Well-seismic model constraint-based sedimentary unit horizon interpretation method and device

Abstract

The present application relates to the technical field of seismic geologic horizon fine interpretation, and particularly relates to a sedimentary unit horizon interpretation method and device based on well-seismic model constraint. The method comprises: according to seismic data, fault interpretation is carried out based on structure mode and forward simulation; well-seismic time-depth relationship is established according to synthetic seismic record, and oil layer group level seismic standard layer is determined; initial stratum framework model is established according to fault interpretation result and seismic standard layer; three-dimensional space-variable velocity field is established according to framework model and time-depth relationship; based on framework model, each sedimentary unit geological layer at well point, seismic data and three-dimensional velocity field, convolution neural network algorithm is used to extract each sedimentary unit horizon, and horizon interpretation result is obtained. The present application can obtain sedimentary unit horizon interpretation result which is consistent with geological layer at well point and follows seismic waveform characteristics between wells, and adjacent sedimentary unit interpretation horizon, especially fault, has no channeling layer phenomenon, so that the interpretation horizon result is more fine and accurate.

Description

Technical Field

[0001] This invention relates to the field of fine interpretation of seismic geological strata, and in particular to a method and apparatus for interpreting sedimentary unit strata based on well-seismic model constraints. Background Technology

[0002] As oilfields enter the mid-to-late stages of development, conducting precise seismic reservoir prediction requires establishing accurate sedimentary unit-level stratigraphic frameworks. The accuracy of interpreting sedimentary units near faults, in particular, directly impacts the effectiveness of tapping into scattered remaining oil in efficient wells near faults. Typically, limited by the resolution and quality of seismic data, the scale for direct stratigraphic interpretation using seismic data is limited to oil-bearing formations or sandstone formations, corresponding to stratigraphic thicknesses ranging from hundreds to tens of meters. For the interpretation of sedimentary unit-level stratigraphic positions with a thickness of only a few meters or tens of meters, two methods are generally used: One method is to obtain the target sedimentary unit stratigraphic position by using a proportional linear interpolation method under the control of the seismic standard layer of the oil reservoir group. However, this method is suitable for the interpretation of sedimentary unit stratigraphic positions with relatively simple structures and stable stratigraphic thickness. It can maintain the structural trend of the strata between wells, but it cannot guarantee that the interpreted sedimentary unit stratigraphic position matches the geological strata. The other method is to assume that the depth ratio of the sedimentary unit is equal to the time domain ratio. Under the control of the seismic standard layer of the oil reservoir group, each sedimentary unit stratigraphic position in the time domain is divided according to the depth ratio, and the position of the sedimentary unit stratigraphic position is obtained by using a linear interpolation algorithm. This method can basically guarantee that the geological strata at the well point are consistent with the seismically interpreted stratigraphic position when there are no well-developed faults and the formation velocity plane does not change much. However, it cannot maintain the structural trend of the upper and lower stratigraphic positions, especially in fault-developed areas, where there is a phenomenon of "upper and lower stratigraphic position shift" between adjacent sedimentary unit stratigraphic positions. Overall, existing stratigraphic interpretation methods can interpret sedimentary units and have been applied in actual production. However, when the structure is relatively complex and the stratigraphic thickness varies greatly, the results of sedimentary unit-level stratigraphic interpretation still have many problems and need to be further improved. Summary of the Invention

[0003] This invention proposes a method and apparatus for interpreting sedimentary unit stratigraphic levels based on well-seismic model constraints. This addresses the shortcomings of existing methods for interpreting sedimentary unit stratigraphic levels at the meter or tens of meter level, which cannot guarantee that the interpreted sedimentary unit stratigraphic level matches the geological stratification, cannot maintain the tectonic trend of the upper and lower stratigraphic levels, and exhibit the phenomenon of "upper and lower stratigraphic level crossing" between adjacent sedimentary units.

[0004] According to one aspect of the present invention, a method for interpreting sedimentary unit stratigraphic positions based on a well-seismic model is provided, comprising: Obtain geological stratification data, seismic data, and synthetic seismic records for each sedimentary unit at the well points in the study area; Based on the earthquake data, fault interpretation is performed using tectonic models and forward modeling. Based on the synthesized seismic records, the well-seismic time-depth relationship was established, and the standard seismic layers of the oil reservoir group were determined; Based on the fault interpretation results and the earthquake standard layers, an initial stratigraphic framework model is established. Based on the initial stratigraphic framework model and the time-depth relationship, a three-dimensional spatially variable velocity field is established. Based on the initial stratigraphic framework model, geological stratification data of each sedimentary unit at the well point, seismic data, and the three-dimensional space-varying velocity field, the layer of each sedimentary unit in the target layer is extracted using a convolutional neural network algorithm to obtain the layer interpretation results of the sedimentary unit.

[0005] Preferably, before acquiring the seismic data, the seismic data is preprocessed, and the method includes: By using structural guided filtering and wavelet frequency division transform, the original seismic data of the target layer is preprocessed to obtain preprocessed seismic data.

[0006] Preferably, the method for preprocessing raw seismic data of the target layer using structurally guided filtering includes: Dip-guided filtering is performed on the raw seismic data using stratigraphic reflection dip angle information and anisotropic diffusion filtering technology.

[0007] Preferably, the method for preprocessing the original seismic data of the target layer using wavelet frequency division transform includes: Using the Morlet mother wavelet, spectral decomposition is performed within the bandwidth of the seismic data according to a predetermined step size to obtain a series of frequency-divided seismic data volumes.

[0008] Preferably, the method for fault interpretation of the seismic data based on tectonic models and forward modeling includes: Among the frequency-divided seismic data volumes obtained from seismic data, the frequency-divided data volume that best reflects the geological characteristics is selected, and different tectonic attribute volumes are extracted from it; Based on the existing fault knowledge in the study area, fault geological models with different fault displacements, intersections, and combinations were established in a hierarchical and classified manner. Combined with the dominant frequency of the selected frequency-division data volume, forward modeling was performed using Ricker wavelet to obtain a three-dimensional fault forward modeling data volume. The different structural attribute volumes and the three-dimensional fault forward modeling data volume are fused to obtain a structural attribute fused volume; Fault identification is performed based on the constructed attribute fusion body to obtain fault interpretation results.

[0009] Preferably, the method for performing fault identification based on the constructed attribute fusion body to obtain fault interpretation results includes: Different fault interpretation labels are defined on the constructed attribute fusion body, and they are input into the convolutional neural network model to train the model; The trained convolutional neural network model is used for tomography identification to obtain tomography interpretation results.

[0010] Preferably, the method for establishing the well-seismic-time-depth relationship based on the synthesized seismic record includes: An initial synthetic seismic record was created using acoustic waveforms and the Ricker wavelet; Extract seismic traces near the target layer, and use a sliding time window to perform correlation analysis on the initial synthetic seismic record and the seismic traces near the well. Determine the time window position with the largest correlation coefficient as the time-depth calibration result to obtain the initial time-depth relationship. Based on the initial time-depth relationship, the synthetic seismic records of all wells are locally stretched or compressed to ensure that the waveform of the synthetic seismic record has the maximum correlation with the well-side data, thus obtaining the final well-seismic time-depth relationship.

[0011] Preferably, the method for determining the oil reservoir group-level seismic standard layer based on the synthetic seismic record includes: Based on the geological stratification data and the characteristics of the synthetic seismic record, strong wave crests or troughs that are relatively continuous and stable in the lateral direction are preferred as the seismic standard layer. Using the fault interpretation results as spatial constraints, the earthquake standard layers are traced to complete the trace interpretation of the standard layers.

[0012] Preferably, the method for establishing an initial stratigraphic framework model based on the fault interpretation results and the seismic standard layers includes: Establish stratigraphic contact relationships for different seismic standard layers; Based on the fault interpretation results, seismic standard layers, and the stratigraphic contact relationships, an initial stratigraphic framework model is established.

[0013] Preferably, the method for establishing a three-dimensional space-varying velocity field based on the initial stratigraphic framework model and the time-depth relationship includes: A multi-well consistency analysis was performed on the time-depth relationship to remove data from wells with abnormal time-depth relationships in the initial formation framework model; Using the initial stratigraphic framework model as the initial velocity framework model, spatial velocity interpolation is performed using inverse distance weighting or kriging methods to establish a three-dimensional spatially variable velocity field.

[0014] Preferably, the three-dimensional spatial velocity field is subjected to profile and planar slice analysis to determine whether it meets the requirements. If not, the well data is analyzed to determine whether the velocity is abnormal due to encountering faults, failure to penetrate the target layer, inaccurate well trajectory of the deviated well, or inaccurate sonic curve. If so, the well data is removed and the three-dimensional spatial velocity field is re-established.

[0015] Preferably, the method for extracting the depositional unit layers of the target layer using a convolutional neural network algorithm includes: A convolutional neural network model was established, the model was trained, and the trained model was used to extract the deposition unit layers of the target layer. The method for training the model includes: The geological stratification data of each sedimentary unit at the well point are divided into a training set and a test set according to a predetermined ratio. The training set data, along with the corresponding initial stratigraphic framework model, seismic data, and the three-dimensional spatial velocity field, are input into the established convolutional neural network model. The geological stratification data of each sedimentary unit at the well point are used as label data to train the model, and the test set data is used for testing to obtain the trained convolutional neural network model.

[0016] Preferably, the method for quality control of the obtained sedimentary unit stratigraphic interpretation results includes: Determine whether the interpretation results of the sedimentary unit stratigraphy meet the condition that the time difference between the stratigraphy interpretation results and the seismic data is within a predetermined range. If not, the unmet condition is removed, and the compliant condition is added to the initial stratigraphic framework model. The model is then used to re-establish the three-dimensional spatially variable velocity field and to re-extract stratigraphy using a convolutional neural network algorithm to obtain the corresponding sedimentary unit stratigraphy interpretation results. This quality control process is repeated until the time difference between the interpretation results of the sedimentary unit stratigraphy interpretation results and the seismic data is within a predetermined range.

[0017] According to one aspect of the present invention, a well-seismic model-constrained sedimentary unit stratigraphic interpretation apparatus is provided, comprising: The acquisition unit is used to acquire geological stratification data, seismic data, and synthetic seismic records of each sedimentary unit at the well point in the study area. The fault interpretation unit is used to interpret faults based on the seismic data, tectonic models, and forward modeling. The time-depth relationship and standard layer determination unit is used to establish the well-seismic time-depth relationship and determine the oil reservoir group-level seismic standard layer based on the synthetic seismic record. The initial stratigraphic framework model building unit is used to build an initial stratigraphic framework model based on the fault interpretation results and the seismic standard layers. A three-dimensional velocity field establishment unit is used to establish a three-dimensional spatially variable velocity field based on the initial stratigraphic framework model and the time-depth relationship. The stratigraphic extraction unit is used to extract the stratigraphic layers of each sedimentary unit in the target layer based on the initial stratigraphic framework model, geological stratification data of each sedimentary unit at the well point, seismic data, and the three-dimensional space-varying velocity field, using a convolutional neural network algorithm, to obtain the stratigraphic interpretation results of the sedimentary units.

[0018] The present invention has at least the following beneficial effects: This invention proposes a method and apparatus for interpreting sedimentary unit stratigraphic levels based on a well-seismic model. By establishing an initial stratigraphic framework model and a three-dimensional spatially variable velocity field, and combining it with a convolutional neural network algorithm for stratigraphic extraction, the method obtains sedimentary unit stratigraphic interpretation results that match the geological stratification at the well point and follow the seismic waveform characteristics between wells. There is no stratigraphic crossing phenomenon, especially at faults, in the interpretation of stratigraphic levels between adjacent sedimentary units, making the interpretation results more refined and accurate. Attached Figure Description

[0019] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present invention and, together with the specification, serve to explain the technical solutions of the present invention.

[0020] Figure 1 A flowchart illustrating the method for interpreting sedimentary unit stratigraphic positions based on a well-seismic model according to an embodiment of the present invention is shown. Figure 2 The following are cross-sectional views comparing seismic data preprocessing before and after processing, according to an embodiment of the present invention: Figure A is a cross-sectional view before preprocessing, and Figure B is a cross-sectional view after preprocessing. Figure 3 This diagram illustrates the seismic standard layer and its interpretation cross-section according to an embodiment of the present invention. Figure 4 A cross-sectional view of the initial geological framework model according to an embodiment of the present invention is shown; Figure 5 A cross-sectional view showing the automatic interpretation results of sedimentary unit layers according to an embodiment of the present invention is shown. Detailed Implementation

[0021] Various exemplary embodiments, features, and aspects of the present invention will now be described in detail with reference to the accompanying drawings. The same reference numerals in the drawings denote elements that have the same or similar functions. Although various aspects of the embodiments are shown in the drawings, they are not necessarily drawn to scale unless specifically indicated otherwise.

[0022] The term “exemplary” as used herein means “serving as an example, embodiment, or illustration.” Any embodiment illustrated herein as “exemplary” is not necessarily to be construed as superior to or better than other embodiments.

[0023] In this document, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Furthermore, the term "at least one" in this document means any combination of at least two of any one or more elements. For example, including at least one of A, B, and C can mean including any one or more elements selected from the set consisting of A, B, and C.

[0024] Furthermore, to better illustrate the present invention, numerous specific details are set forth in the following detailed embodiments. Those skilled in the art will understand that the present invention can be practiced without certain specific details. In some instances, methods, means, elements, and circuits well known to those skilled in the art have not been described in detail in order to highlight the spirit of the invention.

[0025] Figure 1 A flowchart illustrating the method for interpreting sedimentary unit stratigraphic positions based on a well-seismic model according to an embodiment of the present invention is shown. Figure 2 The following are cross-sectional views comparing seismic data preprocessing before and after processing, according to an embodiment of the present invention: Figure A is a cross-sectional view before preprocessing, and Figure B is a cross-sectional view after preprocessing. Figure 3 This diagram illustrates the seismic standard layer and its interpretation cross-section according to an embodiment of the present invention. Figure 4 A cross-sectional view of the initial geological framework model according to an embodiment of the present invention is shown; Figure 5 A cross-sectional view showing the results of automatic interpretation of sedimentary unit stratigraphy according to an embodiment of the present invention is shown. Figures 1-5 As shown, a method for interpreting sedimentary unit stratigraphic positions based on a well-seismic model constrained by seismic data includes: Step S01: acquiring geological stratification data, seismic data, and synthetic seismic records for each sedimentary unit at the well point in the study area; Step S02: interpreting faults based on the seismic data, using tectonic models and forward modeling; Step S03: establishing a well-seismic time-depth relationship based on the synthetic seismic records and determining the seismic standard layer at the oil layer group level; Step S04: establishing an initial stratigraphic framework model based on the fault interpretation results and the seismic standard layer; Step S05: establishing a three-dimensional space-varying velocity field based on the initial stratigraphic framework model and the time-depth relationship; Step S06: extracting the stratigraphic positions of each sedimentary unit in the target layer using a convolutional neural network algorithm based on the initial stratigraphic framework model, geological stratification data, seismic data, and the three-dimensional space-varying velocity field at the well point, to obtain the sedimentary unit stratigraphic interpretation results.

[0026] The method for interpreting sedimentary unit stratigraphic positions based on well-seismic model constraints provided in this invention specifically includes the following steps: Step S01: Obtain geological stratification data, seismic data, and synthetic seismic records for each sedimentary unit at the well point in the study area.

[0027] In this invention, before acquiring the seismic data, the seismic data is preprocessed. The method includes: using structural steering filtering and wavelet frequency division transform to preprocess the original seismic data of the target layer to obtain preprocessed seismic data.

[0028] In this embodiment of the invention, the raw seismic data undergoes denoising and enhancement preprocessing. Using tilt-guided filtering in structural steering filtering and wavelet frequency division transform, the target layer seismic data is preprocessed to improve the signal-to-noise ratio and lateral continuity of the phase axis.

[0029] In this invention, the method for preprocessing raw seismic data of the target layer using structurally guided filtering includes: performing dip-guided filtering on the raw seismic data using stratum reflection dip angle information and anisotropic diffusion filtering technology.

[0030] In this embodiment of the invention, GeoEast software is first used to perform dip-guided filtering on the original seismic data by combining stratigraphic reflection dip angle information with anisotropic diffusion filtering technology. This improves the signal-to-noise ratio of the seismic data of the target layer, further enhances the continuity of the seismic reflection phase axis, and highlights the waveform differences at fault boundaries.

[0031] In this invention, the method for preprocessing the original seismic data of the target layer using wavelet frequency division transform includes: using Morlet mother wavelet to perform spectral decomposition within the frequency bandwidth of the seismic data according to a predetermined step size to obtain a series of frequency-divided seismic data volumes.

[0032] In this embodiment of the invention, a continuous wavelet frequency division transform method is used. The dipping-filtered seismic data is taken as input, preferably using the Morlet mother wavelet. Within the bandwidth of the seismic data, spectral decomposition is performed at a step size of 5Hz to obtain a series of frequency-divided seismic data volumes such as 5Hz, 10Hz, 15Hz, etc. Figure 2 The images show a comparison of seismic data preprocessing before and after the preprocessing process. Image A is the preprocessing profile, and Image B is the postprocessing profile.

[0033] Step S02: Based on the earthquake data, fault interpretation is performed using tectonic models and forward modeling.

[0034] In this invention, the method for fault interpretation of seismic data based on tectonic models and forward modeling includes: selecting the frequency-divided seismic data volume that best reflects geological characteristics from the frequency-divided seismic data volume obtained from the seismic data, and extracting different tectonic attribute volumes from it; establishing fault geological models with different fault displacements, intersections, and combinations based on existing fault knowledge in the study area, and performing forward modeling using Ricker wavelets in conjunction with the dominant frequency of the selected frequency-divided data volume to obtain a three-dimensional fault forward modeling data volume; fusing the different tectonic attribute volumes and the three-dimensional fault forward modeling data volume to obtain a tectonic attribute fusion body; and performing fault identification based on the tectonic attribute fusion body to obtain fault interpretation results.

[0035] In this embodiment of the invention, preprocessed seismic data is used as input, and an AI model-driven fault attribute body is formed based on a tectonic model library and forward modeling. Different fault interpretation labels are defined to perform detailed interpretation of the target fault layer.

[0036] Based on the reservoir and structural development characteristics of the target layer, from all the frequency-division data volumes obtained in step S01, the frequency-division data volume that best reflects geological features such as faults and reservoirs on its profile and plane is selected, and it is used as input to extract various structural attribute volumes such as coherence volume and ant volume from the selected frequency-division data volume.

[0037] Based on existing fault knowledge in the study area, namely, the fault displacement and cross-cutting relationships obtained from known well drilling data, a series of fault geological models with different cross-cutting and combination relationships, such as fault displacement of 3m, 5m, 10m, and 20m, were established. Combined with the dominant frequency of the selected frequency-divided seismic data volume, forward modeling was performed using Ricker wavelet to obtain a three-dimensional fault forward modeling data volume.

[0038] By integrating various structural attribute volumes such as coherence volume and ant volume, as well as three-dimensional fault forward modeling data, a convolutional neural network algorithm (based on VVA) is used to fuse them to obtain a structural attribute volume fusion body that reflects fault characteristics.

[0039] In this invention, the method for performing tomography identification based on the constructed attribute fusion body to obtain tomography interpretation results includes: defining different tomography interpretation labels on the constructed attribute fusion body and inputting them into a convolutional neural network model to train the model; and using the trained convolutional neural network model to perform tomography identification to obtain tomography interpretation results.

[0040] In this embodiment of the invention, different fault interpretation labels are defined on the constructed attribute fusion body, including key and major faults. The constructed attribute fusion body with defined fault labels is input into a convolutional neural network model, and the model training parameters are reasonably set and optimized. The model is trained, and based on the trained model, fault interpretation is performed on other unlabeled parts, completing the learning and tracking of fault samples to obtain the fault interpretation results.

[0041] The method for quality control of the tomographic interpretation results output by the convolutional neural network model includes: if the accuracy and reasonableness of a certain part of the tomographic interpretation result do not meet the requirements, then locally encrypt the tomographic interpretation label at the unreasonable position, input the encrypted tomographic interpretation label data into the convolutional neural network model for retraining, and repeat this process until the output tomographic interpretation result meets the requirements and the quality control is passed.

[0042] Step S03: Based on the synthesized seismic record, establish the well-seismic time-depth relationship and determine the standard seismic layer of the oil reservoir group.

[0043] In this invention, the method for establishing well-seismic time-depth relationship based on the synthetic seismic record includes: generating an initial synthetic seismic record using acoustic curves and Ricker wavelets; extracting seismic traces near the target layer near the well, performing correlation analysis on the initial synthetic seismic record and the seismic traces near the well using a sliding time window, determining the time window position with the largest correlation coefficient as the time-depth calibration result, and obtaining the initial time-depth relationship; based on the initial time-depth relationship, performing local stretching or compression adjustments on the synthetic seismic records of all wells to ensure that the waveform of the synthetic seismic record has the largest correlation with the seismic traces near the well, and obtaining the final well-seismic time-depth relationship.

[0044] In this embodiment of the invention, a detailed well-seismic time-depth relationship is established. Using synthetic seismic records, the locations and reflection characteristics of each oil-bearing formation are precisely calibrated, thereby identifying at least two oil-bearing formation-level seismic standard layers. A fine-grid interpretation of these standard layers is then completed using an automatic tracking method.

[0045] Based on the convolution principle, the initial synthetic seismic record is obtained using the acoustic curve of the well and the Ricker wavelet; the dominant frequency of the Ricker wavelet is consistent with the selected frequency-division seismic data volume, the time sampling rate is 1ms, and the wavelength is 100ms.

[0046] The well-side seismic traces near the target layer in the seismic data are extracted. The Pearson correlation coefficient method is used to perform correlation analysis on the initial synthetic seismic record (700-1000ms) and the well-side seismic traces using a sliding time window. The position with the largest correlation coefficient is found as the result of time-depth calibration, and the initial time-depth relationship is obtained.

[0047] The initial time-depth relationship obtained above is imported into the software. The synthetic seismic records of all wells are checked, and the local synthetic seismic records of individual wells are fine-tuned by stretching or compressing to ensure that the waveform of the synthetic seismic record has the greatest correlation with the well sidetrack, thus obtaining the final time-depth relationship.

[0048] In this invention, the method for determining the seismic standard layer of oil layer group based on the synthetic seismic record includes: selecting, based on the geological stratification data and the characteristics of the synthetic seismic record, a strong wave peak or trough reflection that is relatively continuous and stable in the lateral direction as the seismic standard layer; and using the fault interpretation results as spatial constraints to track the seismic standard layer and complete the tracking interpretation of the standard layer.

[0049] In this embodiment of the invention, based on geological strata (geological stratification data) and the characteristics of synthetic seismic records, the seismic reflection characteristics of each oil layer group or sandstone group are determined, and strong wave peaks or valleys that are relatively continuous and stable in the lateral direction are selected as the seismic standard layer. Using the fault interpretation results from step S02 as spatial constraints, the seismic standard layers are labeled with layer grids from coarse (64CDP×64CDP) to fine (16CDP×16CDP) according to the quality of the seismic data. Then, using automatic tracking based on waveform or energy similarity or deep learning algorithms, a fine-grained 1CDP×1CDP standard layer interpretation is completed. The results are as follows: Figure 3 As shown.

[0050] Step S04: Based on the fault interpretation results and the earthquake standard layers, establish an initial stratigraphic framework model.

[0051] In this invention, the method for establishing an initial stratigraphic framework model based on the fault interpretation results and the seismic standard layers includes: setting stratigraphic contact relationships for different seismic standard layers; and establishing an initial stratigraphic framework model based on the fault interpretation results, the seismic standard layers, and the stratigraphic contact relationships.

[0052] In this embodiment of the invention, an initial stratigraphic framework model is established. Using the fault interpretation results from steps S02 and S03 and the seismic standard layers as input, stratigraphic contact relationships such as stratigraphic integration, parallel to the top, and parallel to the bottom between different standard layers are given. Using a framework model establishment method in software (such as GeoEast), an initial stratigraphic framework model is established, and the result is as follows: Figure 4 As shown.

[0053] Step S05: Based on the initial stratigraphic framework model and the time-depth relationship, establish a three-dimensional spatially variable velocity field.

[0054] In this invention, the method for establishing a three-dimensional spatially varying velocity field based on the initial formation framework model and the time-depth relationship includes: performing multi-well consistency analysis on the time-depth relationship to remove data from wells with abnormal time-depth relationships in the initial formation framework model; using the initial formation framework model as an initial velocity framework model, and performing spatial velocity interpolation using inverse distance weighting or kriging methods to establish a three-dimensional spatially varying velocity field.

[0055] In this embodiment of the invention, a three-dimensional space-varying velocity field is established based on an initial stratigraphic framework model. Using the initial stratigraphic framework model obtained in step 4 and the time-depth relationship from step 3 as inputs, a refined three-dimensional space-varying velocity field is established using methods such as inverse distance weighting.

[0056] The specific approach is as follows: First, perform a multi-well consistency analysis on the time-depth relationship obtained in step S03 to remove well data with abnormal time-depth relationships. Second, use the initial formation framework model obtained in step S04 as the initial velocity framework model, and use inverse distance weighting or kriging methods to perform spatial velocity interpolation to establish a three-dimensional spatially variable velocity field.

[0057] In this invention, the three-dimensional spatial velocity field is subjected to profile and planar slice analysis to determine whether it meets the requirements. If not, the well data is analyzed to determine whether the velocity is abnormal due to encountering faults, not penetrating the target layer, inaccurate well trajectory of the deviated well, or inaccurate sonic curve. If so, the well data is discarded and the three-dimensional spatial velocity field is re-established.

[0058] In this embodiment of the invention, the established three-dimensional spatially variable velocity field is subjected to cross-sectional and planar slice analysis. If there are no abnormal phenomena such as bullseye on the cross-section or plane, the requirements are met. If the requirements are not met, abnormal data is analyzed and eliminated, and the three-dimensional velocity field is re-established.

[0059] Step S06: Based on the initial stratigraphic framework model, geological stratification data of each sedimentary unit at the well point, seismic data, and the three-dimensional spatial velocity field, the layer of each sedimentary unit in the target layer is extracted using a convolutional neural network algorithm to obtain the layer interpretation results of the sedimentary unit.

[0060] In this invention, the method for extracting the layers of each sedimentary unit in the target layer using a convolutional neural network algorithm includes: establishing a convolutional neural network model, training the model, and using the trained model to extract the layers of each sedimentary unit in the target layer; wherein, the method for training the model includes: dividing the geological stratification data of each sedimentary unit at the well point into a training set and a test set according to a predetermined ratio; inputting the training set data, its corresponding initial stratigraphic framework model, seismic data, and the three-dimensional spatial velocity field into the established convolutional neural network model; using the geological stratification data of each sedimentary unit at the well point as label data to train the model; and testing the model using the test set data to obtain the trained convolutional neural network model.

[0061] In this embodiment of the invention, sedimentary unit strata are automatically extracted. Combining the initial stratigraphic framework model from step S04, the geological stratification data of each sedimentary unit at the well point, and seismic waveform characteristics, along with various feature information from the three-dimensional space-varying velocity field obtained in step S05, the Unet convolutional neural network algorithm is used to extract the strata of each sedimentary unit in the target layer.

[0062] The specific process is as follows: An initial stratigraphic framework model, geological stratification data of each sedimentary unit at the well point, seismic waveform characteristics, and a three-dimensional spatially varying velocity field together constitute a sample set. This sample set is divided into 80% training data and 20% validation data. The training data is input into the Unet convolutional neural network model, using the geological stratification data of each product unit at the well point as label data to train the model. The remaining validation data is used to validate the model.

[0063] Using the trained model, the stratigraphic units of each sedimentary unit in the target layer between wells in the entire study area were extracted, and the stratigraphic extraction results, i.e., the stratigraphic interpretation results of the sedimentary units, were obtained.

[0064] In this invention, the method for quality control of the obtained sedimentary unit stratigraphic interpretation results includes: determining whether the obtained sedimentary unit stratigraphic interpretation results satisfy the condition that the stratigraphic interpretation time difference with the seismic data is within a predetermined range; if not, the unsatisfactory part is removed, the satisfactory part is added to the initial stratigraphic framework model, and the three-dimensional spatially variable velocity field is re-established using it, and the stratigraphic extraction is re-performed using a convolutional neural network algorithm to obtain the corresponding sedimentary unit stratigraphic interpretation results; the quality control process is repeated until the stratigraphic interpretation time difference between the sedimentary unit stratigraphic interpretation results and the seismic data is within a predetermined range.

[0065] In this embodiment of the invention, the interpretation effect of sedimentary unit stratigraphy is evaluated. If the blind well geological stratification obtained in step S06, i.e., the time difference between the interpretation result of the sedimentary unit geological stratigraphy and the stratigraphy interpretation of the seismic data, is within a predetermined range, such as ±2ms, it meets the requirements. Otherwise, the process returns to step S04. Based on the previously established framework model, the interpretation results of the main sedimentary unit stratigraphy of the part that meets the quality control requirements obtained in step S06 are added and used as the control to update the initial stratigraphic framework model. Then, steps S05 and S06 are executed again to perform quality control on the results obtained in S06. If the requirements are still not met, the process of steps S04 to S06 is repeated. The initial stratigraphic framework model is iteratively updated using the data of the part that meets the quality control requirements obtained in S06 until the sedimentary unit stratigraphy interpretation results that meet the requirements are obtained. The results are as follows. Figure 5 As shown.

[0066] It is understood that the various method embodiments mentioned above in this invention can be combined with each other to form combined embodiments without violating the principle and logic. Due to space limitations, this invention will not elaborate further.

[0067] The execution entity of the well-seismic model-constrained sedimentary unit stratigraphic interpretation method can be a well-seismic model-constrained sedimentary unit stratigraphic interpretation device. For example, the well-seismic model-constrained sedimentary unit stratigraphic interpretation method can be executed by a terminal device, a server, or other processing devices. The terminal device can be user equipment (UE), a mobile device, a user terminal, a computing device, etc. In some possible implementations, this well-seismic model-constrained sedimentary unit stratigraphic interpretation method can be implemented by a processor calling computer-readable instructions stored in memory.

[0068] Those skilled in the art will understand that, in the above-described method of the specific implementation, the order in which each step is written does not imply a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of each step should be determined by its function and possible internal logic.

[0069] In this invention, a well-seismic model-constrained sedimentary unit stratigraphic interpretation device includes: an acquisition unit for acquiring geological stratification data, seismic data, and synthetic seismic records of each sedimentary unit at a well point in the study area; a fault interpretation unit for performing fault interpretation based on the seismic data, tectonic models, and forward modeling; a time-depth relationship and standard layer determination unit for establishing a well-seismic time-depth relationship and determining the oil-bearing group-level seismic standard layer based on the synthetic seismic records; an initial stratigraphic framework model establishment unit for establishing an initial stratigraphic framework model based on the fault interpretation results and the seismic standard layers; a three-dimensional velocity field establishment unit for establishing a three-dimensional space-varying velocity field based on the initial stratigraphic framework model and the time-depth relationship; and a stratigraphic extraction unit for extracting the stratigraphic layers of each sedimentary unit in the target layer using a convolutional neural network algorithm based on the initial stratigraphic framework model, geological stratification data of each sedimentary unit at the well point, seismic data, and the three-dimensional space-varying velocity field, to obtain the sedimentary unit stratigraphic interpretation results.

[0070] In some embodiments, the functions or modules and units included in the apparatus provided by the present invention can be used to execute the methods described in the above method embodiments. The specific implementation can be referred to the description of the above method embodiments, and for the sake of brevity, it will not be repeated here.

[0071] The Fuyu oil reservoir is a crucial target for increasing reserves and production, requiring improvements in reservoir understanding and description accuracy to meet development needs. However, factors such as fault development, sparse well network, and seismic data quality hinder the accuracy of inter-well stratigraphic interpretation, limiting the understanding of structural features and inter-well sand body connectivity. Therefore, a combination of well and seismic analysis is needed to further improve the accuracy of sedimentary unit stratigraphic interpretation, thereby guiding precise prediction of narrow and thin reservoirs and meeting production requirements.

[0072] One traditional method for interpreting sedimentary unit stratigraphic horizons involves identifying two seismic standard layers near the target layer based on fine well-seismic calibration results. Using these standard layers as the top and bottom, commercial software (such as SMI and VVA) is used to obtain multiple proportionally scaled slices, with the slice closest to the target sedimentary unit at the well point being selected as the target unit. Another method assumes that the depth proportion of sedimentary units is equal to their time proportion. Under the control of the top and bottom seismic standard layers, the proportion of each sedimentary unit's formation thickness at each well point to the total thickness of the formation between the two standard layers is calculated. This proportion is then interpolated using a grid to obtain a planar distribution map of the target unit's time grid proportion. This map is then added to the time grid of the seismic standard layer at the top of the target layer to obtain the interpretation result of the target sedimentary unit's horizon. However, when interpreting sedimentary unit horizons using traditional methods, especially when the inter-well structure is complex and the formation thickness varies significantly, the interpreted horizons of adjacent sedimentary units exhibit a "layer shifting" phenomenon, severely impacting the accuracy of fine reservoir prediction.

[0073] This invention first utilizes guided filtering and wavelet frequency division transform methods to preprocess the seismic data of the target layer, improving the signal-to-noise ratio, resolution, and lateral continuity of the phase axis of the seismic data. Figure 2 As shown. Based on this, an AI model-driven intelligent and refined fault interpretation is implemented; secondly, synthetic seismic records are used to establish a refined time-depth relationship, clarify the location and reflection characteristics of each oil layer group, and then determine at least two oil layer group-level seismic standard layers. The 1CDP×1CDP refined grid interpretation of this standard layer is completed using an automatic tracking method, as shown. Figure 3 As shown; then, using the interpretation results of faults and seismic standard layers as constraints, an initial stratigraphic framework model is established, as follows. Figure 4 As shown, and a three-dimensional space-varying velocity field; finally, integrating the initial stratigraphic framework model, geological stratification and seismic waveform characteristics of each sedimentary unit at the well point, and various feature information of the three-dimensional velocity field, algorithms such as convolutional neural networks are used to automatically extract and evaluate the quality of each sedimentary unit layer in the target layer, outputting the target sedimentary unit layer that meets the quality requirements, such as... Figure 5 As shown.

[0074] Compared with conventional interpretation methods, the sedimentary unit stratigraphic interpretation obtained by applying this invention matches the geological strata at the well point, follows the seismic waveform characteristics between wells, and shows no stratigraphic crossing between adjacent sedimentary units. In particular, the interpreted stratigraphics obtained at faults are more refined and accurate, laying a scientific geological foundation for the precise prediction of reservoirs at the sedimentary unit level.

[0075] The various embodiments of the present invention have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or technical improvements to the embodiments in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A method for interpreting sedimentary unit stratigraphic positions based on a well-seismic model, characterized in that, include: Obtain geological stratification data, seismic data, and synthetic seismic records for each sedimentary unit at the well points in the study area; Based on the earthquake data, fault interpretation is performed using tectonic models and forward modeling. Based on the synthesized seismic records, the well-seismic time-depth relationship was established, and the standard seismic layers of the oil reservoir group were determined; Based on the fault interpretation results and the earthquake standard layers, an initial stratigraphic framework model is established. Based on the initial stratigraphic framework model and the time-depth relationship, a three-dimensional spatially variable velocity field is established. Based on the initial stratigraphic framework model, geological stratification data of each sedimentary unit at the well point, seismic data, and the three-dimensional space-varying velocity field, the layer of each sedimentary unit in the target layer is extracted using a convolutional neural network algorithm to obtain the layer interpretation results of the sedimentary unit.

2. The method for interpreting sedimentary unit stratigraphic positions based on a well-seismic model according to claim 1, characterized in that, Before acquiring the seismic data, the seismic data is preprocessed, and the method includes: By using structural guided filtering and wavelet frequency division transform, the original seismic data of the target layer is preprocessed to obtain preprocessed seismic data.

3. The method for interpreting sedimentary unit stratigraphic positions based on a well-seismic model according to claim 2, characterized in that, The method for preprocessing raw seismic data of the target layer using structurally guided filtering includes: Dip-guided filtering is performed on the raw seismic data using stratigraphic reflection dip angle information and anisotropic diffusion filtering technology.

4. The method for interpreting sedimentary unit stratigraphic positions based on a well-seismic model according to claim 2, characterized in that, The method for preprocessing raw seismic data of the target layer using wavelet frequency division transform includes: Using the Morlet mother wavelet, spectral decomposition is performed within the bandwidth of the seismic data according to a predetermined step size to obtain a series of frequency-divided seismic data volumes.

5. The method for interpreting sedimentary unit stratigraphic positions based on a well-seismic model according to claim 4, characterized in that, The method for fault interpretation of the seismic data based on tectonic models and forward modeling includes: Among the frequency-divided seismic data volumes obtained from seismic data, the frequency-divided data volume that best reflects geological characteristics is selected, and different tectonic attribute volumes are extracted from it; Based on the existing fault knowledge in the study area, fault geological models with different fault displacements, intersections, and combinations were established in a hierarchical and classified manner. Combined with the dominant frequency of the selected frequency-division data volume, forward modeling was performed using Ricker wavelet to obtain a three-dimensional fault forward modeling data volume. The different structural attribute volumes and the three-dimensional fault forward modeling data volume are fused to obtain a structural attribute fused volume; Fault identification is performed based on the constructed attribute fusion body to obtain fault interpretation results.

6. The method for interpreting sedimentary unit stratigraphic positions based on a well-seismic model according to claim 5, characterized in that, The method for performing fault identification based on the constructed attribute fusion body to obtain fault interpretation results includes: Different fault interpretation labels are defined on the constructed attribute fusion body, and they are input into the convolutional neural network model to train the model; The trained convolutional neural network model is used for tomography identification to obtain tomography interpretation results.

7. The method for interpreting sedimentary unit stratigraphic positions based on a well-seismic model according to claim 1, characterized in that, The method for establishing well-seismic time-depth relationships based on the synthetic seismic record includes: An initial synthetic seismic record was created using acoustic waveforms and the Ricker wavelet; Extract seismic traces near the target layer, and use a sliding time window to perform correlation analysis on the initial synthetic seismic record and the seismic traces near the well. Determine the time window position with the largest correlation coefficient as the time-depth calibration result to obtain the initial time-depth relationship. Based on the initial time-depth relationship, the synthetic seismic records of all wells are locally stretched or compressed to ensure that the waveform of the synthetic seismic record has the maximum correlation with the well-side data, thus obtaining the final well-seismic time-depth relationship.

8. The method for interpreting sedimentary unit stratigraphic positions based on a well-seismic model according to claim 1, characterized in that, The method for determining the oil reservoir group-level seismic standard layer based on the synthetic seismic record includes: Based on the geological stratification data and the characteristics of the synthetic seismic record, strong wave crests or troughs that are relatively continuous and stable in the lateral direction are preferred as the seismic standard layer. Using the fault interpretation results as spatial constraints, the earthquake standard layers are traced to complete the trace interpretation of the standard layers.

9. The method for interpreting sedimentary unit stratigraphic positions based on a well-seismic model according to claim 1, characterized in that, The method for establishing an initial stratigraphic framework model based on the fault interpretation results and the seismic standard layers includes: Establish stratigraphic contact relationships for different seismic standard layers; Based on the fault interpretation results, seismic standard layers, and the stratigraphic contact relationships, an initial stratigraphic framework model is established.

10. The method for interpreting sedimentary unit stratigraphic positions based on a well-seismic model according to claim 1, characterized in that, The method for establishing a three-dimensional space-varying velocity field based on the initial stratigraphic framework model and the time-depth relationship includes: A multi-well consistency analysis was performed on the time-depth relationship to remove data from wells with abnormal time-depth relationships in the initial formation framework model; Using the initial stratigraphic framework model as the initial velocity framework model, spatial velocity interpolation is performed using inverse distance weighting or kriging methods to establish a three-dimensional spatially variable velocity field.

11. The method for interpreting sedimentary unit stratigraphic levels based on a well-seismic model according to claim 10, characterized in that: The three-dimensional spatial velocity field is subjected to profile and planar slice analysis to determine whether it meets the requirements. If not, the well data is analyzed to determine whether the velocity anomaly is caused by encountering faults, not penetrating the target layer, inaccurate well trajectory of the deviated well, or inaccurate sonic curve. If so, the well data is removed and the three-dimensional spatial velocity field is re-established.

12. The method for interpreting sedimentary unit stratigraphic positions based on a well-seismic model according to claim 1, characterized in that, The method for extracting the layers of each depositional unit in the target layer using a convolutional neural network algorithm includes: A convolutional neural network model was established, the model was trained, and the trained model was used to extract the deposition unit layers of the target layer. The method for training the model includes: The geological stratification data of each sedimentary unit at the well point are divided into a training set and a test set according to a predetermined ratio. The training set data, along with the corresponding initial stratigraphic framework model, seismic data, and the three-dimensional spatial velocity field, are input into the established convolutional neural network model. The geological stratification data of each sedimentary unit at the well point are used as label data to train the model, and the test set data is used for testing to obtain the trained convolutional neural network model.

13. The method for interpreting sedimentary unit stratigraphic positions based on well-seismic model constraints according to any one of claims 1-12, characterized in that, The method for quality control of the obtained sedimentary unit stratigraphic interpretation results includes: Determine whether the interpretation results of the sedimentary unit stratigraphy meet the condition that the time difference between the stratigraphy interpretation results and the seismic data is within a predetermined range. If not, the unmet condition is removed, and the compliant condition is added to the initial stratigraphic framework model. The model is then used to re-establish the three-dimensional spatially variable velocity field and to re-extract stratigraphy using a convolutional neural network algorithm to obtain the corresponding sedimentary unit stratigraphy interpretation results. This quality control process is repeated until the time difference between the interpretation results of the sedimentary unit stratigraphy interpretation results and the seismic data is within a predetermined range.

14. A device for interpreting sedimentary unit stratigraphic positions based on a well-seismic model, characterized in that, include: The acquisition unit is used to acquire geological stratification data, seismic data, and synthetic seismic records of each sedimentary unit at the well point in the study area. The fault interpretation unit is used to interpret faults based on the seismic data, tectonic models, and forward modeling. The time-depth relationship and standard layer determination unit is used to establish the well-seismic time-depth relationship and determine the oil reservoir group-level seismic standard layer based on the synthetic seismic record. The initial stratigraphic framework model building unit is used to build an initial stratigraphic framework model based on the fault interpretation results and the seismic standard layers. A three-dimensional velocity field establishment unit is used to establish a three-dimensional spatially variable velocity field based on the initial stratigraphic framework model and the time-depth relationship. The stratigraphic extraction unit is used to extract the stratigraphic layers of each sedimentary unit in the target layer based on the initial stratigraphic framework model, geological stratification data of each sedimentary unit at the well point, seismic data, and the three-dimensional space-varying velocity field, using a convolutional neural network algorithm, to obtain the stratigraphic interpretation results of the sedimentary units.

Images