Deposited reservoir prediction method and system based on three-dimensional seismic data

Abstract

The invention discloses a sedimentary reservoir prediction method based on three-dimensional seismic data. The method comprises the steps of 1, feature extraction, 2, feature dynamic fusion, 3, model establishment and training, 4, storage prediction result optimization and 5, visual interaction. The system comprises a data processing module, a feature fusion module, a model training module, a reservoir prediction module and a visual interaction module. According to the method, multi-dimensional features such as seismic waveforms, frequency spectrums and geometric attributes are fused through a dynamic weight distribution algorithm, the vertical resolution of the thin interbed is remarkably improved, the recognition problem of the hidden reservoir is solved, and the leakage risk is reduced; a two-channel neural network model is established to automatically extract seismic features and sedimentary facies knowledge, the deviation of artificial feature engineering is avoided, the rapid modeling requirement of a complex sedimentary environment is met, a prediction result is constrained and corrected, and the continuity of a vertical sedimentary sequence is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of oil and gas exploration and development, and particularly to a method and system for sedimentary reservoir prediction based on three-dimensional seismic data. Background Art

[0002] In the field of oil and gas exploration, three-dimensional seismic data interpretation is the core technical means for reservoir prediction; the current mainstream prediction methods mainly rely on seismic attribute analysis, inversion technology and machine learning algorithms, aiming to infer the distribution of underground sedimentary facies and reservoir physical property parameters through seismic response characteristics; however, with the transfer of exploration targets to complex lithologies, deep-ultra deep and unconventional reservoirs, the existing technical system faces severe challenges in terms of prediction accuracy, geological consistency and calculation efficiency.

[0003] Traditional methods mostly use linear weighting to fuse seismic attributes (such as amplitude, frequency and geometric attributes), lacking the ability to model the non-linear correlation between attributes, resulting in errors in identification; at the same time, the existing intelligent prediction methods using CNN models and mainstream deep learning frameworks have obvious limitations. The input features of the former are usually manually screened by experts, resulting in the generalization ability of the model being limited by the geological characteristics of the training work area. The latter framework does not embed sedimentary dynamics laws, and there are problems with the prediction results violating the principle of geological space continuity.

[0004] For the identification of three-dimensional modeling, although the existing systems support three-dimensional modeling, the update of the geological knowledge base lags behind, and it is difficult to automatically identify special sedimentary patterns. Therefore, the present invention proposes a method and system for sedimentary reservoir prediction based on three-dimensional seismic data to solve the problems existing in the prior art. Summary of the Invention

[0005] In view of the above problems, the object of the present invention is to propose a method and system for sedimentary reservoir prediction based on three-dimensional seismic data. The method and system for sedimentary reservoir prediction based on three-dimensional seismic data fuse multi-dimensional features such as seismic waveform, spectrum and geometric attributes through a dynamic weight allocation algorithm, significantly improving the vertical resolution of thin interbeds, solving the problem of identifying subtle reservoirs, reducing the leakage risk; establishing a dual-channel neural network model to automatically extract seismic features and sedimentary facies knowledge, avoiding the deviation of artificial feature engineering, meeting the rapid modeling requirements of complex sedimentary environments, and correcting the prediction results to ensure the continuity of the vertical sedimentary sequence.

[0006] To achieve the object of the present invention, the present invention is realized through the following technical solutions: A method for sedimentary reservoir prediction based on three-dimensional seismic data, comprising the following steps:

[0007] Step 1: Feature extraction. Preprocess the collected 3D seismic data, and then extract multi-dimensional features including seismic waveform, spectrum, and geometric attributes from the preprocessed 3D seismic data.

[0008] Step 2: Feature dynamic fusion. Fusion the extracted multi-dimensional features through a dynamic weight allocation algorithm to generate optimized 3D seismic attribute data.

[0009] Step 3: Model establishment and training. Construct a dual-channel deep neural network model, and use known well control data and seismic data to construct a training set, and train the model using a deep learning algorithm.

[0010] Step 4: Optimization of storing prediction results. Input the seismic data of the area to be predicted into the trained model, and based on the 3D geological knowledge graph containing sedimentary patterns and diagenetic rules, generate spatial constraint conditions through a graph neural network to correct and optimize the prediction results output by the model, and obtain the prediction results of the sedimentary reservoir distribution in this area.

[0011] Step 5: Visualization interaction. Dynamically display the corrected reservoir prediction results through VR interaction in the interaction module, and provide uncertainty analysis, multi-scheme comparison, and real-time optimization selection of well location deployment.

[0012] Further improvement lies in: The data preprocessing in Step 1 includes using an improved Curvelet transform algorithm to suppress anisotropic noise of the data, using a median filtering method to filter the data, and performing amplitude compensation processing.

[0013] Further improvement lies in: The amplitude compensation processing is specifically as follows:

[0014] S1. First, perform seismic wave excision processing on the prestack seismic single-shot record data to obtain seismic first arrival wavelet data.

[0015] S2. Then, perform autocorrelation processing on the seismic first arrival wavelet data to obtain seismic record wavelet data.

[0016] S3. Subsequently, decompose the amplitude value of each seismic trace in the seismic record wavelet data to obtain an amplitude compensation factor.

[0017] S4. Finally, perform amplitude compensation on the prestack seismic single-shot record data according to the amplitude compensation factor.

[0018] Further improvement lies in: The dynamic weight allocation algorithm in Step 2 satisfies the following formula:

[0019]

[0020] Where S attris the seismic attribute matrix, W is the weight matrix, and S well is the seismic data matrix beside the well, F represents the Frobenius norm, Λ is the regularization matrix reflecting the geometric constraints between attributes, α and β are adaptive adjustment coefficients, and α + β = 1. α controls the matching accuracy between seismic attributes and well data, and β controls the strength of geological rule constraints.

[0021] A further improvement lies in that: the dual-channel deep neural network model in step three includes a first channel layer, a second channel layer, and a feature fusion layer. The first channel layer uses a 3D convolutional neural network to extract the spatial features of seismic data, the second channel layer uses a graph neural network to process the sedimentary facies knowledge graph, and the feature fusion layer introduces an attention mechanism to perform weighted fusion on the outputs of the first channel layer and the second channel layer.

[0022] A further improvement lies in that: the spatial constraint conditions in step four include vertical sedimentary sequence constraint and reservoir physical property parameter constraint. Among them, the vertical sedimentary sequence constraint is that for any spatial position (i, j, k), the prediction result yk+1 needs to satisfy the vertical transfer rule of sedimentary facies:

[0023]

[0024] where T(·) represents the preset set of sedimentary facies transfer relationships, and N t is the number of vertical sampling points;

[0025] The reservoir physical property parameter constraint is that the porosity and the wave impedance Z satisfy the following non-linear mapping relationship:

[0026]

[0027] The parameters a and b are calibration parameters obtained by calibrating well logging data. ∈ represents the random error term, and N(0, σ 2 ) represents the normal distribution, and σ 2 represents the variance, quantifying the error fluctuation range.

[0028] A sedimentary reservoir prediction system based on three-dimensional seismic data includes a data processing module, a feature fusion module, a model training module, a reservoir prediction module, and a visualization interaction module. The data processing module is used to process the collected three-dimensional seismic data and extract multi-dimensional features. The feature fusion module performs feature fusion processing based on dynamic weight allocation. The model training module is used to construct a dual-channel deep neural network model and train it. The storage prediction module is used to perform constraint optimization on the predicted results of the sedimentary reservoir distribution output by the model. The visualization interaction module dynamically interacts and displays the predicted results of the sedimentary reservoir distribution based on VR technology and provides multiple optimization selection schemes.

[0029] A further improvement lies in that: the model training module includes a model establishment sub-module, a training set construction sub-module, and a training sub-module. The model establishment sub-module establishes a dual-channel deep neural network model based on a 3D convolutional neural network, a graph neural network, and by introducing an attention mechanism. The training set construction sub-module constructs a training set based on known well control data and seismic data. The training sub-module trains the dual-channel deep neural network model based on a deep learning algorithm.

[0030] A further improvement lies in that: the storage prediction module includes a model application sub-module and a constraint correction sub-module. The model application sub-module imports the established dual-channel deep neural network model into the prediction system, uses the seismic data of the area to be predicted as input for prediction applications, and outputs the prediction results. The constraint correction sub-module generates spatial constraint conditions based on a three-dimensional geological knowledge graph through a graph neural network to correct and optimize the prediction results.

[0031] A further improvement lies in that: the visualization interaction module includes a VR visualization sub-module and an optimization selection sub-module. The VR visualization sub-module visually displays the prediction results of sedimentary reservoir distribution based on VR visualization technology. The optimization selection sub-module is used to provide uncertainty analysis, multi-scheme comparison, and real-time optimization selection schemes for well location deployment.

[0032] The beneficial effects of the present invention are as follows: The present invention combines multi-dimensional features such as seismic waveform, spectrum, and geometric attributes through a dynamic weight allocation algorithm, significantly improving the vertical resolution of thin interbeds, solving the problem of identifying subtle reservoirs, and reducing the leakage risk.

[0033] A dual-channel neural network model is established to automatically extract seismic features and sedimentary facies knowledge, avoiding the deviation of manual feature engineering, meeting the rapid modeling requirements of complex sedimentary environments, and correcting and constraining the prediction results based on a three-dimensional geological knowledge graph to ensure the continuity of the vertical sedimentary sequence and avoid economic losses caused by geological contradictions in well location deployment.

[0034] Finally, through VR interaction technology, "geologist - engineer - algorithm" collaborative decision-making is realized, improving the exploration success rate of complex blocks. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a flowchart of the method in Embodiment 1 of the present invention.

[0036] Figure 2 It is a system architecture diagram of Embodiment 2 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0037] In order to deepen the understanding of the present invention, the present invention will be further described in detail below in conjunction with embodiments. These embodiments are only used to explain the present invention and do not constitute a limitation to the protection scope of the present invention.

[0038] Example 1

[0039] According to Figure 1 As shown, this example provides a method for predicting sedimentary reservoirs based on three-dimensional seismic data, including the following steps:

[0040] Step 1: Preprocess the collected three-dimensional seismic data, and then extract multi-dimensional features including seismic waveform, spectrum, and geometric attributes from the preprocessed three-dimensional seismic data;

[0041] Among them, the preprocessing of the data includes suppressing anisotropic noise of the data using an improved Curvelet transform algorithm, filtering the data using a median filtering method, and performing amplitude compensation processing;

[0042] The amplitude compensation processing is specifically as follows:

[0043] S1: First, perform seismic wave excision processing on the prestack seismic single-shot record data to obtain seismic first arrival wavelet data;

[0044] S2: Then, perform autocorrelation processing on the seismic first arrival wavelet data to obtain seismic record wavelet data;

[0045] S3: Subsequently, decompose the amplitude value of each seismic trace in the seismic record wavelet data to obtain an amplitude compensation factor;

[0046] S4: Finally, perform amplitude compensation on the prestack seismic single-shot record data according to the amplitude compensation factor.

[0047] Step 2: Fusion the extracted multi-dimensional features through a dynamic weight allocation algorithm to generate optimized three-dimensional seismic attribute data;

[0048] Among them, the dynamic weight allocation algorithm satisfies the following formula:

[0049]

[0050] Where S attr is the seismic attribute matrix, with dimensions of M×K (M represents the spatial position, such as each seismic trace or voxel; K represents the seismic attribute category, such as instantaneous amplitude, curvature, spectral slope, etc.),

[0051] W is the weight matrix, with dimensions of K×P (K corresponds to the number of seismic attribute categories; P is the dimension of the target attribute after fusion, usually defined by geological requirements, such as reservoir thickness, porosity, etc.);

[0052] S well is the seismic data matrix beside the well, with the same dimensions as S attr Same;

[0053] F represents the Frobenius norm;

[0054] Λ is a regularization matrix for geometric constraints between reaction attributes, and the diagonal element Λ ii represents the geological contribution degree weight of the i-th type of attribute, and the off-diagonal elements can encode the spatial correlation between attributes (such as the correlation between the curvature attribute and the fault distance);

[0055] α and β are adaptive adjustment coefficients. α controls the matching accuracy between seismic attributes and well data, and β controls the intensity of geological rule constraints (the larger the value, the more the attribute fusion conforms to the prior geological rules). α + β = 1 for normalization to ensure optimization stability.

[0056] Step 3: Construct a dual-channel deep neural network model, and use the known well-controlled data and seismic data to construct a training set, and train the model using a deep learning algorithm;

[0057] The dual-channel deep neural network model includes a first channel layer, a second channel layer, and a feature fusion layer. The first channel layer uses a 3D convolutional neural network to extract the spatial features of seismic data. The second channel layer uses a graph neural network to process the sedimentary facies knowledge graph. The feature fusion layer introduces an attention mechanism to perform weighted fusion on the outputs of the first channel layer and the second channel layer.

[0058] Step 4: Input the seismic data of the area to be predicted into the trained model, and based on the three-dimensional geological knowledge graph containing sedimentary patterns and diagenetic rules, generate spatial constraint conditions through a graph neural network, and correct and optimize the prediction results output by the model to obtain the prediction results of the sedimentary reservoir distribution in this area;

[0059] The spatial constraint conditions include vertical sedimentary sequence constraints and reservoir physical property parameter constraints. Among them, the vertical sedimentary sequence constraint is that for any spatial position (i, j, k), the prediction result yk+1 needs to satisfy the sedimentary facies vertical transfer rule:

[0060]

[0061] where T(·) represents the preset set of sedimentary facies transfer relationships, and N t is the number of vertical sampling points, which determines the number of iterations of the vertical sequence constraint. For example, N t = 500, then it is necessary to perform sedimentary facies transfer verification on 499 adjacent horizons;

[0062] The reservoir physical property parameter constraint is that the porosity and the wave impedance Z satisfy the following non-linear mapping relationship:

[0063]

[0064] The parameters a and b are calibration parameters obtained by calibrating well logging data. Parameter a controls the sensitivity of wave impedance to porosity (e.g., a > 0 indicates that an increase in impedance is accompanied by a decrease in porosity), and parameter b is a correction term to eliminate the influence of impedance baseline shift.

[0065] ∈ represents the random error term, and N(0, σ 2 ) represents a normal distribution, where 0 is the mean, indicating that the error expectation is zero, and σ 2 represents the variance, which quantifies the error fluctuation range. σ is the standard deviation, which directly affects the width of the confidence interval for porosity prediction (e.g., σ = 0.5% means that the 95% confidence interval is ±1%).

[0066] Step Five: Dynamically display the corrected reservoir prediction results through VR interaction in the interaction module, and provide uncertainty analysis, multi - scenario comparison, and real - time optimization selection for well placement.

[0067] The three - dimensional dynamic rendering of VR interaction realizes volume rendering through GPU acceleration, providing three - dimensional visualization of the reservoir, well placement design, and real - time optimization functions.

[0068] Example 2

[0069] According to Figure 2 As shown, this example provides a sedimentary reservoir prediction system based on 3D seismic data, including a data processing module, a feature fusion module, a model training module, a reservoir prediction module, and a visualization interaction module. The data processing module is used to process the collected 3D seismic data and extract multi - dimensional features. The feature fusion module performs feature fusion processing based on dynamic weight assignment. The model training module is used to construct and train a dual - channel deep neural network model. The storage prediction module is used to constrain and optimize the predicted results of the sedimentary reservoir distribution output by the model. The visualization interaction module dynamically displays the predicted results of the sedimentary reservoir distribution based on VR technology and provides multiple optimization selection schemes.

[0070] The model training module includes a model establishment sub - module, a training set construction sub - module, and a training sub - module. The model establishment sub - module establishes a dual - channel deep neural network model based on 3D convolutional neural network, graph neural network, and by introducing an attention mechanism. The training set construction sub - module constructs a training set based on known well - controlled data and seismic data. The training sub - module trains the dual - channel deep neural network model based on deep learning algorithms.

[0071] The storage prediction module includes a model application sub - module and a constraint correction sub - module. The model application sub - module imports the established dual - channel deep neural network model into the prediction system, uses the seismic data of the area to be predicted as input for prediction application, and outputs the prediction results. The constraint correction sub - module generates spatial constraint conditions based on the three - dimensional geological knowledge graph and through graph neural network to correct and optimize the prediction results.

[0072] The visualization and interaction module includes a VR visualization sub-module and an optimization selection sub-module. The VR visualization sub-module visually displays the prediction results of the sedimentary reservoir distribution based on VR visualization technology, and realizes volume rendering through GPU acceleration. The optimization selection sub-module is used to provide uncertainty analysis, multi-scenario comparison, and real-time optimization selection solutions for well location deployment.

[0073] The systematic design makes the whole prediction process more automated and improves work efficiency.

[0074] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. What is described in the above embodiments and the specification only illustrates the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A sedimentary reservoir prediction method based on three-dimensional seismic data, characterized in that, It includes the following steps: Step 1, Feature extraction: Preprocess the collected 3D seismic data, and then extract multi-dimensional features including seismic waveform, spectrum, and geometric attributes from the preprocessed 3D seismic data; Step 2, Feature dynamic fusion: Fusion the extracted multi-dimensional features through a dynamic weight allocation algorithm to generate optimized 3D seismic attribute data; Step 3, Model establishment and training: Construct a two-channel deep neural network model, and use known well control data and seismic data to construct a training set, and train the model using a deep learning algorithm; Step 4, Optimization of storage prediction results: Input the seismic data of the area to be predicted into the trained model, and based on the 3D geological knowledge graph containing sedimentary patterns and diagenetic rules, generate spatial constraint conditions through a graph neural network to correct and optimize the prediction results output by the model, and obtain the prediction results of the sedimentary reservoir distribution in this area; Step 5, Visualization interaction: Dynamically display the corrected reservoir prediction results through VR interaction in the interaction module, and provide uncertainty analysis, multi-scheme comparison, and real-time optimization selection of well location deployment.

2. A method for predicting sedimentary reservoirs based on three-dimensional seismic data according to claim 1, characterized in that: The data preprocessing in Step 1 includes using an improved Curvelet transform algorithm to suppress anisotropic noise of the data, using a median filtering method to filter the data, and performing amplitude compensation processing.

3. The sedimentary reservoir prediction method based on three-dimensional seismic data according to claim 2, wherein: The amplitude compensation processing is specifically as follows: S1. First, perform seismic wave excision processing on the pre-stack seismic single-shot record data to obtain seismic first arrival wavelet data; S2. Then perform autocorrelation processing on the seismic first arrival wavelet data to obtain seismic record wavelet data; S3. Subsequently, decompose the amplitude value of each seismic trace in the seismic record wavelet data to obtain an amplitude compensation factor; S4. Finally, perform amplitude compensation on the pre-stack seismic single-shot record data according to the amplitude compensation factor.

4. A method for predicting sedimentary reservoirs based on 3D seismic data according to claim 1, characterized in that: The dynamic weight allocation algorithm in Step 2 satisfies the following formula Among which S attr is the seismic attribute matrix, W is the weight matrix, S well is the seismic data matrix beside the well, F represents the Frobenius norm, Λ is the regularization matrix reflecting the geometric constraints between attributes, α and β are adaptive adjustment coefficients, and α + β = 1. α controls the matching accuracy between seismic attributes and well data, and β controls the intensity of geological rule constraints.

5. A method for predicting sedimentary reservoirs based on 3D seismic data according to claim 1, characterized in that: The two-channel deep neural network model in Step 3 includes a first channel layer, a second channel layer, and a feature fusion layer. The first channel layer uses a 3D convolutional neural network to extract the spatial features of seismic data, the second channel layer uses a graph neural network to process the sedimentary facies knowledge graph, and the feature fusion layer introduces an attention mechanism to perform weighted fusion on the outputs of the first channel layer and the second channel layer.

6. A method for predicting sedimentary reservoirs based on three-dimensional seismic data according to claim 1, characterized in that: The spatial constraint conditions in Step 4 include vertical sedimentary sequence constraint and reservoir physical property parameter constraint. Among them, the vertical sedimentary sequence constraint is that for any spatial position (i, j, k), the prediction result yk+1 needs to satisfy the vertical transfer rule of sedimentary facies: where T(·) represents a preset set of sedimentary facies transfer relationships, and N t is the number of vertical sampling points; The reservoir physical property parameters are constrained to porosity and the wave impedance Z satisfies the following non-linear mapping relationship: Parameters a and b are calibration parameters obtained by calibrating well logging data. ∈ represents a random error term, and N(0, σ 2 ) represents a normal distribution, where σ 2 represents the variance, quantifying the error fluctuation range.

7. A sedimentary reservoir prediction system based on 3D seismic data, characterized in that: It includes a data processing module, a feature fusion module, a model training module, a reservoir prediction module, and a visualization interaction module. The data processing module is used to process the collected three-dimensional seismic data and extract multi-dimensional features. The feature fusion module performs feature fusion processing based on dynamic weight allocation. The model training module is used to construct and train a dual-channel deep neural network model. The storage prediction module is used to constrain and optimize the sedimentary reservoir distribution prediction results output by the model. The visualization interaction module dynamically interacts and displays the sedimentary reservoir distribution prediction results based on VR technology and provides multiple optimization selection schemes.

8. A sedimentary reservoir prediction system based on three-dimensional seismic data according to claim 7, characterized in that: The model training module includes a model establishment sub-module, a training set construction sub-module, and a training sub-module. The model establishment sub-module establishes a dual-channel deep neural network model based on 3D convolutional neural network, graph neural network, and by introducing an attention mechanism. The training set construction sub-module constructs a training set based on known well control data and seismic data. The training sub-module trains the dual-channel deep neural network model based on a deep learning algorithm.

9. The sedimentary reservoir prediction system based on 3D seismic data according to claim 7, characterized in that: The storage prediction module includes a model application sub-module and a constraint correction sub-module. The model application sub-module imports the established dual-channel deep neural network model into the prediction system, uses the seismic data of the area to be predicted as input for prediction application, and outputs the prediction results. The constraint correction sub-module generates spatial constraint conditions based on a three-dimensional geological knowledge graph and through a graph neural network to correct and optimize the prediction results.

10. A sedimentary reservoir prediction system based on three-dimensional seismic data according to claim 7, characterized in that: The visualization interaction module includes a VR visualization sub-module and an optimization selection sub-module. The VR visualization sub-module visually displays the sedimentary reservoir distribution prediction results based on VR visualization technology. The optimization selection sub-module is used to provide uncertainty analysis, multi-scheme comparison, and real-time optimization selection schemes for well location deployment.

Images