processed to obtain multimodal data; The multimodal data is embedded to obtain feature information data, wherein the feature information data includes multiple feature information sequences.
2. The knowledge-driven multimodal single-well geological intelligent interpretation method according to claim 1, characterized in that, The step of embedding the multimodal data to obtain feature information data includes: Geological feature label data is embedded into the multimodal data to obtain feature information data, wherein the geological feature label data includes at least one of porosity, permeability, fracture information, lithology information and oil saturation.
3. The knowledge-driven multimodal single-well geological intelligent interpretation method according to claim 2, characterized in that, The step of generating feature information links based on the feature information data and noise data includes: The feature information sequences that meet the preset similarity conditions in the feature information data are linked to obtain sample feature links, and The noise data is processed using a normal distribution to obtain multiple noise feature vectors; The plurality of noise feature vectors are input into the target generative adversarial model, so that the generative model in the target generative adversarial model generates noise feature links based on the plurality of noise feature vectors, wherein the feature information links include the sample feature links and the noise feature links.
4. The knowledge-driven multimodal single-well geological intelligent interpretation method according to claim 3, characterized in that, The step of training the target generative adversarial model based on the feature information link, so that the trained target generative adversarial model generates multimodal geological interpretation data corresponding to the target region, includes: The sample feature links and the noise feature links are input into the target generative adversarial model, so that the discriminant model in the target generative adversarial model outputs the feature link discrimination result; The target generation adversarial model is trained based on the feature link discrimination results, so that the trained target generation adversarial model outputs multimodal geological interpretation data corresponding to the target area.
5. The knowledge-driven multimodal single-well geological intelligent interpretation method according to claim 4, characterized in that, The step of training the target generative adversarial model based on the feature link discrimination result, so that the trained target generative adversarial model outputs multimodal geological interpretation data corresponding to the target region, includes: Based on the feature link discrimination results, the loss functions for the sample feature links and the noise feature links are determined; The generative model and the discriminative model are trained according to the loss function so that the trained generative model generates multimodal geological interpretation data corresponding to the target area.
6. The knowledge-driven multimodal single-well geological intelligent interpretation method according to claim 5, characterized in that, The step of training the generative model and the discriminative model according to the loss function, so that the trained generative model generates multimodal geological interpretation data corresponding to the target area, includes: The generative model and the discriminative model are trained according to the loss function, so that the trained discriminative model outputs a discrimination result that meets the preset discrimination conditions. The trained generative model generates target data that satisfies the preset discrimination conditions, wherein the multimodal geological interpretation data includes the target data.
7. A knowledge-driven multimodal single-well geological intelligent interpretation device, characterized in that, It is configured to perform the knowledge-driven multimodal single-well geological intelligent interpretation method according to any one of claims 1 to 6.
8. A machine-readable storage medium, characterized in that, The machine-readable storage medium stores instructions that, when executed by a processor, cause the processor to implement the knowledge-driven multimodal single-well geological intelligent interpretation method according to any one of claims 1 to 6.