A seismic facies driven hydrocarbon source rock prediction method, device and equipment

Abstract

The present application relates to the technical field of oil and gas exploration, and discloses a method, device and equipment for predicting source rocks driven by seismic facies. The present application can generate a source rock facies classification data body in advance, and then perform weighted fusion after normalizing the source rock facies classification data body and a well-constrained model data body respectively, so that the obtained fusion data body can effectively guarantee the longitudinal and lateral resolution of the prediction result of the source rocks. The present application uses the fusion data body to generate a source rock sensitive factor data body for predicting the source rocks, so as to accurately depict the spatial distribution of the underground source rocks, enhance the identification ability of complex tectonic geological bodies, and improve the accuracy of the prediction of the spatial distribution of the source rocks.

Description

Technical Field

[0001] This invention relates to the field of oil and gas exploration technology, and in particular to a method, apparatus and equipment for predicting source rocks driven by seismic facies. Background Technology

[0002] Source rocks are sedimentary rocks rich in organic matter and are key reservoirs for the formation and preservation of oil and gas. Source rock prediction is of great significance for oil and gas exploration.

[0003] It should be noted that seismic data contains underlying information at different scales, and low-frequency data within seismic data includes geological information such as large-scale sedimentary background. Related techniques generally use low-frequency data to build an initial model, and then solve the inverse problem of the initial model to achieve hydrocarbon source rock prediction.

[0004] However, the initial models constructed using low-frequency data often show inconsistent stratigraphic attitudes between the model and the seismic profile in the vertical direction, and a "bull's eye" phenomenon around the well in the horizontal direction, which does not match the actual spatial distribution of the stratigraphy and results in low accuracy of source rock prediction. Summary of the Invention

[0005] This invention provides a method, apparatus, and equipment for predicting source rocks driven by seismic facies, in order to address the shortcomings of low accuracy in source rock prediction in related technologies and improve the accuracy of source rock prediction.

[0006] In a first aspect, the present invention proposes a method for predicting source rocks driven by seismic facies, the method comprising:

[0007] Multiple seismic attribute data volumes are generated using pre-stack seismic data from the target exploration area;

[0008] The multiple seismic attribute data volumes are input into a trained seismic facies classification model to classify source rocks, resulting in a source rock facies classification data volume output by the seismic facies classification model; wherein, the seismic facies classification model is obtained by machine learning based on multiple sample seismic attribute data volumes and a pre-trained classification model;

[0009] The source rock facies classification data volume and the well constraint model data volume are normalized respectively to obtain the normalized source rock data volume and the normalized well constraint model data volume.

[0010] The normalized source rock facies classification data volume and the normalized well constraint model data volume are weighted and fused to generate the first fused data volume to be verified.

[0011] If the first fused data volume matches the paleogeographic feature data volume, then the first fused data volume is determined to have passed verification, and the corresponding source rock sensitive factor data volume is predicted using the first fused data volume.

[0012] Optionally, the target data volume includes multiple three-dimensional attribute data, each of which includes three-dimensional spatial coordinate data and corresponding attribute values;

[0013] The target data volume is normalized to obtain a normalized target data volume, including:

[0014] The attribute values ​​in all three-dimensional attribute data of the target data volume are normalized to obtain a normalized target data volume. The normalized target data volume includes multiple normalized three-dimensional attribute data, and each normalized three-dimensional attribute data includes the three-dimensional spatial coordinate data and the normalized attribute value.

[0015] The target data volume is either the hydrocarbon source rock facies classification data volume or the well constraint model data volume.

[0016] Optionally, the weighted fusion of the normalized source rock facies classification data volume and the normalized well constraint model data volume to generate a first fused data volume to be verified includes:

[0017] Set a first weighting coefficient and a second weighting coefficient corresponding to the normalized source rock facies classification data body and the normalized well constraint model data body, respectively;

[0018] The first fused data volume is generated by weighting and summing the normalized source rock facies classification data volume and the normalized well constraint model data volume using the first weighting coefficient and the second weighting coefficient.

[0019] Optionally, after generating the first fused data volume, the method further includes:

[0020] If the first fused data volume does not match the paleogeographic feature data volume, then the first fused data volume is determined to have failed verification.

[0021] The first weighting coefficient and / or the second weighting coefficient are adjusted to obtain the adjusted weighting coefficient, and the adjusted weighting coefficient is used to perform a weighted summation on the normalized source rock data volume and the normalized well constraint model data volume to generate a second fused data volume to be verified.

[0022] If the second fused data body does not match the paleogeographic feature data body, it is determined that the second fused data body has failed the verification. The adjusted weighting coefficients are then adjusted until the newly generated fused data body matches the paleogeographic feature data body, and the newly generated fused data body is determined to have passed the verification.

[0023] The newly generated fused data volume is used to predict the corresponding source rock sensitive factor data volume.

[0024] Optionally, the step of predicting the corresponding source rock sensitive factor data volume using the first fused data volume includes:

[0025] Within the framework of Bayesian theory, the first fused data volume is inverted to generate the corresponding angular elastic impedance data volume;

[0026] The source rock sensitive factor data volume is generated by inversion using the angular elastic impedance data volume.

[0027] Optionally, after generating the source rock sensitivity factor data volume, the method further includes:

[0028] Obtain the wellbore curve corresponding to the source rock sensitive factor data volume;

[0029] Calculate the correlation value based on the wellbore curve and the actual well logging interpretation results;

[0030] If the correlation value is greater than the preset correlation threshold, then the source rock sensitive factor data volume is determined to be a highly reliable data volume.

[0031] Secondly, the present invention proposes a seismically driven hydrocarbon source rock prediction device, the device comprising:

[0032] The first generation unit is used to generate multiple corresponding seismic attribute data volumes using pre-stack seismic data of the target exploration area;

[0033] The input unit is used to input the multiple seismic attribute data volumes into a trained seismic facies classification model for source rock classification, and obtain the source rock data volume output by the seismic facies classification model; wherein, the seismic facies classification model is obtained by machine learning based on multiple sample seismic attribute data volumes and a pre-trained classification model;

[0034] The normalization unit is used to normalize the source rock facies classification data body and the well constraint model data body respectively, so as to obtain the normalized source rock facies classification data body and the normalized well constraint model data body.

[0035] The weighted fusion unit is used to perform weighted fusion on the normalized source rock facies classification data body and the normalized well constraint model data body to generate a first fused data body to be verified.

[0036] The first determining unit is configured to determine that the first fused data volume passes verification if the first fused data volume matches the paleogeographic feature data volume.

[0037] The prediction unit is used to predict the corresponding source rock sensitive factor data volume using the first fused data volume.

[0038] Optionally, the weighted fusion unit is further configured to set a first weighting coefficient and a second weighting coefficient corresponding to the normalized source rock data volume and the normalized well constraint model data volume, respectively.

[0039] The weighted fusion unit is further configured to use the first weighting coefficient and the second weighting coefficient to perform a weighted summation on the normalized source rock data volume and the normalized well constraint model data volume to generate the first fused data volume.

[0040] Optionally, the device further includes an adjustment unit:

[0041] The first determining unit is further configured to determine that the first fused data body has failed verification if the first fused data body does not match the paleogeographic feature data body;

[0042] The adjustment unit is used to adjust the first weighting coefficient and / or the second weighting coefficient to obtain the adjusted weighting coefficient;

[0043] The weighted fusion unit is further configured to use the adjusted weighting coefficients to perform a weighted summation on the normalized source rock data volume and the normalized well constraint model data volume to generate a second fused data volume to be verified.

[0044] The first determining unit is further configured to determine that the second fused data body has failed verification if the second fused data body does not match the paleogeographic feature data body;

[0045] The adjustment unit is also used to adjust the adjusted weighting coefficients until the newly generated fused data volume matches the paleogeographic feature data volume.

[0046] The first determining unit is further configured to determine that the newly generated fused data volume has passed verification;

[0047] The prediction unit is also used to predict the corresponding source rock sensitive factor data volume using the newly generated fused data volume.

[0048] Thirdly, the present invention proposes a computer device comprising: a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the seismically driven source rock prediction method of the first aspect or any corresponding embodiment described above.

[0049] The seismic facies-driven source rock prediction method, apparatus, and equipment proposed in this invention can generate multiple seismic attribute data volumes using pre-stack seismic data of the target exploration area; input these multiple seismic attribute data volumes into a trained seismic facies classification model to classify source rocks, obtaining a source rock facies classification data volume output by the seismic facies classification model; wherein, the seismic facies classification model is obtained by machine learning based on multiple sample seismic attribute data volumes of a pre-trained classification model; normalize the source rock facies classification data volume and the well constraint model data volume respectively, obtaining normalized source rock facies classification data volume and normalized well constraint model data volume; perform weighted fusion of the normalized source rock facies classification data volume and the normalized well constraint model data volume to generate a first fused data volume to be verified; if the first fused data volume matches the paleogeographic feature data volume, the first fused data volume is determined to pass verification, and the corresponding source rock sensitive factor data volume is predicted using the first fused data volume. This invention can first generate a source rock data volume, and then perform weighted fusion on the source rock data volume and the well constraint model data volume after normalization. The resulting fused data volume can effectively ensure the vertical and horizontal resolution of the distribution map. Using the fused data volume to predict source rocks and generate source rock sensitive factor data volume can effectively achieve accurate characterization of the spatial distribution of underground source rocks, enhance the ability to identify complex geological structures, and improve the accuracy of source rock spatial distribution prediction. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0051] Figure 1 One of the flowcharts for a seismically driven hydrocarbon source rock prediction method provided in an embodiment of the present invention;

[0052] Figure 2 The second flowchart of the seismically driven source rock prediction method provided in the embodiments of the present invention;

[0053] Figure 3The third flowchart of the seismically driven source rock prediction method provided in the embodiments of the present invention;

[0054] Figure 4 A machine learning phase classification profile provided in an embodiment of the present invention;

[0055] Figure 5 An initial model profile diagram established using conventional well constraint model data volume is provided as an embodiment of the present invention;

[0056] Figure 6 A cross-sectional view of a fusion model modeled using a first fusion data volume, provided as an embodiment of the present invention;

[0057] Figure 7 A cross-sectional view corresponding to an initial model of elastic impedance at a small angle provided in an embodiment of the present invention;

[0058] Figure 8 A cross-sectional view corresponding to an initial model of elastic impedance at a medium angle provided in an embodiment of the present invention;

[0059] Figure 9 A cross-sectional view corresponding to an initial model of elastic impedance at a large angle provided in an embodiment of the present invention;

[0060] Figure 10 A cross-sectional view corresponding to the elastic impedance inversion result of small-angle seismic data provided in an embodiment of the present invention;

[0061] Figure 11 A cross-sectional view corresponding to the elastic impedance inversion result of mid-angle seismic data provided in an embodiment of the present invention;

[0062] Figure 12 A cross-sectional view corresponding to the elastic impedance inversion result of large-angle seismic data provided in an embodiment of the present invention;

[0063] Figure 13 A cross-sectional view of a source rock sensitive factor data volume obtained by elastic impedance inversion from three angles, provided as an embodiment of the present invention;

[0064] Figure 14 This is a schematic diagram of the structure of a seismically driven hydrocarbon source rock prediction device provided in an embodiment of the present invention;

[0065] Figure 15 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0066] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0067] The following is combined with Figures 1-13 This invention describes a seismically driven method for predicting hydrocarbon source rocks.

[0068] like Figure 1 As shown, this embodiment proposes a first method for predicting source rocks driven by seismic facies, which may include the following steps:

[0069] S101. Generate multiple corresponding seismic attribute data volumes using pre-stack seismic data of the target exploration area.

[0070] Specifically, the target exploration area can be an undrilled area that needs to be explored for oil and gas.

[0071] Specifically, in this embodiment, pre-stack seismic data of the target exploration area can be collected first, and then processed to generate corresponding post-stack seismic data. The pre-stack seismic data needs to undergo dynamic correction, static correction, and denoising before being stacked at partial incident angles to obtain partial stacking angle gather data, i.e., post-stack seismic data, to maintain a high signal-to-noise ratio and fidelity. Subsequently, this embodiment can use the post-stack seismic data to calculate more than 40 attribute data volumes, including amplitude-based attributes, frequency-based attributes, time-spectrum attributes, and absorption attenuation attributes.

[0072] It should be noted that each attribute data volume calculated in this embodiment is a basic data volume that can be used for seismic facies classification.

[0073] Specifically, a seismic attribute data volume can include multiple three-dimensional attribute data, each of which includes three-dimensional spatial coordinate data and corresponding attribute values.

[0074] It is understood that the data structure of all data volumes in this embodiment can be similar to that of the seismic attribute data volume, that is, it consists of three-dimensional spatial coordinate data and corresponding attribute values ​​(parameter values).

[0075] S102. Input multiple seismic attribute data volumes into the trained seismic facies classification model to classify source rocks, and obtain the source rock facies classification data volume output by the seismic facies classification model; wherein, the seismic facies classification model is obtained by machine learning based on the pre-trained classification model using multiple sample seismic attribute data volumes.

[0076] It should be noted that in this embodiment, sample seismic attribute data volumes can be collected first in the drilled area. The seismic facies categories corresponding to these sample seismic attribute data volumes, such as source rocks (medium-deep lakes, shallow lakes, etc.) and other non-source rocks (e.g., deltas, channels, etc.), are then identified. The sample seismic attribute data volumes identifying source rocks are selected based on paleogeomorphological analysis, prioritizing attributes that reflect the external shape, internal characteristics, and evolutionary patterns of medium-deep lakes, shallow lakes, and deltas. Subsequently, the identified sample seismic attribute data volumes are used to perform machine learning on a pre-trained classification model, enabling the model to filter out source rock facies classification data volumes from the sample seismic attribute data volumes. When the classification model's classification ability meets the requirements, this embodiment can then identify it as the trained seismic facies classification model.

[0077] Specifically, in this embodiment, the generated multiple seismic attribute data volumes can be input into a trained seismic facies classification model for source rock classification, so that the seismic facies classification model can select source rock facies classification data volumes from the multiple seismic attribute data volumes.

[0078] S103. Normalize the source rock facies classification data volume and the well constraint model data volume respectively to obtain the normalized source rock data volume and the normalized well constraint model data volume.

[0079] It should be noted that this embodiment can collect stratigraphic and well location data of the target exploration area, and construct a well constraint model data volume based on the stratigraphic and logging data. The stratigraphic data needs to be smooth and accurate, and the logging data needs to include data such as P-wave and S-wave velocities, density, and sonic transit time.

[0080] Optionally, the target data volume includes multiple three-dimensional attribute data, each of which includes three-dimensional spatial coordinate data and corresponding attribute values;

[0081] The target data volume is normalized to obtain the normalized target data volume, which includes:

[0082] The attribute values ​​in all three-dimensional attribute data of the target data volume are normalized to obtain the normalized target data volume. The normalized target data volume includes multiple normalized three-dimensional attribute data, and each normalized three-dimensional attribute data includes three-dimensional spatial coordinate data and normalized attribute values.

[0083] The target data volume is either a hydrocarbon source rock facies classification data volume or a well constraint data model data volume.

[0084] It should be noted that both the source rock facies classification data volume and the well constraint model data volume can include multiple three-dimensional attribute data, and the three-dimensional spatial data of the three-dimensional attribute data in the source rock data volume and the well constraint model data volume can be consistent. It is understood that this embodiment can construct the well constraint model data volume based on the three-dimensional spatial data of the three-dimensional attribute data in the source rock data volume, as well as based on the aforementioned stratigraphic data and well location data.

[0085] It should be noted that seismic records can reflect characteristics such as the lithology and contact relationships of subsurface media. Different sedimentary bodies exhibit varying amplitudes, frequencies, and continuity on seismic profiles. Based on differences in reflection characteristics, the classification results of multiple seismic facies in three-dimensional space obtained through machine learning are encoded and characterized according to different seismic facies. The three-dimensional distribution of different codes effectively reveals the distribution of different seismic facies in three-dimensional space. Different characterization codes exist at the boundaries of different seismic facies, and changes in the encoded data indicate sedimentary changes, i.e., sedimentary boundaries. These sedimentary boundaries can effectively characterize special subsurface geological bodies and enhance the lateral resolution of seismic inversion modeling.

[0086] S104. The normalized source rock facies classification data volume and the normalized well constraint model data volume are weighted and fused to generate the first fused data volume to be verified.

[0087] Specifically, the first fused data volume is the data volume obtained by weighted fusion of the normalized source rock facies classification data volume and the normalized well constraint model data volume in this embodiment.

[0088] Optional, such as Figure 2 As shown, step S104 may include steps S1041 and S1042. Wherein:

[0089] S1041. Set the first weighting coefficient and the second weighting coefficient corresponding to the normalized source rock facies classification data volume and the normalized well constraint model data volume, respectively.

[0090] S1042. Using the first weighting coefficient and the second weighting coefficient, the normalized source rock facies classification data volume and the normalized well constraint model data volume are weighted and summed to generate the first fused data volume.

[0091] Specifically, the first weighting coefficient is the weighting coefficient of the normalized source rock facies classification data body, and the second weighting coefficient is the weighting coefficient of the normalized well constraint model data body.

[0092] It is understandable that weighted summation refers to the weighted summation of attribute values ​​of the same three-dimensional spatial data in two data volumes.

[0093] S105. If the first fused data volume matches the paleogeographic feature data volume, then the first fused data volume is determined to have passed the verification.

[0094] Specifically, the three-dimensional spatial data of the paleogeographic feature data volume and the three-dimensional attribute data in the first fused data volume can be consistent.

[0095] Specifically, in this embodiment, the paleogeographic feature data body can be determined based on the three-dimensional spatial data of the three-dimensional attribute data in the first fused data body.

[0096] It should be noted that when the first fused data volume matches the paleogeographic feature data volume, this embodiment can determine that the first fused data volume has passed verification and has a certain data accuracy for predicting source rocks.

[0097] S106. Predict the corresponding source rock sensitive factor data volume using the first fused data volume.

[0098] Specifically, in this embodiment, the first fused data volume can be verified, and the first fused data volume can be used to predict the source rock sensitive factor data volume.

[0099] Specifically, in this embodiment, after the first fused data volume is verified, the first fused data volume can be used to predict source rocks and obtain the corresponding source rock sensitive factor data volume.

[0100] It should be noted that related technologies generally only use initial models constructed from low-frequency data, such as conventional well-constrained model data volumes, for source rock prediction. These initial models use variograms to simply describe the correlation between two points underground, thus characterizing the spatial structure of sedimentary strata. This approach struggles to accurately represent the structural characteristics of semi-deep lacustrine organic-rich mudstone in complex tectonic regions. Furthermore, the distribution maps corresponding to these initial models exhibit inconsistencies between the stratigraphic attitudes shown in the model vertically and those on seismic profiles, and a "bull's-eye" phenomenon around the well horizontally, failing to reflect the actual spatial distribution of strata. This embodiment, however, can first generate a source rock data volume. After normalizing both the source rock facies classification data volume and the well-constrained model data volume, a weighted fusion is performed. The resulting fused data volume significantly improves both vertical and horizontal resolution. Using this fused data volume to generate source rock sensitive factor data volumes effectively and accurately characterizes the spatial distribution of underground source rocks, enhances the identification of complex tectonic geological bodies, and improves the accuracy of source rock spatial distribution prediction.

[0101] The seismic facies-driven source rock prediction method proposed in this embodiment can generate multiple seismic attribute data volumes using pre-stack seismic data of the target exploration area. These seismic attribute data volumes are then input into a trained seismic facies classification model to classify source rocks, resulting in a source rock facies classification data volume output by the model. The seismic facies classification model is obtained by machine learning from a pre-trained classification model based on multiple sample seismic attribute data volumes. The source rock facies classification data volume and the well-constrained model data volume are normalized to obtain normalized source rock facies classification data volume and normalized well-constrained model data volume, respectively. The normalized source rock facies classification data volume and the normalized well-constrained model data volume are then weighted and fused to generate a first fused data volume to be verified. If the first fused data volume matches the paleogeographic feature data volume, the first fused data volume is verified, and the corresponding source rock sensitive factor data volume is predicted using the first fused data volume. In this embodiment, a source rock facies classification data volume can be generated first. After normalizing the source rock facies classification data volume and the well constraint model data volume respectively, a weighted fusion is performed. The resulting fused data volume can effectively ensure vertical and horizontal resolution. Using the fused data volume to predict source rocks and generate source rock sensitive factor data volume can effectively achieve accurate characterization of the spatial distribution of underground source rocks, enhance the ability to identify complex tectonic geological bodies, and improve the accuracy of source rock spatial distribution prediction.

[0102] based on Figure 1 This embodiment proposes a second method for predicting source rocks driven by seismic facies. This method, after step S104 above, may further include:

[0103] If the first fused data volume does not match the paleogeographic feature data volume, then the first fused data volume is determined to have failed the verification.

[0104] The first weighting coefficient and / or the second weighting coefficient are adjusted to obtain the adjusted weighting coefficient. The adjusted weighting coefficient is then used to perform a weighted summation on the normalized source rock facies classification data volume and the normalized well constraint model data volume to generate the second fused data volume to be verified.

[0105] If the second fused data volume does not match the paleogeographic feature data volume, the second fused data volume is determined to have failed the verification. The adjusted weighting coefficients are then adjusted until the newly generated fused data volume matches the paleogeographic feature data volume, and the newly generated fused data volume is determined to have passed the verification.

[0106] The newly generated fused data volume is used to predict the corresponding source rock sensitive factor data volume.

[0107] Specifically, when the fused data volume does not match the paleogeographic feature data volume, this embodiment can determine that the accuracy of the fused data volume is insufficient and that it has failed the verification.

[0108] The seismic-driven source rock prediction method proposed in this embodiment can adjust the weighting coefficients of the normalized source rock facies classification data volume and the normalized well-constrained model data volume when the weighted fusion data volume does not match the paleogeomorphological feature data volume. This adjustment continues until the normalized source rock facies classification data volume and the normalized well-constrained model data volume can be weighted and fused to produce a fusion data volume that matches the paleogeomorphological feature data volume. This ensures the horizontal and vertical resolution of the distribution map corresponding to the fusion data volume, thereby ensuring the accurate depiction of the spatial distribution of underground source rocks, enhancing the ability to identify complex tectonic geological bodies, and further improving the accuracy of source rock spatial distribution prediction.

[0109] based on Figure 1 This embodiment proposes a third method for predicting source rocks driven by seismic facies. In this method, step S106 above may include:

[0110] Within the framework of Bayesian theory, inversion is performed based on the first fused data volume to generate the corresponding angular elastic impedance data volume.

[0111] The source rock sensitivity factor data volume is generated by using the angular elastic impedance data volume for calculation.

[0112] It should be noted that this embodiment, within the framework of Bayesian theory, assumes that the noise concentrated in the partial angular superimposed seismic traces is independent and conforms to a Gaussian distribution, thus obtaining the likelihood function. Then, the likelihood function is multiplied by the prior probability to obtain the posterior probability distribution. Maximizing the posterior probability distribution yields the objective function. Differentiating the objective function with respect to the model parameters and setting the derivative to zero yields the inversion equation. Inversion based on the first fused data volume yields the angular elastic impedance data volume.

[0113] Specifically, this embodiment can establish the relationship between the angular elastic impedance data volume and the source rock sensitive factor data volume through linear rock physics model analysis, and use statistical algorithms for clustering and density estimation, such as the Expectation-Maximization (EM) Gaussian Mixture Model (GM) algorithm, to achieve statistical rock physics inversion of the source rock sensitive factor data volume from the angular elastic impedance data volume.

[0114] It should also be noted that, based on rock physical analysis, this embodiment can clarify the relationship between elastic impedance and source rock sensitivity factor, and use the EM-GM algorithm to realize the statistical rock physical inversion from elastic impedance to source rock sensitivity factor.

[0115] In practical applications, the initial model is the most crucial part of the seismic inversion process. The richer the low-frequency information considered, the more reliable the inversion results. Related techniques typically involve interpolating and extrapolating processed well logging data under stratigraphic constraints to obtain a low-frequency model. This method is only suitable for structurally simple and relatively flat strata. Due to its low lateral resolution and the "bull's-eye effect" (also known as the trend effect), related techniques are generally ineffective in identifying complex underground geological bodies. The purpose of this embodiment is to start with pre-stack seismic data, use machine learning methods to classify seismic attribute data volumes, and then use stratigraphic constraints and well logging data interpolation to obtain an initial model, namely the aforementioned well-constrained model data volume. Through a data fusion algorithm, the normalized seismic facies classification data volume and the well-constrained model data volume are fused to generate a fused data volume with high lateral and vertical resolution. Then, an inversion algorithm can be used to obtain seismic prediction results, thereby achieving the goal of hydrocarbon source rock prediction.

[0116] It is understood that this embodiment proposes a pre-stack seismic inversion technique driven by machine learning seismic facies. This inversion technique makes the inversion results more consistent with the actual geological structure, enhances the accuracy and stability of interpreting geological bodies in complex tectonic regions, and can be used for source rock prediction, thus improving the accuracy of evaluating high-quality source rocks. Compared with conventional methods, the machine learning seismic facies-driven seismic inversion technique constructed in this embodiment considers more low-frequency information and enhances the ability to identify complex tectonic geological bodies. It can solve the "bull's-eye effect" in conventional methods, improve the lateral resolution of seismic interpretation results, and enhance the stability and reliability of interpretation results based on pre-stack seismic data while better conforming to the geological background.

[0117] The seismic-driven source rock prediction method proposed in this embodiment can obtain a source rock sensitive factor data volume by inversion based on the first fused data volume, effectively ensuring the prediction of source rocks based on the first fused data volume and ensuring the accuracy of source rock prediction.

[0118] like Figure 3 As shown in this embodiment, another method for predicting source rocks driven by seismic facies is proposed, which may include:

[0119] Prestack seismic data from both drilled and undrilled wells within a designated area are acquired. Attribute analysis is performed on the prestack seismic data from drilled areas, generating multiple seismic attribute data volumes. Seismic facies analysis is then conducted in conjunction with geological research findings. Based on the research and facies analysis results, seismic reflection characteristics reflecting source rocks are selected. Various seismic facies reflecting source rocks (medium-deep lakes, shallow lakes, etc.) and non-source rocks (deltas, channels, etc.) in drilled areas are manually labeled and categorized, and used as training samples to train a seismic facies classification model using machine learning algorithms. For undrilled prestack seismic data, multiple seismic attribute data volumes are calculated and input into the seismic facies classification model for source rock classification, yielding the aforementioned source rock facies classification data volumes.

[0120] This embodiment can acquire stratigraphic and well data within a specified area, and construct a well data interpolation model, i.e., the aforementioned well-constrained model data volume, based on the stratigraphic and well data. The source rock facies classification data volume and the well-constrained model data volume are then normalized to obtain normalized source rock facies classification and well-constrained model data volumes. These normalized source rock facies classification and well-constrained model data volumes are then weighted and fused to obtain the initial fused model, i.e., the first fused data volume. If the first fused data volume matches the paleogeomorphological feature data volume, the first fused data volume is deemed to have passed verification. Elastic impedance pre-stack seismic inversion is then performed using the first fused data volume to sequentially obtain the angular elastic impedance data volume and the source rock sensitivity factor data volume.

[0121] based on Figure 1 This embodiment proposes a fourth method for predicting source rocks driven by seismic facies. This method, after step S106, may further include:

[0122] Obtain the wellbore curves corresponding to the source rock sensitive factor data volume;

[0123] Calculate the correlation value based on the wellside curve and the actual well logging interpretation results;

[0124] If the correlation value is greater than the preset correlation threshold, the source rock sensitive factor data volume is determined to be a high-confidence data volume.

[0125] Specifically, this embodiment can compare the source rock sensitive factor data with the actual well logging interpretation results of drilled wells to ensure that the prediction results are consistent with the actual underground geological conditions, thereby achieving quality control of the prediction results.

[0126] Specifically, in this embodiment, the wellbore curve corresponding to the source rock sensitive factor data volume can be extracted, and the actual logging interpretation results of the drilled well can be correlated with the curve. When the correlation value is greater than the preset correlation threshold, such as 85%, this embodiment can confirm that the prediction result is highly reliable.

[0127] It should also be noted that this embodiment can compare the source rock sensitive factor prediction data volume with the actual well logging interpretation results, and select the source rock sensitive factor data volume with the highest inversion quality as the basis for source rock identification, so as to realize the prediction of pre-stack seismic source rocks driven by seismic facies and effectively ensure the accuracy of the prediction.

[0128] The seismic-driven source rock prediction method proposed in this embodiment can perform correlation analysis between the source rock sensitive factor data volume and the actual well logging interpretation results of drilled wells after obtaining the source rock sensitive factor data volume. When the correlation value is greater than the preset correlation threshold, the source rock sensitive factor data volume is confirmed to be highly reliable, thereby further improving the accuracy of source rock prediction.

[0129] To better illustrate the beneficial effects of this embodiment compared to related technologies, this embodiment proposes... Figures 4 to 13 Please provide an explanation. Figures 4 to 13 The vertical axis Time[s] represents the two-way travel time of the seismic wave, the horizontal axis Trace Number represents the seismic data trace number, Seismic Face indicates the seismic facies classification, EI refers to elastic wave impedance (unit: kg / ^3*m / s), WELL-A refers to well number A, and IF hrs refers to the source rock sensitivity factor.

[0130] in, Figures 4 to 6 These are, respectively, profile diagrams for machine learning phase classification, initial model built using conventional well-constrained model data volumes, and fusion model built using the first fusion data volume. The initial model profile diagram is the profile diagram corresponding to the initial model, and the fusion model profile diagram is the profile diagram corresponding to the fusion model. Figure 4 In the diagram, different color depth areas can be used to identify different seismic facies classification codes, namely good source rock seismic facies, poor source rock seismic facies, and non-source rock seismic facies. Figure 5 and Figure 6 In the diagram, different color depth areas can be used to indicate the range of elastic impedance values.

[0131] Specifically, Figure 4 The results of several seismic facies classifications obtained through machine learning (low-frequency continuous strong amplitude, low-frequency medium amplitude subparallel reflection, and others) are introduced as new inversion constraints into the initial model established using conventional well constraint model data volume, thus obtaining the fusion model corresponding to the first fusion data volume. Figure 5 The initial model profile obtained by conventional well constraint modeling shows that the lateral resolution is not high. The effect is good near well A, but the well data constraint capability weakens and the resolution decreases as the well moves away from the well location. Figure 6 A cross-sectional view of the fusion model is shown, from which it can be seen that... Figure 6 Overcame Figure 5The problem of low lateral resolution of the same phase axis in the mid-section far from the well location.

[0132] Figures 7 to 9 The images show cross-sectional views of the initial elastic impedance models at small, medium, and large angles, demonstrating that these models achieve good lateral resolution after incorporating machine learning seismic facies classification results. The fused model, which integrates machine learning seismic facies classification results, maintains high resolution at distant well locations while also ensuring a high degree of agreement between the model and well data near the well location, proving the rationality of the fused initial model and meeting the requirements for further seismic inversion.

[0133] Figures 10 to 12 The results of elastic impedance retrieval (EIP) from small-angle, medium-angle, and large-angle seismic data are presented respectively. As shown in the figures, the EIP results at all three angles exhibit high resolution at distant well locations, while also maintaining a high degree of agreement with the well data near well A. This demonstrates the rationality of the EIP results and meets the requirements for further inversion of source rock sensitivity factors. Based on the profile diagrams corresponding to the angle EIP retrieval results in this embodiment, the spatial distribution of organic-rich mudstone can be clearly depicted in stages, showcasing the lateral variations of source rocks in different geological periods.

[0134] Figure 13 This is a cross-sectional view of the source rock sensitive factor data volume obtained from elastic impedance inversion at three angles, which can be used for source rock prediction and evaluation. The color scales in the figure represent the numerical distribution range of attribute values ​​in the source rock sensitive factor data volume. From... Figure 13 As can be seen, the results of the source rock sensitivity factor inversion are in good agreement with the well data. Based on the source rock sensitivity factor results obtained by elastic impedance inversion, the spatial distribution of organic mudstone can be clearly depicted in stages, showing the lateral variation of source rocks in different geological periods.

[0135] Specifically, this embodiment can calculate the source rock sensitivity factor based on elastic impedance from three angles, which can be used for the prediction and evaluation of source rocks. Comparison with actual well logging interpretation results shows that the inversion results have a high degree of agreement with the actual well logging interpretation results, and the correlation between the inversion results and the actual well logging interpretation results is greater than a preset correlation threshold.

[0136] This embodiment uses cross-sectional views of seismic facies classification data, an initial model created by related technologies, and a first fused data volume to determine the cross-sectional view corresponding to the first fused data volume, thus overcoming the problem of low lateral resolution of the same phase axis on the cross-sectional view corresponding to the initial model created by related technologies.

[0137] like Figure 14As shown in the figure, this embodiment proposes a seismically driven source rock prediction device, which may include:

[0138] The first generation unit 101 is used to generate multiple corresponding seismic attribute data volumes using pre-stack seismic data of the target exploration area;

[0139] Input unit 102 is used to input multiple seismic attribute data volumes into a trained seismic facies classification model for source rock classification, and obtain the source rock facies classification data volume output by the seismic facies classification model; wherein, the seismic facies classification model is obtained by machine learning based on multiple sample seismic attribute data volumes and a pre-trained classification model;

[0140] Normalization unit 103 is used to normalize the source rock facies classification data volume and the well constraint model data volume respectively, so as to obtain the normalized source rock data volume and the normalized well constraint model data volume.

[0141] The weighted fusion unit 104 is used to perform weighted fusion on the normalized source rock facies classification data volume and the normalized well constraint model data volume to generate the first fused data volume to be verified.

[0142] The first determining unit 105 is used to determine that the first fused data volume passes the verification if the first fused data volume matches the paleogeographic feature data volume.

[0143] The prediction unit is used to predict the corresponding source rock sensitive factor data volume using the first fused data volume.

[0144] It should be noted that the processing procedures and beneficial effects of the first generation unit 101, input unit 102, normalization unit 103, weighted fusion unit 104, first determination unit 105, and prediction unit can be referred to respectively. Figure 1 The relevant explanations for each step are not repeated here.

[0145] Optionally, the target data volume includes multiple three-dimensional attribute data, each of which includes three-dimensional spatial coordinate data and corresponding attribute values;

[0146] The target data volume is normalized to obtain the normalized target data volume, which is set as follows:

[0147] The attribute values ​​in all three-dimensional attribute data of the target data volume are normalized to obtain the normalized target data volume. The normalized target data volume includes multiple normalized three-dimensional attribute data, and each normalized three-dimensional attribute data includes three-dimensional spatial coordinate data and normalized attribute values.

[0148] The target data volume is either a hydrocarbon source rock facies classification data volume or a well constraint model data volume.

[0149] Optionally, the weighted fusion unit 104 is also used to set a first weighting coefficient and a second weighting coefficient corresponding to the normalized source rock facies classification data body and the normalized well constraint model data body, respectively.

[0150] The weighted fusion unit 104 is also used to perform weighted summation on the normalized source rock facies classification data volume and the normalized well constraint model data volume using the first weighting coefficient and the second weighting coefficient to generate the first fused data volume.

[0151] Optionally, the device may also include an adjustment unit:

[0152] The first determining unit 105 is further configured to determine that the first fused data body has failed verification if the first fused data body does not match the paleogeographic feature data body.

[0153] An adjustment unit is used to adjust the first weighting coefficient and / or the second weighting coefficient to obtain the adjusted weighting coefficient;

[0154] The weighted fusion unit 104 is also used to perform weighted summation on the normalized source rock facies classification data volume and the normalized well constraint model data volume using the adjusted weighting coefficients to generate a second fusion data volume to be verified.

[0155] The first determining unit 105 is further configured to determine that the second fused data body has failed verification if the second fused data body does not match the paleogeographic feature data body;

[0156] The adjustment unit is also used to adjust the weighting coefficients until the newly generated fused data volume matches the paleogeographic feature data volume.

[0157] The first determining unit 105 is also used to determine whether the newly generated fused data volume passes verification.

[0158] The prediction unit is also used to predict the corresponding source rock sensitive factor data volume using the newly generated fused data volume.

[0159] Optionally, the prediction unit is also used to: perform inversion based on the first fused data volume within the Bayesian theoretical framework to generate the corresponding angular elastic impedance data volume; and use the angular elastic impedance data volume to perform calculations to generate the source rock sensitivity factor data volume.

[0160] Optionally, the above-mentioned device further includes:

[0161] The acquisition unit is used to acquire the wellbore curves corresponding to the source rock sensitive factor data volume;

[0162] The calculation unit is used to calculate the correlation value based on the wellside curve and the actual well logging interpretation results;

[0163] The second determining unit is used to determine the source rock sensitive factor data volume as a high-confidence data volume if the correlation value is greater than the preset correlation threshold.

[0164] This embodiment proposes a seismic facies-driven source rock prediction device, which can generate multiple seismic attribute data volumes using pre-stack seismic data of the target exploration area. These seismic attribute data volumes are then input into a trained seismic facies classification model for source rock facies classification, resulting in a source rock facies classification data volume output by the seismic facies classification model. The seismic facies classification model is obtained through machine learning on a pre-trained classification model based on multiple sample seismic attribute data volumes. The source rock facies classification data volume and the well constraint model data volume are normalized to obtain normalized source rock facies classification data volume and normalized well constraint model data volume, respectively. The normalized source rock facies classification data volume and the normalized well constraint model data volume are then weighted and fused to generate a first fused data volume to be verified. If the first fused data volume matches the paleogeographic feature data volume, the first fused data volume is verified, and the corresponding source rock sensitive factor data volume is predicted using the first fused data volume. In this embodiment, a source rock data volume can be generated first. After normalizing the source rock data volume and the well constraint model data volume respectively, a weighted fusion is performed. The resulting fused data volume can effectively ensure vertical and horizontal resolution. Using the fused data volume to generate source rock sensitive factor data volume for source rock prediction can effectively achieve accurate characterization of the spatial distribution of underground source rocks, enhance the ability to identify complex geological structures, and improve the accuracy of source rock spatial distribution prediction.

[0165] In this embodiment, the seismic phase-driven source rock prediction device is presented in the form of a functional unit. Here, a unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0166] This invention also provides a computer device having the above-described features. Figure 14 The device shown is a seismically driven source rock prediction device.

[0167] Please see Figure 15 , Figure 15 This is a schematic diagram of the structure of a computer device provided in an optional embodiment of the present invention, such as... Figure 15As shown, the computer device includes one or more processors 10, memory 20, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components communicate with each other via different buses and can be mounted on a common motherboard or otherwise installed as needed. The processors can process instructions executed within the computer device, including instructions stored in or on memory to display graphical information of a GUI on external input / output devices (such as display devices coupled to the interfaces). In some alternative implementations, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple computer devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). Figure 15 Take a processor 10 as an example.

[0168] Processor 10 may be a central processing unit, a network processor, or a combination thereof. Processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The programmable logic device may be a complex programmable logic device (CAMP), a field-programmable gate array (FPGA), a general-purpose array logic (GDA), or any combination thereof.

[0169] The memory 20 stores instructions executable by at least one processor 10 to cause at least one processor 10 to perform the method shown in the above embodiments.

[0170] The memory 20 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function. The data storage area may store data created based on the use of the computer device. Furthermore, the memory 20 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, the memory 20 may optionally include memory remotely located relative to the processor 10, which can be connected to the computer device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0171] Memory 20 may include volatile memory, such as random access memory. Memory may also include non-volatile memory, such as flash memory, hard disk, or solid-state drive. Memory 20 may also include combinations of the above types of memory.

[0172] The computer device also includes a communication interface 30 for communicating with other devices or communication networks.

[0173] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code, which, when accessed and executed by the computer, processor, or hardware, implements the methods shown in the above embodiments.

[0174] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A seismic facies-driven method of predicting source rock, characterized in that, The method includes: Multiple seismic attribute data volumes are generated using pre-stack seismic data from the target exploration area; The multiple seismic attribute data volumes are input into a trained seismic facies classification model to classify source rocks, resulting in a source rock facies classification data volume output by the seismic facies classification model; wherein, the seismic facies classification model is obtained by machine learning based on multiple sample seismic attribute data volumes and a pre-trained classification model; The source rock facies classification data volume and the well constraint model data volume are normalized respectively to obtain the normalized source rock facies classification data volume and the normalized well constraint model data volume. The normalized source rock facies classification data volume and the normalized well constraint model data volume are weighted and fused to generate the first fused data volume to be verified. If the first fused data volume matches the paleogeographic feature data volume, then the first fused data volume is determined to have passed the verification, and the corresponding source rock sensitive factor data volume is predicted using the first fused data volume. The step of weightedly fusing the normalized source rock facies classification data volume and the normalized well constraint model data volume to generate a first fused data volume to be verified includes: Set a first weighting coefficient and a second weighting coefficient corresponding to the normalized source rock facies classification data body and the normalized well constraint model data body, respectively; The first fused data volume is generated by weighting and summing the normalized source rock facies classification data volume and the normalized well constraint model data volume using the first weighting coefficient and the second weighting coefficient. The method further includes, after generating the first fused data volume: If the first fused data volume does not match the paleogeographic feature data volume, then the first fused data volume is determined to have failed verification. The first weighting coefficient and / or the second weighting coefficient are adjusted to obtain the adjusted weighting coefficient. The adjusted weighting coefficient is then used to perform a weighted summation on the normalized source rock facies classification data body and the normalized well constraint model data body to generate a second fusion data body to be verified. If the second fused data body does not match the paleogeographic feature data body, it is determined that the second fused data body has failed the verification. The adjusted weighting coefficients are then adjusted until the newly generated fused data body matches the paleogeographic feature data body, and the newly generated fused data body is determined to have passed the verification. The newly generated fused data volume is used to predict the corresponding source rock sensitive factor data volume.

2. The method according to claim 1, characterized in that, The target data volume includes multiple three-dimensional attribute data, and each three-dimensional attribute data includes three-dimensional spatial coordinate data and corresponding attribute values; The target data volume is normalized to obtain a normalized target data volume, including: The attribute values ​​in all three-dimensional attribute data of the target data volume are normalized to obtain a normalized target data volume. The normalized target data volume includes multiple normalized three-dimensional attribute data, and each normalized three-dimensional attribute data includes the three-dimensional spatial coordinate data and the normalized attribute value. The target data volume is either the hydrocarbon source rock facies classification data volume or the well constraint model data volume.

3. The method according to claim 1, characterized in that, The step of predicting the corresponding source rock sensitive factor data volume using the first fused data volume includes: Within the framework of Bayesian theory, inversion is performed based on the first fused data volume to generate the corresponding angular elastic impedance data volume; The source rock sensitivity factor data volume is generated by using the angular elastic impedance data volume for calculation.

4. The method according to claim 1, characterized in that, After generating the source rock sensitivity factor data volume, the method further includes: Obtain the wellbore curve corresponding to the source rock sensitive factor data volume; Calculate the correlation value based on the wellbore curve and the actual well logging interpretation results; If the correlation value is greater than the preset correlation threshold, then the source rock sensitive factor data volume is determined to be a highly reliable data volume.

5. A seismically driven hydrocarbon source rock prediction device, characterized in that, The device includes: The first generation unit is used to generate multiple corresponding seismic attribute data volumes using pre-stack seismic data of the target exploration area; The input unit is used to input the multiple seismic attribute data volumes into a trained seismic facies classification model for source rock classification, and obtain the source rock facies classification data volume output by the seismic facies classification model; wherein, the seismic facies classification model is obtained by machine learning based on multiple sample seismic attribute data volumes and a pre-trained classification model; The normalization unit is used to normalize the source rock facies classification data body and the well constraint model data body respectively, so as to obtain the normalized source rock facies classification data body and the normalized well constraint model data body. The weighted fusion unit is used to perform weighted fusion on the normalized source rock facies classification data body and the normalized well constraint model data body to generate a first fused data body to be verified. The first determining unit is configured to determine that the first fused data volume passes verification if the first fused data volume matches the paleogeographic feature data volume. The prediction unit is used to predict the corresponding source rock sensitive factor data volume using the first fused data volume; The weighted fusion unit is further configured to set a first weighting coefficient and a second weighting coefficient corresponding to the normalized source rock data volume and the normalized well constraint model data volume, respectively. The weighted fusion unit is further configured to use the first weighting coefficient and the second weighting coefficient to perform a weighted summation on the normalized source rock data volume and the normalized well constraint model data volume to generate the first fused data volume; The device further includes an adjustment unit: The first determining unit is further configured to determine that the first fused data body has failed verification if the first fused data body does not match the paleogeographic feature data body; The adjustment unit is used to adjust the first weighting coefficient and / or the second weighting coefficient to obtain the adjusted weighting coefficient; The weighted fusion unit is further configured to use the adjusted weighting coefficients to perform a weighted summation on the normalized source rock data volume and the normalized well constraint model data volume to generate a second fused data volume to be verified. The first determining unit is further configured to determine that the second fused data body has failed verification if the second fused data body does not match the paleogeographic feature data body; The adjustment unit is also used to adjust the adjusted weighting coefficients until the newly generated fused data volume matches the paleogeographic feature data volume. The first determining unit is further configured to determine that the newly generated fused data volume has passed verification; The prediction unit is also used to predict the corresponding source rock sensitive factor data volume using the newly generated fused data volume.

6. A computer device, characterized in that, include: A memory and a processor are interconnected, the memory storing computer instructions, and the processor executing the computer instructions to perform the seismically driven source rock prediction method according to any one of claims 1 to 4.

Images