Reservoir parameter prediction method and system, processing equipment and storage medium

Through small sample learning network and data augmentation technology based on RelationNet, combined with seismic and well logging data, a reservoir parameter prediction model is constructed, which solves the problem of insufficient number of well points and overfitting of deep learning models in the case of small samples in the existing technology. , realize efficient and accurate prediction of reservoir parameters.

CN118897320AInactive Publication Date: 2024-11-05CHINA NATIONAL OFFSHORE OIL (CHINA) CO LTD +1

Patent Information

Application Number
CN202411029147.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-30
Publication Date
2024-11-05
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing reservoir parameter prediction method relies on well point data. Insufficient well points affect the prediction accuracy, there are multiple solutions and high computing resource consumption. Deep learning models are prone to overfitting in small samples, making it difficult to effectively predict complex non-complexing. Linear relationship.

Method used

Using a small sample learning network based on RelationNet, combining seismic data and logging data, low-frequency model of reservoir parameters is enhanced and constructed, porosity, permeability and saturation prediction models are constructed, and the samples are expanded by SMOTE method, and spatial interpolation and Feature enhancement to improve prediction accuracy.

Benefits of technology

In the case of small samples, efficient and accurate prediction of reservoir parameters is achieved, overfitting is avoided, and the generalization performance and prediction efficiency of the model are improved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118897320A_ABST
    Figure CN118897320A_ABST
Patent Text Reader

Abstract

The invention relates to a reservoir parameter prediction method and system, processing equipment and a storage medium, and the method is characterized in that the method comprises the steps: obtaining post-stack seismic data in geophysical exploration; the obtained post-stack seismic data are input into a pre-constructed reservoir parameter prediction model, reservoir parameters of the post-stack seismic data are predicted, the reservoir parameter prediction model is constructed based on a RelationNet small sample learning network and comprises a porosity prediction model, a permeability prediction model and a saturation prediction model, and the permeability prediction model comprises a permeability prediction model and a saturation prediction model; the method can effectively solve the problem of uneven sample distribution, improves the prediction accuracy, and can be widely applied to the technical field of oil and gas resource exploration.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of oil and gas resource exploration, and in particular to a reservoir parameter prediction method, system, processing equipment and storage medium. Background Art

[0002] Reservoir prediction is the basis of oil and gas exploration and development. It is an important means to evaluate and predict the geological structure, lithology distribution, physical parameters and fluid characteristics of the reservoir by comprehensively utilizing geological, seismic, drilling and logging exploration and development data. It is also the main basis for trap evaluation and detailed description of oil and gas reservoirs. The prediction of reservoir parameters is a hot spot and difficulty in oil and gas exploration and development research. Reservoir parameters refer to parameters such as porosity, permeability, saturation, etc. used to describe the physical properties of the reservoir and the properties of the fluid therein. These parameters can accurately and quantitatively characterize the size of the reservoir pores, the ability of the reservoir to allow fluids to pass through, and the percentage of the pore volume occupied by each phase of fluid (oil, gas or water) in the oil and gas reservoir rock. These parameters have important applications in oil and gas exploration and development and are one of the important bases for evaluating reservoir quality.

[0003] At present, the reservoir parameters at the well point can be basically accurately obtained through the methods of drilling core measurement and logging interpretation. However, compared with the entire work area, the reservoir parameters at the well point cannot represent the changes in the reservoir parameters of the entire work area. In particular, tight reservoirs have the characteristics of thin reservoirs and strong lateral heterogeneity. When solving such problems, the existing method mainly uses the well-seismic joint method to find the regression relationship between seismic data and reservoir parameters, so as to predict the spatial distribution of reservoir parameters. The main methods currently used are as follows: (1) Reservoir parameter prediction method based on seismic inversion: This method uses logging data and seismic data to infer the physical properties of underground reservoirs. This method uses the propagation characteristics of seismic waves in the underground and converts seismic data into information reflecting the physical properties of the formation (such as porosity, permeability, etc.) through mathematical models. Before performing seismic inversion, it is usually necessary to establish a rock physics model to describe the relationship between seismic wave impedance and reservoir physical properties. This process usually involves a large amount of laboratory measurements and theoretical research. Then, seismic inversion is performed under well constraints to obtain wave impedance data. (2) Reservoir physical property parameter prediction method based on deep learning model: This method is a reservoir physical property parameter prediction method based on deep learning model. It mainly uses the powerful nonlinear mapping and learning ability of neural network to model the complex relationship between wellbore physical property parameters and seismic data, and then predicts the physical property parameters of the reservoir in the entire work area.

[0004] However, the existing methods also have some technical problems. In method (1), first, the method relies on well point data, so its application is affected by the distribution of well points. If the number of well points is insufficient, the accuracy of the prediction results may be affected. Second, the method has multiple solutions. Due to the non-uniqueness of the seismic inversion problem, there may be multiple different inversion solutions for the same set of seismic data. In addition, seismic inversion is a complex mathematical problem that requires a lot of computing resources and time. The steps of this method are cumbersome, the prediction time is long, and the efficiency is low. In method (2), first, deep learning models usually require a large amount of sample data for training, and the acquisition of samples can only rely on well data, which is expensive and difficult to obtain. In addition, deep learning models with small samples may have potential problems such as overfitting when dealing with complex nonlinear relationships, which will affect the accuracy of the prediction results. Finally, the complexity of the model and the large amount of computing requirements may lead to a large time consumption in the training and inference process.

[0005] In addition, the current success of machine learning mainly relies on massive amounts of labeled data. However, for reservoir parameter prediction, reservoir parameter labels can only be obtained through well logging data. The number of well logging data accounts for a very small proportion of the entire work area. Therefore, reservoir parameter prediction is also a small sample problem. Popular deep learning models are prone to overfitting when dealing with small sample reservoir parameter predictions. Summary of the invention

[0006] In view of the above problems, the purpose of the present invention is to provide a reservoir parameter prediction method, system, processing equipment and storage medium, which can efficiently and accurately predict reservoir parameters.

[0007] To achieve the above object, the present invention adopts the following technical solutions: In a first aspect, a reservoir parameter prediction method is provided, comprising: Acquisition of post-stack seismic data in geophysical exploration; The acquired post-stack seismic data is input into a pre-built reservoir parameter prediction model to predict the reservoir parameters of the post-stack seismic data, wherein the reservoir parameter prediction model is constructed based on the RelationNet small sample learning network, including a porosity prediction model, a permeability prediction model and a saturation prediction model.

[0008] Furthermore, the acquisition of reservoir parameter prediction well data and processing thereof include: Acquiring and processing reservoir parameter prediction well data, wherein the reservoir parameter prediction well data includes well logging curve data, well logging interpretation data and time-depth relationship data; Time-depth conversion is performed on the well logging curve data and well logging interpretation data in the processed reservoir parameter prediction well data.

[0009] Furthermore, the construction process of the reservoir parameter prediction model is as follows: Acquire post-stack seismic data and seismic horizon data and process them; Obtain reservoir parameter prediction well data and process it; Resampling the processed post-stack seismic data and reservoir parameter prediction well data; Based on the resampled post-stack seismic data and well logging curve data as well as the well logging interpretation data after time-depth conversion, a learning sample set is constructed, and data enhancement and optimization are performed; Based on the processed seismic horizon data and the resampled reservoir parameter prediction well data, a low-frequency model of reservoir parameters is constructed; Based on the constructed low-frequency model of reservoir parameters and the optimized learning sample set, the RelationNet small sample learning network model is trained to obtain a trained reservoir parameter prediction model.

[0010] Furthermore, the resampling of the processed post-stack seismic data and reservoir parameter prediction well data includes: Resample the amplitude and phase of the processed post-stack seismic data; The porosity, permeability, and saturation curves in the time-depth converted logging data are resampled, and the sampling rate is the same as that of the post-stack seismic data.

[0011] Furthermore, the method constructs a learning sample set based on the resampled post-stack seismic data and well logging curve data and the well logging interpretation data after time-depth conversion, and performs data enhancement and optimization, including: Obtain the spatial position of each sampling point on the predicted well according to the well point coordinates and well deviation data; Extract the nearest seismic trace data and seismic attribute trace data according to the spatial position to form the well-side seismic trace data; A learning sample set is constructed based on the resampled post-stack seismic data and well logging curve data as well as the well logging interpretation data after time-depth conversion; Perform data augmentation on the learning sample set; Optimize the learning sample set after data enhancement; The optimized learning sample set is loaded into the post-stack seismic data, and the wellside seismic traces are extracted for comparison and inspection with the obtained wellside seismic trace data.

[0012] Furthermore, the reservoir parameter low-frequency model is constructed based on the processed seismic horizon data and the resampled reservoir parameter prediction well data, including: Interpolate the seismic layer data and arrange them spatially from deep to shallow to obtain three-dimensional stratigraphic framework data in the time domain; Based on the processed logging curve data in the reservoir parameter prediction well data, spatial interpolation is performed under the control of the three-dimensional stratigraphic grid in the time domain to form a low-frequency model of reservoir parameters.

[0013] Furthermore, the RelationNet small sample learning network model is trained based on the constructed reservoir parameter low-frequency model and the optimized learning sample set to obtain a trained reservoir parameter prediction model, including: splicing seismic attributes and low-frequency models of reservoir parameters into multi-channel samples; Input the labeled data and the data to be predicted in the multi-channel samples into the deep learning neural network model respectively, and output the feature vector; The output feature vectors are concatenated to obtain a merged feature vector; The obtained feature vector is input into the relational network and the similarity measure is output; Repeat the above steps until all the labeled data and the data to be predicted are processed, and the final reservoir parameter prediction results are output to obtain the trained reservoir parameter prediction models, including the porosity prediction model, the permeability prediction model and the saturation prediction model.

[0014] In a second aspect, a reservoir parameter prediction system is provided, comprising: A data acquisition module for acquiring post-stack seismic data in geophysical exploration; The reservoir parameter prediction module is used to input the acquired post-stack seismic data into a pre-built reservoir parameter prediction model to predict the reservoir parameters of the post-stack seismic data, wherein the reservoir parameter prediction model is constructed based on the RelationNet small sample learning network, including a porosity prediction model, a permeability prediction model and a saturation prediction model.

[0015] In a third aspect, a processing device is provided, comprising computer program instructions, wherein the computer program instructions, when executed by the processing device, are used to implement the steps corresponding to the above-mentioned reservoir parameter prediction method.

[0016] In a fourth aspect, a computer-readable storage medium is provided, on which computer program instructions are stored, wherein the computer program instructions, when executed by a processor, are used to implement the steps corresponding to the above-mentioned reservoir parameter prediction method.

[0017] The present invention adopts the above technical solution, which has the following advantages: 1. The machine learning methods commonly used for reservoir parameter prediction are all developed based on traditional statistical theory and are based on the asymptotic theory in traditional statistical theory. A large number of training samples are required to obtain higher prediction accuracy, but when faced with small sample problems, it is often difficult to achieve ideal generalization effects and adaptability due to its inherent principles and property limitations. Samples in the field of oil and gas exploration mainly come from well data, and samples are often limited. The present invention uses the powerful self-learning ability of the small sample learning algorithm model to mine effective information from seismic data and well logging data itself to identify reservoir parameters. In the case of a small number of samples, reservoir porosity and saturation parameters can be predicted.

[0018] 2. When the currently popular deep learning model is directly applied to the joint well-seismic reservoir parameter identification, it faces the problem of insufficient sample quantity. The present invention uses a machine learning method based on SMOTE to expand the sample data, enhances the sample feature space by adding seismic attributes, and effectively solves the problem of uneven sample distribution.

[0019] 3. At present, when deep learning is used to directly predict reservoir parameters, the model is prone to overfitting and has poor generalization performance. The present invention constructs three-dimensional stratigraphic grid data in the time domain through seismic layer data, and forms a low-frequency model of reservoir parameters by performing spatial interpolation under the control of the three-dimensional stratigraphic grid in the time domain to improve the prediction accuracy.

[0020] In summary, the present invention can be widely used in the technical field of oil and gas resource exploration. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Throughout the accompanying drawings, the same reference numerals are used to represent the same components. In the accompanying drawings: Figure 1 It is a schematic diagram of a method flow provided by an embodiment of the present invention; Figure 2 It is a schematic diagram of the structure of a reservoir parameter prediction model provided by an embodiment of the present invention; Figure 3 is a schematic diagram of a porosity prediction cross section provided by an embodiment of the present invention; Figure 4 is a schematic diagram of a saturation prediction profile provided by an embodiment of the present invention; Figure 5 It is a schematic diagram of a permeability prediction profile provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0022] The exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

[0023] It should be understood that the terms used herein are only for the purpose of describing specific example embodiments and are not intended to be limiting. Unless the context clearly indicates otherwise, the singular forms "one", "an" and "said" as used herein may also be meant to include plural forms. The terms "include", "comprise", "contain", and "have" are inclusive, and therefore specify the existence of stated features, steps, operations, elements and / or parts, but do not exclude the existence or addition of one or more other features, steps, operations, elements, parts, and / or combinations thereof. The method steps, processes, and operations described herein are not interpreted as necessarily requiring them to be performed in the specific order described or illustrated, unless the execution order is clearly indicated. It should also be understood that additional or alternative steps may be used.

[0024] Although the terms first, second, third, etc. can be used in the text to describe multiple elements, components, regions, layers and / or sections, these elements, components, regions, layers and / or sections should not be limited by these terms. These terms can only be used to distinguish an element, component, region, layer or section from another region, layer or section. Unless the context clearly indicates, terms such as "first", "second" and other numerical terms do not imply order or sequence when used in the text. Therefore, the first element, component, region, layer or section discussed below can be referred to as the second element, component, region, layer or section without departing from the teaching of the example embodiments.

[0025] At present, in the reservoir parameter prediction method based on seismic inversion disclosed in the prior art, firstly, the method depends on well point data, so its application is affected by the distribution of well points. If the number of well points is insufficient, the accuracy of the prediction results may be affected; secondly, the method has multiple solutions. Due to the non-uniqueness of the seismic inversion problem, there may be multiple different inversion solutions for the same set of seismic data. In addition, seismic inversion is a complex mathematical problem that requires a lot of computing resources and time. The steps of this method are cumbersome, the prediction time is long, and the efficiency is low. In the reservoir physical property parameter prediction method based on deep learning model disclosed in the prior art, firstly, the deep learning model usually requires a large amount of sample data for training, and the acquisition of samples can only rely on well data, and the acquisition of well data is expensive and difficult to meet; in addition, the deep learning model with small samples may have potential problems caused by small samples such as overfitting when dealing with complex nonlinear relationships, which will affect the accuracy of the prediction results; finally, the complexity of the model and the large amount of computing requirements may lead to a large time consumption in the training and inference process. In addition, the success of machine learning currently mainly relies on massive labeled data, but for reservoir parameter prediction, reservoir parameter labels can only be obtained through well logging data, and the number of well logging accounts for a very small proportion of the entire work area. Therefore, reservoir parameter prediction is also a small sample problem. The popular deep learning model is prone to overfitting when dealing with small sample reservoir parameter prediction. The embodiment of the present invention provides a reservoir parameter prediction method, including: obtaining post-stack seismic data in geophysical exploration; inputting the acquired post-stack seismic data into a pre-built reservoir parameter prediction model to predict the reservoir parameters of the post-stack seismic data, wherein the reservoir parameter prediction model includes a porosity prediction model, a permeability prediction model and a saturation prediction model. In recent years, small sample learning machine learning methods have shown great potential in dealing with problems with a small number of data samples. Small sample learning machine learning algorithms can accurately predict new data through in-depth analysis and learning of data in the case of only a small number of data samples. Therefore, the present invention applies small sample learning machine learning methods to the field of reservoir parameter prediction to solve the problems existing in the prior art.

[0026] Example 1 like Figure 1 As shown, this embodiment provides a reservoir parameter prediction method, comprising the following steps: 1) Obtain and process post-stack seismic data and seismic horizon data to improve the post-stack seismic data and seismic horizon data, specifically: 1.1) The post-stack seismic data are de-noised, weak signal enhancement is performed on the target layer, and seismic amplitude and phase attributes are extracted.

[0027] 1.2) Check the quality of seismic layer data from the perspectives of plane, section, and three-dimensional, and correct the layers with obvious cross-layer interpretation and incomplete plane in the seismic layer data to ensure the accuracy of the seismic layer data.

[0028] 2) Obtaining and processing reservoir parameter prediction well data, wherein the reservoir parameter prediction well data includes well logging curve data, well logging interpretation data and time-depth relationship data.

[0029] Specifically, reservoir parameter prediction well data include well logging curve data including well name, coordinates, well inclination, core elevation, porosity, permeability, saturation, target layer stratification data, logging interpretation data of lithology data (mud logging), and time-depth relationship data.

[0030] Specifically, the processing includes checking the well logging data (the well logging interpretation data and deep relationship data are processed results and do not need to be processed again), standardizing the porosity, permeability and saturation curves in the well logging data, and processing outliers.

[0031] 3) Perform time-depth conversion on the logging curve data and logging interpretation data in the processed reservoir parameter prediction well data.

[0032] Specifically, through the time-depth relationship of each well, the logging curve data and logging interpretation data in the processed reservoir parameter prediction well data in the depth domain are converted to the time domain. If there is no time-depth relationship, it is necessary to calibrate the seismic synthetic records through acoustic wave curves and density curves to obtain the time-depth relationship of the well.

[0033] 4) Resample the processed post-stack seismic data and the time-depth converted logging curve data, specifically: 4.1) Resample the amplitude and phase of the processed post-stack seismic data to 1 millisecond.

[0034] 4.2) Resample the porosity, permeability and saturation curves in the time-depth converted logging data, with the same sampling rate as the post-stack seismic data.

[0035] 5) Based on the resampled post-stack seismic data and well logging curve data and the well logging interpretation data after time-depth conversion, a learning sample set is constructed, and data enhancement and optimization are performed, specifically: 5.1) Obtain the spatial position of each sampling point on the predicted well based on the well point coordinates and well inclination data.

[0036] 5.2) Extract the nearest seismic trace data and seismic attribute trace data according to the spatial position to form the wellside seismic trace data.

[0037] 5.3) Construct a learning sample set based on the resampled post-stack seismic data and well logging curve data as well as the well logging interpretation data after time-depth conversion.

[0038] 5.4) Perform data augmentation on the learning sample set.

[0039] Specifically, due to the small number of well data samples, the sample diversity becomes low. When the amount of data is limited, the sample diversity can be improved through data enhancement. Data enhancement refers to the data expansion or feature enhancement of the original small sample data set with the help of auxiliary data or auxiliary information. Data expansion is to add new data to the original data set, which can be unlabeled data or synthetic labeled data; feature enhancement is to add features that are easy to classify in the feature space of the original sample to increase feature diversity.

[0040] Specifically, this embodiment can use a sample augmentation method based on the SMOTE machine learning method to perform data augmentation on the learning sample set. The SMOTE (Synthetic Minority Over-sampling Technique) method is a data synthesis method based on feature space, which is used to solve the problem of data category imbalance. New samples are generated by randomly weighting and summing small sample data categories with neighboring points in the feature space to make the data distribution tend to be balanced. The specific process is: ①For the minority class sample in the learning sample set, calculate the distance between it and all other minority class samples and find its K nearest neighbors.

[0041] ② Randomly select a sample from the K nearest neighbors and calculate the difference between the sample and the current minority class sample.

[0042] ③ Based on the calculated difference, a new synthetic sample is generated, which is located on the line between the two samples.

[0043] ④ Repeat the above steps ① to ③ until a specified number of synthetic samples are generated to obtain a learning sample set after data enhancement.

[0044] Specifically, this embodiment may also adopt a feature enhancement method based on earthquake attributes, that is, extracting multiple earthquake attributes and inputting external attributes, and improving the diversity of the learning sample set by enhancing the feature space of the learning sample set.

[0045] 5.5) Optimize the learning sample set after data enhancement, specifically: 5.5.1) Balance the learning sample set after data enhancement, retain some samples with small label values ​​and the remaining samples with large label values ​​in the learning sample set to ensure the balance of the learning sample set.

[0046] Specifically, the reservoir parameter prediction well data is resampled after the aligned logging data, and the logging samples are unevenly distributed. Taking porosity as an example, the label value of most sample points is below 0.001, which makes it difficult to fit the model during training and tends to be 0 during prediction. By balancing the learning sample set after data enhancement, only some samples with small label values ​​and the rest of the samples with large label values ​​are retained, so that the balance of the learning sample set can be guaranteed.

[0047] 5.5.2) The frequency distribution of logging labels is inconsistent with that of seismic data. The high-resolution seismic data obtained after resampling only increases the number of sampling points, which lacks high-frequency information and cannot reflect the rapidly changing part of the reservoir parameter prediction well data. Therefore, the logging curve data of porosity, permeability, and saturation in the reservoir parameter prediction well data are low-pass filtered to remove the high-frequency components so that the effective frequency bands of the input and labels coincide.

[0048] 5.6) Load the optimized learning sample set into the post-stack seismic data, extract the wellbore seismic traces and compare them with the wellbore seismic trace data obtained in step 5.2) to ensure the accuracy of the learning sample set data.

[0049] 6) Based on the processed seismic layer data and the reservoir parameter prediction well data after time-depth conversion, a low-frequency model of reservoir parameters is constructed.

[0050] Specifically, when the amount of training data is small, the deep neural network model training is prone to overfitting, and the model is difficult to learn the implicit rules in the data. By combining the reservoir parameter prediction well data to establish a low-frequency model of reservoir parameters, and inputting it into the deep learning neural network model together with a variety of seismic attributes, prior information can be introduced into the model training to enhance the network's ability to capture reservoir parameter trends. Constraining the deep learning neural network model with the reservoir parameter low-frequency model can enhance the model's prediction accuracy and geological continuity, thereby improving the reliability of reservoir parameter characterization. The specific construction process of the reservoir parameter low-frequency model is as follows: 6.1) Interpolate the seismic layer data and arrange them spatially in order from deep to shallow to obtain three-dimensional stratigraphic grid data in the time domain.

[0051] 6.2) Based on the processed logging curve data in the reservoir parameter prediction well data, spatial interpolation is performed under the control of the three-dimensional stratigraphic grid in the time domain to form a low-frequency model of reservoir parameters.

[0052] 7) Based on the constructed reservoir parameter low-frequency model and the optimized learning sample set, the RelationNet small sample learning network model is trained to obtain a trained reservoir parameter prediction model, such as Figure 2 shown.

[0053] Specifically, RelationNet is a small sample learning network based on metric learning. This embodiment uses the constructed reservoir parameter low-frequency model to train the RelationNet small sample learning network model. The specific process is: 7.1) Seismic attributes such as seismic amplitude and phase are spliced ​​with low-frequency models of reservoir parameters into multi-channel samples.

[0054] 7.2) The labeled data in the multi-channel samples and the data to be predicted They are respectively input into the deep learning neural network model and the feature vector is output. The deep learning neural network model is a multi-layer one-dimensional convolutional neural network composed of cascaded convolutional layers (conv), normalization layers (ln) and activation layers (ReLU).

[0055] 7.3) Concatenate the output feature vectors to obtain the merged feature vector.

[0056] 7.4) Input the obtained feature vector into the relationship network and output the similarity measure , where the relational network is a multi-layer fully connected neural network consisting of cascaded fully connected layers (fc) and activation layers (ReLU).

[0057] 7.5) Repeat the above steps 7.1) to 7.4) until all the labeled data are completed. and the data to be predicted Process and output the final reservoir parameter prediction results , and obtain the trained reservoir parameter prediction models, including porosity prediction model, permeability prediction model and saturation prediction model (the porosity in the logging curve data is used to train the RelationNet small sample learning network model to obtain the porosity prediction model, and the permeability prediction model and saturation prediction model can be obtained similarly).

[0058] Specifically, for qualitative analysis, the reservoir parameter prediction result calculation formula is: (1) Specifically, for quantitative analysis, the reservoir parameter prediction result calculation formula is: (2) in, is the similarity measurement function; is the sample serial number.

[0059] 8) Obtain post-stack seismic data in geophysical exploration and input it into a pre-built reservoir parameter prediction model to predict reservoir parameters such as porosity, permeability and saturation from the post-stack seismic data. Specifically: 8.1) Acquisition of post-stack seismic data in geophysical exploration.

[0060] 8.2) Input the acquired post-stack seismic data into the porosity prediction model to predict the porosity, such as Figure 3 shown.

[0061] 8.3) Input the acquired post-stack seismic data into the permeability prediction model to predict the permeability, such as Figure 5 shown.

[0062] 8.4) Input the acquired post-stack seismic data into the saturation prediction model to predict the saturation, such as Figure 4 shown.

[0063] Example 2 This embodiment provides a reservoir parameter prediction system, including: The data acquisition module is used to acquire post-stack seismic data in geophysical exploration.

[0064] The reservoir parameter prediction module is used to input the acquired post-stack seismic data into a pre-built reservoir parameter prediction model to predict the reservoir parameters of the post-stack seismic data. The reservoir parameter prediction model is built based on the RelationNet small sample learning network, including a porosity prediction model, a permeability prediction model and a saturation prediction model.

[0065] The system provided in this embodiment is used to execute the above-mentioned method embodiments. Please refer to the above-mentioned embodiments for specific processes and detailed contents, which will not be repeated here.

[0066] Example 3 This embodiment provides a processing device corresponding to the reservoir parameter prediction method provided in Embodiment 1. The processing device can be applicable to a client processing device, such as a mobile phone, a laptop computer, a tablet computer, a desktop computer, etc., to execute the method of Embodiment 1.

[0067] The processing device includes a processor, a memory, a communication interface and a bus, and the processor, the memory and the communication interface are connected through the bus to complete mutual communication. The memory stores a computer program that can be run on the processing device, and the processing device executes the reservoir parameter prediction method provided in this embodiment 1 when running the computer program.

[0068] In some implementations, the memory may be a high-speed random access memory (RAM), and may also include a non-volatile memory, such as at least one disk storage.

[0069] In some other implementations, the processor may be a central processing unit (CPU), a digital signal processor (DSP), or other general-purpose processors of various types, which are not limited herein.

[0070] In addition, the logic instructions in the above-mentioned memory can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on such an understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0071] Those skilled in the art will understand that the structure of the above-mentioned computing device is only a partial structure related to the solution of the present invention, and does not constitute a limitation on the computing device to which the solution of the present invention is applied. The specific computing device may include more or fewer components, or combine certain components, or have a different arrangement of components.

[0072] Example 4 This embodiment provides a computer program product corresponding to the reservoir parameter prediction method provided in Embodiment 1. The computer program product may include a computer-readable storage medium on which computer-readable program instructions for executing the reservoir parameter prediction method described in Embodiment 1 are loaded.

[0073] Computer readable storage media can be tangible devices that hold and store instructions used by instruction execution devices. Computer readable storage media can be, for example, but not limited to, electronic storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any combination thereof.

[0074] The above embodiment provides a computer-readable storage medium, whose implementation principle and technical effect are similar to those of the above method embodiment, and will not be repeated here.

[0075] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0076] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0077] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0078] The above embodiments are only used to illustrate the present invention, wherein the structure, connection mode and manufacturing process of each component may be changed. Any equivalent transformations and improvements based on the technical solution of the present invention should not be excluded from the protection scope of the present invention.

Claims

1. A reservoir parameter prediction method, characterized in that: include: Acquisition of post-stack seismic data in geophysical exploration; The acquired post-stack seismic data is input into a pre-built reservoir parameter prediction model to predict the reservoir parameters of the post-stack seismic data, wherein the reservoir parameter prediction model is constructed based on the RelationNet small sample learning network, including a porosity prediction model, a permeability prediction model and a saturation prediction model.

2. A reservoir parameter prediction method according to claim 1, characterized in that: The method of obtaining reservoir parameter prediction well data and processing the data includes: Acquiring and processing reservoir parameter prediction well data, wherein the reservoir parameter prediction well data includes well logging curve data, well logging interpretation data and time-depth relationship data; Time-depth conversion is performed on the well logging curve data and well logging interpretation data in the processed reservoir parameter prediction well data.

3. A reservoir parameter prediction method according to claim 2, characterized in that: The construction process of the reservoir parameter prediction model is as follows: Acquire post-stack seismic data and seismic horizon data and process them; Obtain reservoir parameter prediction well data and process it; Resampling the processed post-stack seismic data and reservoir parameter prediction well data; Based on the resampled post-stack seismic data and well logging curve data as well as the well logging interpretation data after time-depth conversion, a learning sample set is constructed, and data enhancement and optimization are performed; Based on the processed seismic horizon data and the resampled reservoir parameter prediction well data, a low-frequency model of reservoir parameters is constructed; Based on the constructed low-frequency model of reservoir parameters and the optimized learning sample set, the RelationNet small sample learning network model is trained to obtain a trained reservoir parameter prediction model.

4. A reservoir parameter prediction method according to claim 3, characterized in that: The resampling of the processed post-stack seismic data and reservoir parameter prediction well data includes: Resample the amplitude and phase of the processed post-stack seismic data; The porosity, permeability, and saturation curves in the time-depth converted logging data are resampled, and the sampling rate is the same as that of the post-stack seismic data.

5. A reservoir parameter prediction method according to claim 3, characterized in that: The method constructs a learning sample set based on the resampled post-stack seismic data and well logging curve data and the well logging interpretation data after time-depth conversion, and performs data enhancement and optimization, including: Obtain the spatial position of each sampling point on the predicted well according to the well point coordinates and well deviation data; Extract the nearest seismic trace data and seismic attribute trace data according to the spatial position to form the well-side seismic trace data; A learning sample set is constructed based on the resampled post-stack seismic data and well logging curve data as well as the well logging interpretation data after time-depth conversion; Perform data augmentation on the learning sample set; Optimize the learning sample set after data enhancement; The optimized learning sample set is loaded into the post-stack seismic data, and the wellside seismic traces are extracted for comparison and inspection with the obtained wellside seismic trace data.

6. A reservoir parameter prediction method according to claim 2, characterized in that: The method of constructing a reservoir parameter low-frequency model based on the processed seismic horizon data and the resampled reservoir parameter prediction well data includes: Interpolate the seismic layer data and arrange them spatially from deep to shallow to obtain three-dimensional stratigraphic framework data in the time domain; Based on the processed logging curve data in the reservoir parameter prediction well data, spatial interpolation is performed under the control of the three-dimensional stratigraphic grid in the time domain to form a low-frequency model of reservoir parameters.

7. A reservoir parameter prediction method according to claim 2, characterized in that: The RelationNet small sample learning network model is trained based on the constructed reservoir parameter low-frequency model and the optimized learning sample set to obtain a trained reservoir parameter prediction model, including: splicing seismic attributes and low-frequency models of reservoir parameters into multi-channel samples; Input the labeled data and the data to be predicted in the multi-channel samples into the deep learning neural network model respectively, and output the feature vector; The output feature vectors are concatenated to obtain a merged feature vector; The obtained feature vector is input into the relational network and the similarity measure is output; Repeat the above steps until all the labeled data and the data to be predicted are processed, and the final reservoir parameter prediction results are output to obtain the trained reservoir parameter prediction models, including the porosity prediction model, the permeability prediction model and the saturation prediction model.

8. A reservoir parameter prediction system, characterized in that: include: A data acquisition module for acquiring post-stack seismic data in geophysical exploration; The reservoir parameter prediction module is used to input the acquired post-stack seismic data into a pre-built reservoir parameter prediction model to predict the reservoir parameters of the post-stack seismic data, wherein the reservoir parameter prediction model is constructed based on the RelationNet small sample learning network, including a porosity prediction model, a permeability prediction model and a saturation prediction model.

9. A processing device, characterized in that: It comprises computer program instructions, wherein the computer program instructions are used to implement the steps corresponding to the reservoir parameter prediction method according to any one of claims 1 to 7 when executed by a processing device.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer program instructions, wherein the computer program instructions, when executed by a processor, are used to implement the steps corresponding to the reservoir parameter prediction method according to any one of claims 1 to 7.

Citation Information

Patent Citations

  • Seismic logging ultrahigh resolution data fusion implementation method based on small sample machine learning

    CN113267816A

  • Intelligent interpretation method for reservoir lithogenous facies logging

    CN116304921A

  • Intelligent reservoir physical property parameter prediction method and system under small sample condition

    CN116398114A

  • Reservoir physical property parameter prediction method and apparatus

    WO2024087827A1

Cited By

  • Reservoir parameter prediction method suitable for low-permeability reservoir

    CN119644424A

  • Water logging degree and saturation change determination method and device, electronic equipment and storage medium

    CN120145840A

  • Water-flooding degree and saturation change determination method and device, electronic equipment and storage medium

    CN120145840B