Complex reservoir gas-bearing prediction method and device, electronic equipment and medium
By constructing a low-frequency velocity model based on seismic partitioning and a closed-loop convolutional network, combined with the sedimentary characteristics of well logging data, the problem of fluid prediction in complex reservoirs in northern Hubei was solved, and efficient reservoir gas-bearing prediction was achieved under conditions of few or no wells.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-29
- Publication Date
- 2026-03-31
AI Technical Summary
In the complex reservoirs of northern Hubei, existing technologies are insufficient for effective fluid prediction, especially under conditions of few or no wells, where seismic wave energy attenuation is severe and the signal-to-noise ratio of seismic data is inconsistent, making it impossible for traditional methods to accurately predict the gas content of reservoirs.
A low-frequency velocity model based on seismic partitioning is adopted, combined with unsupervised and semi-supervised closed-loop convolutional networks, to perform convolutional encoding and decoding of seismic data. The sedimentary characteristics of well logging data are used as constraints, and gas-bearing capacity of reservoirs is predicted through intelligent inversion of sensitive parameters of deep reservoirs.
It enables efficient fluid prediction of complex reservoirs under conditions of few or no wells, improves the accuracy and reliability of reservoir gas content prediction, and adapts to the complex geological conditions in northern Hubei.
Smart Images
Figure CN121763366A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of predicting gas content in complex reservoirs, and more specifically, to a method, apparatus, electronic device, and medium for predicting gas content in complex reservoirs. Background Technology
[0002] Northern Hubei contains structural belts such as the Yixia Slope and the Yimeng Uplift. Years of exploration have shown that the main oil and gas reservoirs in this region are lithological or structural lithological reservoirs, with the primary reservoirs being fluvial-lacustrine sandstones. These reservoirs exhibit significant lateral variations in rock mass and facies, dramatic velocity changes, and are characterized by complex lithology and poor sedimentary homogeneity. The dense clastic rocks are also complex in lithology and have varied frameworks. The region is highly fractured, with multiple porous media coexisting, resulting in severe seismic wave energy attenuation, classifying it as an absorptive and attenuating medium, making oil and gas reservoir exploration extremely difficult. Regarding high-resolution imaging of absorptive and attenuating medium reservoirs in northern Hubei, geophysicists have conducted extensive research on relatively amplitude-preserving seismic data processing, primarily focusing on the amplitude preservation of various processing techniques. Near-surface Q-compensation can effectively address the consistency of lateral frequency wave groups, but this technique is still in the research and trial stage and cannot improve inconsistencies caused by the signal-to-noise ratio. Furthermore, spherical diffusion is strongly limited by low-frequency energy, and the compensation factor can only meet low-frequency energy compensation needs; for high frequencies, deep compensation is clearly insufficient.
[0003] Currently, the most commonly used fluid prediction methods both domestically and internationally are seismic attribute analysis and inversion. The relevant attributes for fluid prediction are derived from seismic data, specifically measuring geometric, kinematic, dynamic, and statistical characteristics. These attributes are categorized as follows: statistical, mathematical calculation, etc.; attributes reflecting structural and reservoir characteristics (geometric features), including coherence, variance, edge detection, edge protection smoothing filtering, texture, structural steering filtering, energy gradient calculation, dip estimation, curvature, rose diagram, amplitude variation, etc.; attributes reflecting hydrocarbon and reservoir characteristics, including time-frequency analysis, single-frequency analysis, hydrocarbon detection, pre-stack / post-stack formation absorption coefficients, etc.; and analytical methods, including attribute scaling and formation slicing, etc.
[0004] Therefore, it is necessary to develop a method, device, electronic equipment, and medium for predicting the gas content of complex reservoirs.
[0005] The information disclosed in the background section of this invention is intended only to enhance the understanding of the general background of this invention, and should not be construed as an admission or in any way implying that such information constitutes prior art known to those skilled in the art. Summary of the Invention
[0006] This invention proposes a method, device, electronic equipment, and medium for predicting gas content in complex reservoirs. It can achieve fluid prediction under conditions with few wells, which is a key step in the exploration of complex reservoirs and the layout of well locations. It is of great significance for fluid prediction and subsequent processing in complex reservoirs.
[0007] In a first aspect, embodiments of this disclosure provide a method for predicting the gas content of complex reservoirs, including:
[0008] Construct a low-frequency velocity model based on seismic partitioning;
[0009] Seismic data is convolutionally encoded and decoded using an unsupervised closed-loop convolutional network to obtain a sparse representation of the seismic data.
[0010] Based on a semi-supervised closed-loop convolutional network, intelligent inversion of sensitive parameters of deep reservoirs is performed to predict the gas content of the reservoir.
[0011] As a specific implementation of this disclosure, a low-frequency velocity model based on seismic partitioning is constructed by using the Kriging interpolation algorithm and constraining the sedimentary characteristics of well logging data.
[0012] As a specific implementation of this disclosure, the convolution kernels for convolutional encoding and decoding are optimized using the batch gradient descent method.
[0013] As a specific implementation of this disclosure, the intelligent inversion of sensitive parameters of deep reservoirs based on a semi-supervised closed-loop convolutional network includes:
[0014] Based on the error gradient between the semi-supervised closed-loop convolutional network output and the actual seismic data, the low-frequency velocity model is updated to achieve intelligent inversion of sensitive parameters of deep reservoirs.
[0015] As a specific implementation of this disclosure, a semi-supervised closed-loop convolutional network is constructed by using the convolutional decoder in an unsupervised closed-loop convolutional neural network as the seismic wavelet.
[0016] Secondly, embodiments of this disclosure also provide a device for predicting the gas content of complex reservoirs, comprising:
[0017] The module is used to build a low-frequency velocity model based on seismic partitioning.
[0018] The encoding / decoding module performs convolutional encoding and decoding of seismic data using an unsupervised closed-loop convolutional network to obtain a sparse representation of the seismic data.
[0019] The inversion module, based on a semi-supervised closed-loop convolutional network, performs intelligent inversion of sensitive parameters of deep reservoirs to predict reservoir gas content.
[0020] As a specific implementation of this disclosure, a low-frequency velocity model based on seismic partitioning is constructed by using the Kriging interpolation algorithm and constraining the sedimentary characteristics of well logging data.
[0021] As a specific implementation of this disclosure, the convolution kernels for convolutional encoding and decoding are optimized using the batch gradient descent method.
[0022] As a specific implementation of this disclosure, the intelligent inversion of sensitive parameters of deep reservoirs based on a semi-supervised closed-loop convolutional network includes:
[0023] Based on the error gradient between the semi-supervised closed-loop convolutional network output and the actual seismic data, the low-frequency velocity model is updated to achieve intelligent inversion of sensitive parameters of deep reservoirs.
[0024] As a specific implementation of this disclosure, a semi-supervised closed-loop convolutional network is constructed by using the convolutional decoder in an unsupervised closed-loop convolutional neural network as the seismic wavelet.
[0025] Thirdly, embodiments of this disclosure also provide an electronic device, the electronic device comprising:
[0026] Memory, which stores executable instructions;
[0027] A processor that executes the executable instructions in the memory to implement the method for predicting the gas content of complex reservoirs.
[0028] Fourthly, embodiments of this disclosure also provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method for predicting the gas content of complex reservoirs.
[0029] Its beneficial effects are as follows:
[0030] This invention combines the seismic wave propagation theorem with machine learning theory to predict low-frequency velocity models based on seismic facies depositional characteristics. It also conducts research on seismic data convolutional encoding and decoding based on closed-loop convolutional networks, and completes intelligent prediction of reservoir fluids using semi-supervised closed-loop convolutional networks. This provides technical and data support for the prediction of complex reservoir fluids and has practical significance.
[0031] The methods and apparatus of the present invention have other features and advantages that will be apparent from or will be set forth in detail in the accompanying drawings and following detailed description, which together serve to explain the particular principles of the invention. Attached Figure Description
[0032] The above and other objects, features and advantages of the present invention will become more apparent from the more detailed description of exemplary embodiments of the invention in conjunction with the accompanying drawings, wherein the same reference numerals generally represent the same parts.
[0033] Figure 1 A flowchart illustrating the steps of a method for predicting the gas content of complex reservoirs according to an embodiment of the present invention is shown.
[0034] Figure 2 A schematic diagram of an unsupervised closed-loop convolutional network according to an embodiment of the present invention is shown.
[0035] Figure 3 A schematic diagram of a semi-supervised closed-loop convolutional network according to an embodiment of the present invention is shown.
[0036] Figure 4 A schematic diagram of a smart inversion profile of gas-bearing porosity of well 1 according to an embodiment of the present invention is shown.
[0037] Figure 5 A schematic diagram of a smart inversion profile of gas-bearing porosity of well 2 according to an embodiment of the present invention is shown.
[0038] Figure 6 A block diagram of a complex reservoir gas-bearing prediction device according to an embodiment of the present invention is shown.
[0039] Explanation of reference numerals in the attached figures:
[0040] 201. Construction module; 202. Encoding / decoding module; 203. Inversion module. Detailed Implementation
[0041] Preferred embodiments of the invention will now be described in more detail. While preferred embodiments of the invention are described below, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein.
[0042] To facilitate understanding of the solutions and effects of the embodiments of the present invention, six specific application examples are given below. Those skilled in the art should understand that these examples are merely for the purpose of understanding the present invention, and any specific details therein are not intended to limit the present invention in any way.
[0043] Example 1
[0044] Figure 1 A flowchart illustrating the steps of a method for predicting the gas content of complex reservoirs according to an embodiment of the present invention is shown.
[0045] like Figure 1 As shown, the method for predicting the gas content of complex reservoirs includes:
[0046] Step 101: Construct a low-frequency velocity model based on seismic partitioning;
[0047] Step 102: Perform convolutional encoding and decoding of seismic data using an unsupervised closed-loop convolutional network to obtain a sparse representation of the seismic data;
[0048] Step 103: Based on a semi-supervised closed-loop convolutional network, perform intelligent inversion of sensitive parameters of deep reservoirs to achieve gas content prediction.
[0049] In one example, a low-frequency velocity model based on seismic partitioning is constructed by using the Kriging interpolation algorithm and constraining the sedimentary features of well logging data.
[0050] In one example, the convolutional kernels for both encoding and decoding are optimized using batch gradient descent.
[0051] In one example, intelligent inversion of sensitive parameters of deep reservoirs based on a semi-supervised closed-loop convolutional network includes:
[0052] Based on the error gradient between the semi-supervised closed-loop convolutional network output and the actual seismic data, the low-frequency velocity model is updated to achieve intelligent inversion of sensitive parameters of deep reservoirs.
[0053] In one example, a semi-supervised closed-loop convolutional network is constructed using the convolutional decoder in an unsupervised closed-loop convolutional neural network as the seismic wavelet.
[0054] Specifically, the low-frequency model is constructed using the traditional Kriging interpolation algorithm. However, unlike the traditional algorithm, it uses the sedimentary characteristics of well logging data for relevant constraints, thus realizing the construction of a low-frequency model based on sedimentary characteristic constraints.
[0055] This invention, guided by earthquake inversion theory and based on knowledge of seismic wave dynamics and the inherent relationships between different data, constructs a closed-loop convolutional neural network. This convolutional neural network structure consists of an outer loop and an inner loop.
[0056] Figure 2 A schematic diagram of an unsupervised closed-loop convolutional network according to an embodiment of the present invention is shown.
[0057] The outer loop is an unsupervised closed-loop convolutional network, and the execution process is as follows: Figure 2 As shown, a random sample of seismic data is subjected to convolutional encoding and decoding to achieve a sparse representation of the seismic data. During this process, the convolutional kernels for both encoding and decoding are optimized using batch gradient descent. The trained convolutional decoder functions as a seismic wavelet and participates in the inner loop seismic inversion process.
[0058] Figure 3 A schematic diagram of a semi-supervised closed-loop convolutional network according to an embodiment of the present invention is shown.
[0059] Seismic inversion requires well logging data modeling, well-seismic calibration wavelet extraction, and well logging constraints during the inversion process. To adapt seismic inversion to deep reservoirs with few or no wells, this paper utilizes the characteristics of different machine learning frameworks. Under the constraints of a prior model, a convolutional decoder in an unsupervised closed-loop convolutional neural network is used as a bridge to form the inner loop of a closed-loop convolutional network for intelligent deep reservoir inversion. This is a semi-supervised closed-loop convolutional network. The error gradient between the network output and the actual seismic data is calculated to update the low-frequency velocity model values, enabling intelligent inversion of sensitive parameters of deep reservoirs. Figure 3 As shown, this enables the prediction of gas content in reservoirs.
[0060] Example 2
[0061] The present invention also provides a device for predicting the gas content of complex reservoirs, comprising:
[0062] The module is used to build a low-frequency velocity model based on seismic partitioning.
[0063] The encoding / decoding module performs convolutional encoding and decoding of seismic data using an unsupervised closed-loop convolutional network to obtain a sparse representation of the seismic data.
[0064] The inversion module, based on a semi-supervised closed-loop convolutional network, performs intelligent inversion of sensitive parameters of deep reservoirs to predict reservoir gas content.
[0065] In one example, a low-frequency velocity model based on seismic partitioning is constructed by using the Kriging interpolation algorithm and constraining the sedimentary features of well logging data.
[0066] In one example, the convolutional kernels for both encoding and decoding are optimized using batch gradient descent.
[0067] In one example, intelligent inversion of sensitive parameters of deep reservoirs based on a semi-supervised closed-loop convolutional network includes:
[0068] Based on the error gradient between the semi-supervised closed-loop convolutional network output and the actual seismic data, the low-frequency velocity model is updated to achieve intelligent inversion of sensitive parameters of deep reservoirs.
[0069] In one example, a semi-supervised closed-loop convolutional network is constructed using the convolutional decoder in an unsupervised closed-loop convolutional neural network as the seismic wavelet.
[0070] Specifically, the low-frequency model is constructed using the traditional Kriging interpolation algorithm. However, unlike the traditional algorithm, it uses the sedimentary characteristics of well logging data for relevant constraints, thus realizing the construction of a low-frequency model based on sedimentary characteristic constraints.
[0071] This invention, guided by earthquake inversion theory and based on knowledge of seismic wave dynamics and the inherent relationships between different data, constructs a closed-loop convolutional neural network. This convolutional neural network structure consists of an outer loop and an inner loop.
[0072] The outer loop is an unsupervised closed-loop convolutional network, and the execution process is as follows: Figure 2 As shown, a random sample of seismic data is subjected to convolutional encoding and decoding to achieve a sparse representation of the seismic data. During this process, the convolutional kernels for both encoding and decoding are optimized using batch gradient descent. The trained convolutional decoder functions as a seismic wavelet and participates in the inner loop seismic inversion process.
[0073] Seismic inversion requires well logging data modeling, well-seismic calibration wavelet extraction, and well logging constraints during the inversion process. To adapt seismic inversion to deep reservoirs with few or no wells, this paper utilizes the characteristics of different machine learning frameworks. Under the constraints of a prior model, a convolutional decoder in an unsupervised closed-loop convolutional neural network is used as a bridge to form the inner loop of a closed-loop convolutional network for intelligent deep reservoir inversion. This is a semi-supervised closed-loop convolutional network. The error gradient between the network output and the actual seismic data is calculated to update the low-frequency velocity model values, enabling intelligent inversion of sensitive parameters of deep reservoirs. Figure 3 As shown, this enables the prediction of gas content in reservoirs.
[0074] Example 3
[0075] Figure 4 A schematic diagram of a smart inversion profile of gas-bearing porosity of well 1 according to an embodiment of the present invention is shown.
[0076] Figure 5 A schematic diagram of a smart inversion profile of gas-bearing porosity of well 2 according to an embodiment of the present invention is shown.
[0077] This technology was used to predict the gas content of a deep carbonate reservoir. The intelligent inversion profiles of gas porosity in wells 1 and 2 are shown below. Figure 4 , Figure 5 As shown in the figure, the black curve represents the porosity curve containing gas. Observe... Figure 4 , Figure 5 It can be seen that the intelligent inversion results of gas-bearing porosity are consistent with the well logging calculation results, and the results show a zonal distribution in the horizontal direction with clear boundaries.
[0078] Example 4
[0079] Figure 6 A block diagram of a complex reservoir gas-bearing prediction device according to an embodiment of the present invention is shown.
[0080] like Figure 6 As shown, the complex reservoir gas-bearing prediction device includes:
[0081] Module 201 is used to construct a low-frequency velocity model based on seismic partitioning;
[0082] The encoding / decoding module 202 performs convolutional encoding and decoding of seismic data based on an unsupervised closed-loop convolutional network to obtain a sparse representation of the seismic data.
[0083] Inversion module 203, based on a semi-supervised closed-loop convolutional network, performs intelligent inversion of sensitive parameters of deep reservoirs to achieve gas content prediction.
[0084] In one example, a low-frequency velocity model based on seismic partitioning is constructed by using the Kriging interpolation algorithm and constraining the sedimentary features of well logging data.
[0085] In one example, the convolutional kernels for both encoding and decoding are optimized using batch gradient descent.
[0086] In one example, intelligent inversion of sensitive parameters of deep reservoirs based on a semi-supervised closed-loop convolutional network includes:
[0087] Based on the error gradient between the semi-supervised closed-loop convolutional network output and the actual seismic data, the low-frequency velocity model is updated to achieve intelligent inversion of sensitive parameters of deep reservoirs.
[0088] In one example, a semi-supervised closed-loop convolutional network is constructed using the convolutional decoder in an unsupervised closed-loop convolutional neural network as the seismic wavelet.
[0089] Example 5
[0090] This disclosure provides an electronic device, comprising: a memory storing executable instructions; and a processor executing the executable instructions in the memory to implement the aforementioned method for predicting the gas content of complex reservoirs.
[0091] An electronic device according to an embodiment of the present disclosure includes a memory and a processor.
[0092] This memory is used to store non-transitory computer-readable instructions. Specifically, the memory may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may, for example, include random access memory (RAM) and / or cache memory. The non-volatile memory may, for example, include read-only memory (ROM), hard disk, flash memory, etc.
[0093] The processor may be a central processing unit (CPU) or other form of processing unit with data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device to perform desired functions. In one embodiment of this disclosure, the processor is used to execute computer-readable instructions stored in the memory.
[0094] Those skilled in the art will understand that, in order to solve the technical problem of how to achieve a good user experience, this embodiment may also include well-known structures such as communication buses and interfaces, and these well-known structures should also be included within the protection scope of this disclosure.
[0095] For a detailed description of this embodiment, please refer to the corresponding descriptions in the foregoing embodiments, which will not be repeated here.
[0096] Example 6
[0097] This disclosure provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method for predicting the gas content of complex reservoirs.
[0098] A computer-readable storage medium according to embodiments of the present disclosure stores non-transitory computer-readable instructions. When these non-transitory computer-readable instructions are executed by a processor, all or part of the steps of the methods described in the foregoing embodiments of the present disclosure are performed.
[0099] The aforementioned computer-readable storage media include, but are not limited to: optical storage media (e.g., CD-ROM and DVD), magneto-optical storage media (e.g., MO), magnetic storage media (e.g., magnetic tape or portable hard drive), media with built-in rewritable non-volatile memory (e.g., memory card), and media with built-in ROM (e.g., ROM cartridge).
[0100] Those skilled in the art should understand that the above description of the embodiments of the present invention is only intended to illustrate the beneficial effects of the embodiments of the present invention, and is not intended to limit the embodiments of the present invention to any of the examples given.
[0101] The various embodiments of the present invention have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments.
Claims
1. A method for predicting gas content in complex reservoirs, characterized in that, The method comprises the following steps: constructing a low-frequency velocity model based on seismic division; performing convolutional encoding and decoding of seismic data according to an unsupervised closed-loop convolutional network to obtain a sparse representation of the seismic data; performing intelligent inversion of deep reservoir sensitive parameters based on a semi-supervised closed-loop convolutional network to realize prediction of reservoir gas content.
2. The method for complex reservoir gas-in-place prediction of claim 1, wherein, The low-frequency velocity model based on seismic division is constructed by using a Kriging interpolation algorithm and constraining sedimentary characteristics of well logging data.
3. The method for complex reservoir gas-in-place prediction of claim 1, wherein, The convolution kernel of the convolutional encoding and decoding is optimized by using a batch gradient descent method.
4. The method for complex reservoir gas-in-place prediction of claim 1, wherein, The intelligent inversion of deep reservoir sensitive parameters based on the semi-supervised closed-loop convolutional network comprises the following steps: calculating an error gradient of a network output and actual seismic data according to the semi-supervised closed-loop convolutional network, updating the low-frequency velocity model, and realizing intelligent inversion of deep reservoir sensitive parameters.
5. The method for complex reservoir gas-in-place prediction of claim 1, wherein, The semi-supervised closed-loop convolutional network is constructed by taking a convolutional decoder in the unsupervised closed-loop convolutional neural network as a seismic wavelet.
6. A device for predicting gas content in complex reservoirs, characterized in that, The method comprises the following steps: constructing a low-frequency velocity model based on seismic division by a constructing module; performing convolutional encoding and decoding of seismic data according to an unsupervised closed-loop convolutional network by a coding and decoding module to obtain a sparse representation of the seismic data; performing intelligent inversion of deep reservoir sensitive parameters based on a semi-supervised closed-loop convolutional network by an inversion module to realize prediction of reservoir gas content.
7. The apparatus for complex reservoir gas-in-place prediction of claim 6, wherein, The convolution kernel of the convolutional encoding and decoding is optimized by using a batch gradient descent method.
8. The apparatus for complex reservoir gas-in-place prediction of claim 6, wherein, The intelligent inversion of deep reservoir sensitive parameters based on the semi-supervised closed-loop convolutional network comprises the following steps: calculating an error gradient of a network output and actual seismic data according to the semi-supervised closed-loop convolutional network, updating the low-frequency velocity model, and realizing intelligent inversion of deep reservoir sensitive parameters.
9. An electronic device, comprising: The electronic device comprises: a memory storing executable instructions; a processor running the executable instructions in the memory to realize the complex reservoir gas content prediction method in any one of claims 1-5.
10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, which is executed by a processor to realize the complex reservoir gas content prediction method in any one of claims 1-5.