Method for constructing prediction model of oil and gas reserves and method for predicting oil and gas reserves

Abstract

The embodiment of the present application relates to the technical field of oil and gas exploration and development, and discloses a construction method of an oil and gas reserve prediction model and an oil and gas reserve prediction method, comprising: obtaining drilling post-stack data bodies, drilling oil and gas reserves and drilling trajectories of a plurality of reference drillings; performing attribute extraction on the drilling post-stack data bodies to obtain reference measured values of a plurality of preset kinds of attributes; determining a reference attribute plane based on the reference measured values of the attributes of the preset kinds, projecting the drilling trajectories on the reference attribute plane to obtain drilling trajectory projections; sampling along the drilling trajectory projections to obtain reference sampling values of the attributes of the preset kinds; and training a preset machine model by using the reference sampling values of the attributes of the preset kinds and the drilling oil and gas reserves to obtain the oil and gas reserve prediction model.

Description

Technical Field

[0001] This invention relates to the field of oil and gas exploration and development technology, and in particular to a method for constructing an oil and gas reserve prediction model and a method for predicting oil and gas reserves. Background Technology

[0002] Unconventional oil and gas EUR (estimated ultimate recovery) forecasting is a crucial step in oil and gas companies' decision-making process, as they determine extraction locations based on EUR forecasts for different regions.

[0003] Currently, there are two methods for predicting unconventional oil and gas recoverable reserves (EUR). One method indirectly assesses EUR based on the distribution of unconventional oil and gas sweet spots, using these sweet spots as points with higher EUR values. The other method directly assesses EUR using well logging data. However, a complex relationship remains between the distribution of unconventional oil and gas sweet spots and their recoverable reserves, making it difficult to quantify the connection. Therefore, predicting the distribution of unconventional oil and gas sweet spots using seismic data is not equivalent to predicting EUR using seismic data. Assessing and predicting EUR using well logging data primarily employs static and dynamic methods, requiring a large amount of drilling data and thus cannot be extended to areas with only a few wells. Summary of the Invention

[0004] The purpose of this invention is to provide at least one method for constructing a prediction model for oil and gas reserves and a method for predicting oil and gas reserves, which can at least solve the problem of predicting oil and gas reserves in areas with only a few wells.

[0005] To address the aforementioned technical problems, at least one embodiment of this application provides a method for constructing a prediction model for oil and gas reserves, comprising:

[0006] Obtain the back-stack data volume, drilling oil and gas reserves, and drilling trajectory of several reference wells;

[0007] Attribute extraction is performed on the drilling stack data volume to obtain reference measured values ​​of several preset types of attributes;

[0008] A reference attribute plane is determined based on the reference measured values ​​of the attributes of each of the preset types, and each of the drilling trajectories is projected onto the reference attribute plane to obtain the drilling trajectory projection.

[0009] Sampling is performed along the projection of the drilling trajectory to obtain reference sample values ​​of several preset types of attributes;

[0010] The preset machine model is trained using reference sample values ​​of the preset type of attributes and the drilling oil and gas reserves to obtain a prediction model for oil and gas reserves.

[0011] At least one embodiment of this application also provides a method for predicting oil and gas reserves, including:

[0012] Obtain the target seismic data volume of the target well, and the oil and gas reserve prediction model constructed by the oil and gas reserve prediction model construction method described in any of the above embodiments;

[0013] The target seismic data volume is subjected to attribute extraction to obtain the target measured values ​​of several preset types of attributes;

[0014] Based on the target measured values ​​of each of the preset types of attributes, a target attribute plane is determined, and the trajectory of the target well is projected onto the target attribute plane to obtain the target trajectory projection.

[0015] Sampling is performed along the target trajectory projection to obtain target sample values ​​of several preset types of attributes;

[0016] Based on the oil and gas reserve prediction model, the oil and gas reserve prediction value of the target well is obtained by calculating the target sampled values ​​of several preset types of attributes.

[0017] At least one embodiment of this application also provides an apparatus for constructing a prediction model for oil and gas reserves, comprising:

[0018] The drilling data acquisition module is used to acquire the drilling back-stack data volume, drilling oil and gas reserves, and drilling trajectory of several reference wells;

[0019] The post-stack attribute extraction module is used to extract attributes from the post-stack drilling data volume to obtain reference measured values ​​of several preset types of attributes.

[0020] The trajectory projection plane module is used to determine a reference attribute plane based on the reference measured values ​​of the attributes of each preset type, and to project each drilling trajectory onto the reference attribute plane to obtain the drilling trajectory projection.

[0021] The projection attribute sampling module is used to sample along the drilling trajectory projection to obtain reference sample values ​​of several preset types of attributes;

[0022] The machine model training module is used to train a preset machine model using reference sample values ​​of the preset type of attributes and the drilling oil and gas reserves to obtain a prediction model for oil and gas reserves.

[0023] At least one embodiment of this application also provides an oil and gas reserves prediction device, comprising:

[0024] The target seismic data module is used to acquire the target seismic data volume of the target well, and the oil and gas reserve prediction model constructed by the oil and gas reserve prediction model construction method described in any of the above embodiments.

[0025] The target attribute extraction module is used to extract attributes from the target seismic data volume to obtain the target measured values ​​of several preset types of attributes.

[0026] The target trajectory projection module is used to determine the target attribute plane based on the measured values ​​of the target attributes of each preset type, and to project the trajectory of the target drilling onto the target attribute plane to obtain the target trajectory projection.

[0027] The target trajectory sampling module is used to sample along the projection of the target trajectory to obtain target sample values ​​of several preset types of attributes;

[0028] The target reserve prediction module is used to calculate the target well's oil and gas reserve prediction value based on the oil and gas reserve prediction model and the target sample values ​​of several preset types of attributes.

[0029] At least one embodiment of this application also provides an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the above-described method for constructing a prediction model of oil and gas reserves, or the at least one processor to perform the above-described method for predicting oil and gas reserves.

[0030] At least one embodiment of this application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-described method for constructing a prediction model for oil and gas reserves, or the computer program, when executed by a processor, implements the above-described method for predicting oil and gas reserves.

[0031] The method for constructing an oil and gas reserve prediction model provided in the embodiments of this application involves acquiring back-stack drilling data volumes, drilling oil and gas reserves, and drilling trajectories of several reference wells; extracting attributes from the back-stack drilling data volumes to obtain reference measured values ​​of several preset types of attributes; determining a reference attribute plane based on the reference measured values ​​of each preset type of attribute; projecting each drilling trajectories onto the reference attribute plane to obtain a drilling trajectory projection; sampling along the drilling trajectory projection to obtain reference sampled values ​​of several preset types of attributes; and training a preset machine model using the reference sampled values ​​of the preset types of attributes and the drilling oil and gas reserves to obtain an oil and gas reserve prediction model. When applying this oil and gas reserve prediction model, the numerical values ​​of the preset attribute types of the collected seismic data are input into the oil and gas reserve prediction model for prediction. This fully utilizes seismic data to predict oil and gas reserves, reducing reliance on well logging data. On the one hand, this avoids using the distribution of oil and gas sweet spots to assess the distribution of oil and gas reserves, and avoids the impact of the complex relationship between sweet spot distribution and oil and gas reserves on prediction accuracy. On the other hand, reducing reliance on well logging data allows this method to be extended to areas with only a small number of wells. Attached Figure Description

[0032] One or more embodiments are illustrated by way of example with reference to the accompanying drawings, and these illustrative descriptions do not constitute a limitation on the embodiments.

[0033] Figure 1 This is a flowchart of a method for constructing a prediction model for oil and gas reserves in one embodiment;

[0034] Figure 2 A flowchart illustrating the process for predicting unconventional recoverable oil and gas reserves in one embodiment;

[0035] Figure 3 Another flowchart of the process for predicting the planar distribution of recoverable reserves in one embodiment;

[0036] Figure 4 A flowchart of a machine learning-based seismic prediction method for unconventional recoverable oil and gas reserves in one embodiment;

[0037] Figure 5 This is a structural diagram of a planar distribution seismic prediction device for recoverable shale gas reservoirs in one embodiment.

[0038] Figure 6 This is a structural diagram of an electronic device in one embodiment;

[0039] Figure 7 A line graph showing the number of input variables versus the relative error of the model in one embodiment;

[0040] Figure 8This is a shale gas EUR planar prediction map in one embodiment;

[0041] Figure 9 This is a line graph showing the predicted and actual EUR values ​​for different wells in a shale gas field in one embodiment.

[0042] Figure 10 A bar chart showing the importance of input variable features in one embodiment;

[0043] Figure 11 Another planar projection of shale oil EUR in one embodiment;

[0044] Figure 12 This is a line graph showing the predicted and actual EUR values ​​for different wells in a basin shale oil well in one embodiment. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the various embodiments of this application will be described in detail below with reference to the accompanying drawings. However, those skilled in the art will understand that many technical details have been provided in the various embodiments of this application to help readers better understand this application. However, the technical solutions claimed in this application can be implemented even without these technical details and various changes and modifications based on the following embodiments. The division of the various embodiments below is for the convenience of description and should not constitute any limitation on the specific implementation of this application. The various embodiments can be combined with and referenced by each other without contradiction.

[0046] To address the technical problem of predicting oil and gas reserves in areas with only a limited number of wells, this invention proposes a method for constructing a prediction model. The implementation details of the prediction model construction method in this embodiment are described below. The following content is provided for ease of understanding and is not essential for implementing this solution.

[0047] Example 1:

[0048] The method for constructing the oil and gas reserve prediction model in this embodiment can be applied to electronic devices with communication, computing, and data storage capabilities. The specific process can be as follows: Figure 1 As shown, it includes:

[0049] Step 110: Obtain the drilling back-stack data volume, drilling oil and gas reserves, and drilling trajectory of several reference wells.

[0050] In this embodiment, the post-stack drilling data volume reflects the seismic properties of the area where the reference well is located, and the drilling trajectory reflects the extension path of the reference well in the area. By extracting gathers from the pre-stack seismic data volume at different angles and then overlaying the extracted gathers, the post-stack drilling data volume can be obtained. The well's oil and gas reserves are determined by the parameters required to predict the oil and gas reserves using the prediction model. For example, if the oil and gas reserves need to be predicted per unit horizontal segment length using the prediction model, then the specific index of the well's oil and gas reserves is the oil and gas reserves per unit horizontal segment length, i.e., the estimated ultimate recovery (EUR) per unit well.

[0051] Step 120: Extract attributes from the drilling stack data volume to obtain reference measured values ​​of several preset types of attributes.

[0052] In this embodiment, the post-drilling stack data volume records information about different seismic interpretation horizons. To improve the accuracy of model predictions, the seismic attributes of the seismic interpretation horizons in the post-drilling stack data volume that are highly correlated with hydrocarbons are selected as the target horizon. Attributes are extracted from the top, middle, and bottom of the target horizon using time windows. The average value of the same preset attribute type is calculated for the top, middle, and bottom horizons to obtain the reference measured value for that preset attribute type. The same processing is performed on each preset attribute type to obtain reference measured values ​​for several preset attribute types. The preset attribute types include at least one of root mean square amplitude, average amplitude, maximum peak amplitude, and average instantaneous frequency. The more preset attribute types there are, the better the accuracy of model predictions will be.

[0053] Step 130: Determine a reference attribute plane based on the reference measured values ​​of the attributes of each of the preset types, and project each of the drilling trajectories onto the reference attribute plane to obtain the drilling trajectory projection.

[0054] In this embodiment, within the area where the reference well is located, the measured reference values ​​for the same preset type of attribute differ at different locations. The well trajectory extends across different locations within the reference well area, and the measured reference values ​​for the same preset type of attribute differ at different locations along the well trajectory. To more accurately extract the seismic attribute information of the reference well, a reference attribute plane is first determined based on the measured reference values ​​for each preset type of attribute. This reference attribute plane reflects the attribute situation for each preset type at different locations within the reference well area. Each well trajectory is then projected onto the reference attribute plane to determine its distribution.

[0055] Step 140: Sample along the drilling trajectory projection to obtain reference sample values ​​of several preset types of attributes.

[0056] In this embodiment, the drilling trajectory projection is located on a reference attribute plane. The drilling trajectory projection consists of multiple projection points. For each projection point, a reference sample value of a preset type of attribute can be determined based on its position on the reference attribute plane. Sampling is performed along the projection points of the drilling trajectory projection to obtain reference sample values ​​of several preset types of attributes for the drilling trajectory projection. Sampling is performed on the drilling trajectory projection of each reference well to obtain reference sample values ​​of several preset types of attributes for each reference well.

[0057] Step 150: Use the reference sample values ​​of the preset type of attributes and the drilling oil and gas reserves to train the preset machine model to obtain the oil and gas reserve prediction model.

[0058] In this embodiment, reference sample values ​​of preset attribute types and well oil and gas reserves of each reference well are obtained as training samples to train a preset machine model, determine the parameters of the preset machine model, and obtain a prediction model for oil and gas reserves. In subsequent applications, target sample values ​​of several preset attribute types of the target well to be predicted can be collected and input into the prediction model for oil and gas reserves to obtain the predicted value of oil and gas reserves of the target well.

[0059] The method for constructing an oil and gas reserve prediction model in this embodiment involves acquiring back-stack drilling data volumes, drilling oil and gas reserves, and drilling trajectories of several reference wells; extracting attributes from the back-stack drilling data volumes to obtain reference measured values ​​for several preset types of attributes; determining a reference attribute plane based on the reference measured values ​​of each preset type of attribute; projecting each drilling trajectories onto the reference attribute plane to obtain drilling trajectory projections; sampling along the drilling trajectory projections to obtain reference sampled values ​​for several preset types of attributes; and training a preset machine model using the reference sampled values ​​of the preset types of attributes and the drilling oil and gas reserves to obtain an oil and gas reserve prediction model. When applying this oil and gas reserve prediction model, the numerical values ​​of the preset attribute types of the collected seismic data are input into the oil and gas reserve prediction model for prediction. This fully utilizes seismic data to predict oil and gas reserves, reducing reliance on well logging data. On the one hand, this avoids using the distribution of oil and gas sweet spots to assess the distribution of oil and gas reserves, and avoids the impact of the complex relationship between sweet spot distribution and oil and gas reserves on prediction accuracy. On the other hand, reducing reliance on well logging data allows this method to be extended to areas with only a small number of wells.

[0060] In one embodiment, the step of acquiring the post-drilling stack data volume of a plurality of reference wells includes:

[0061] Acquire pre-stack seismic data volumes from several of the aforementioned reference wells;

[0062] Based on the preset number of traces and the preset angle interval, gather extraction is performed on the pre-stack seismic data volume of each reference well to obtain several gather data of each reference well.

[0063] For each of the reference wells, the gather data of each well are superimposed to obtain a post-drilling data volume of several reference wells.

[0064] In this embodiment, to improve the accuracy of model prediction, pre-stack seismic data volumes from various reference wells in the reference well area are collected. For each pre-stack seismic data volume, gathers are extracted from multiple angles. The gather extraction method is determined according to the preset number of gathers and the preset angle interval. The preset number of gathers determines the number of gathers. Each gather has a preset angle interval, and gather extraction is performed within its preset angle interval. For example, gathers are divided into three types according to reflection angle: near gathers, middle gathers, and far gathers. The preset number of gathers is 3, the preset angle interval for near gathers is [0°, 12°), the preset angle interval for middle gathers is [12°, 24°), and the preset angle interval for far gathers is [24°, 36°). The preset number of gathers and the preset angle interval are adjusted according to the actual seismic data quality acquired in the application area. Generally, the span of the preset angle interval is between 5° and 15°, and the preset number of gathers is 3 to 5. Different angles from which gathers are extracted reflect different seismic reflection information, which can reflect the heterogeneity of the underground reservoir of the reference well, increase the amount of stratigraphic information obtained in the area where the reference well is located, and help improve the accuracy of prediction.

[0065] In one embodiment, the step of acquiring pre-stack seismic data volumes of the plurality of reference wells includes:

[0066] Acquire pre-stack seismic data and logging information from several of the aforementioned reference wells;

[0067] For each of the reference wells, the well logging information is used to perform well-seismic calibration on the pre-stack seismic data to obtain a pre-stack seismic data volume of several reference wells.

[0068] In this embodiment, pre-stack seismic data is time-domain information, with a seismic record format of x, y, and t. x and y are used to determine the position of a point in a plane, and t represents time. This seismic record format lacks a depth representation (z), making it impossible to directly locate depth z from the pre-stack seismic data. Well logging information, including well logging curves, is depth-domain information. Its well logging record format includes depth z, allowing for well-seismic calibration of the pre-stack seismic data. This involves converting between time and depth to obtain a well-calibrated pre-stack seismic data volume. This pre-stack seismic data volume more accurately reflects the formation information of the reference well, such as stratigraphy, faults, lithology, porosity, and permeability, which is beneficial for subsequent processing of the pre-stack seismic data volume.

[0069] In one embodiment, the step of extracting attributes from the post-drilling data volume to obtain reference measured values ​​of several preset types of attributes includes:

[0070] Each of the aforementioned drilling post-stack data volumes is subjected to frequency division processing to obtain frequency-divided post-stack data volumes for several preset frequency band intervals of each of the aforementioned reference drillings;

[0071] For each reference well, the target layer attributes are extracted from the frequency-divided stacked data volume of each preset frequency band interval to obtain reference measured values ​​of several preset types of attributes of the target layer of each reference well.

[0072] In this embodiment, the amount of information and resolution reflected by post-stack drilling data volumes in different frequency bands vary. Before seismic attribute extraction, the post-stack drilling data volumes are frequency-divided. For each post-stack drilling data volume, seismic frequency division is performed at fixed intervals, such as 5Hz. The frequency bands and number of frequency bands for frequency division are determined based on the quality of the post-stack drilling data volume, generally divided into no fewer than three frequency bands. Frequency division can improve the utilization rate of the post-stack drilling data volume, extract seismic information at different scales, and obtain frequency-divided post-stack drilling data volumes in different frequency bands after frequency division. For each frequency-division post-stack data volume of the same drilling post-stack data volume, the attribute of the target layer is extracted separately to obtain the measured values ​​of the preset type of attribute of each frequency-division post-stack data volume at the top, middle and bottom of the target layer. The average of the measured values ​​of the preset type of attribute of the same frequency-division post-stack data volume at the top, middle and bottom of the target layer is calculated to obtain the reference measured value of the preset type of attribute of the frequency-division post-stack data volume.

[0073] In one embodiment, the step of training a preset machine model using reference sampled values ​​of the preset type of attributes and the drilling oil and gas reserves to obtain a prediction model for oil and gas reserves includes:

[0074] Calculate the correlation between the reference sample value of each of the preset types of attributes and the drilling oil and gas reserves, and obtain the correlation coefficient of each of the preset types of attributes;

[0075] The attributes of each preset category are sorted in descending order of the absolute value of the correlation coefficient;

[0076] A number of different preset attribute quantities are obtained. For each preset attribute quantity, the reference sample value of the preset type of attribute of the highest preset attribute quantity and the drilling oil and gas reserves are selected to train the preset machine model to obtain a number of initial prediction models.

[0077] Calculate the accuracy of each initial prediction model, and select the initial prediction model with the highest accuracy as the prediction model for the oil and gas reserves.

[0078] In this embodiment, there are multiple preset categories of attributes, and each preset category of attribute has a different degree of correlation with drilled oil and gas reserves. That is, the correlation coefficients of different preset categories of attributes are different. The larger the correlation coefficient, the stronger the correlation between the preset category of attribute and drilled oil and gas reserves. To obtain a better reserve prediction model, preset categories of attributes are selected as indicators for training samples in rounds. First, the preset categories of attributes are sorted in descending order of the absolute value of the correlation coefficient. The earlier the preset category of attribute appears, the stronger its correlation with drilled oil and gas reserves. A number of preset categories of attributes are selected in each round, and the number of preset attributes decreases as the round increases. Each selection is based on the order of the absolute value of the correlation coefficient from largest to smallest, prioritizing the preset categories of attributes with higher correlation. Therefore, as the round increases, preset categories of attributes with smaller absolute values ​​of correlation coefficients are excluded, and these excluded preset categories of attributes are not used as training data in that round. For example, the preset attributes include attribute A, attribute B, attribute C, and attribute D. These attributes are arranged in descending order of the absolute value of their correlation coefficients, resulting in the order: attribute A, attribute B, attribute D, and attribute C. The number of preset attributes in each round is 4, 3, and 2 respectively. In the first three selections, the first round selects attributes A, B, D, and C; the second round selects attributes A, B, and D; and the third round selects attributes A and B. The preset attributes selected in each round are combined with drilling oil and gas reserves as training data for that round, used to train the preset machine learning model and obtain the initial prediction model for that round. Based on model evaluation rules, each initial prediction model is evaluated, and its accuracy is calculated. The initial prediction model with the highest accuracy is selected as the prediction model for the oil and gas reserves, thereby improving prediction accuracy.

[0079] Furthermore, the formula for calculating the correlation coefficient of the attributes of the preset categories is shown in (1):

[0080]

[0081] In equation (1), dcor is the distance correlation coefficient, a decimal; dcov(x,y) is the distance covariance between variables x and y; dvar(x,x) is the distance variance of variable x; and dvar(y,y) is the distance variance of variable y. The distance calculation in the pre-defined correlation coefficient formula uses the following formula: dist=|x i -y j |; i, j=1, 2, 3,..., n.

[0082] Furthermore, the preset machine model includes at least one of the Random Forest algorithm, XGBoost algorithm, and SVR algorithm; wherein, the objective function of the Random Forest algorithm is shown in equation (2):

[0083]

[0084] In one embodiment, the objective function of the XGBoost algorithm is shown in equation (3):

[0085]

[0086] In one embodiment, the objective function of the SVR algorithm is shown in equation (4):

[0087]

[0088] In equation (4), w represents the weight, C is the regularization function, and Ω is the default parameter for each weak learner. In this embodiment, when there are multiple preset machine learning models, the same training data and corresponding sample unit reserves can be used to train various preset machine learning models to obtain different trained models. The model accuracy of each trained model is calculated, and the trained model with the highest accuracy is selected as the reserve prediction model. In this way, by increasing the types of preset machine learning models, more trained models can be trained with the same training data, which is beneficial for selecting a better reserve prediction model.

[0089] Furthermore, the accuracy of the initial prediction model is evaluated using cross-validation rules, which are expressed in equations (5) and (6) for model evaluation.

[0090]

[0091]

[0092] Furthermore, the PSO algorithm or NS algorithm is used to optimize the parameters of the oil and gas reserve prediction model, where the PSO algorithm is shown in equations (7) and (8):

[0093]

[0094] In equations (7) and (8), n is the number of particles, c1 is the individual particle acceleration factor, c2 is the social particle acceleration factor, w is the inertial weight, vi is the particle velocity, xi is the particle position, pbest is the best position traversed by the i-th particle, and gbest is the best position traversed by all particles.

[0095] In one embodiment, pre-stack seismic data, logging information, drilling oil and gas reserves, and drilling trajectories of several reference wells are acquired. For each reference well, the pre-stack seismic data is calibrated using the logging information to obtain pre-stack seismic data volumes for the several reference wells. Based on a preset number of traces and a preset angle interval, gather extraction is performed on the pre-stack seismic data volumes of each reference well to obtain several gather data for each reference well. For each reference well, the gather data are superimposed to obtain post-stack data volumes for the several reference wells. Frequency division processing is performed on each post-stack data volume to obtain frequency-divided post-stack data volumes for several preset frequency band intervals for each reference well. For each reference well, the attribute extraction of the target layer is performed on the frequency-divided post-stack data volumes for each preset frequency band interval to obtain reference measured values ​​of several preset types of attributes of the target layer for each reference well.

[0096] In this embodiment, pre-stack seismic data is time-domain information, while well logging information, including well logging curves, belongs to the depth domain. Well logging curves can be used to perform well-seismic calibration on the pre-stack seismic data, resulting in well-calibrated pre-stack seismic data volumes. For each well-calibrated pre-stack seismic data volume, gather extraction is performed at multiple angles. The gather extraction method is determined based on a preset number of traces and a preset angle interval, thereby obtaining post-stack data volumes for several reference wells. For each post-stack data volume, seismic frequency division processing is performed at fixed intervals to extract seismic information at different scales. After frequency division, post-stack data volumes are obtained in different frequency bands. For each frequency-division post-stack data volume of the same drilling post-stack data volume, the attribute of the target layer is extracted separately to obtain the measured values ​​of the preset type of attribute of each frequency-division post-stack data volume at the top, middle and bottom of the target layer. The average of the measured values ​​of the preset type of attribute of the same frequency-division post-stack data volume at the top, middle and bottom of the target layer is calculated to obtain the reference measured value of the preset type of attribute of the frequency-division post-stack data volume.

[0097] Example 2:

[0098] The method for constructing the oil and gas reserve prediction model in this embodiment can be applied to electronic devices with communication, computing, and data storage capabilities. The specific process includes:

[0099] Step S1: The device's data collection module receives and initially organizes the data.

[0100] In this embodiment, the data to be collected includes: processed pre-stack seismic data, processed post-stack seismic data, seismic interpretation stratigraphic information, wellhead information of mature development areas corresponding to the seismic data, well trajectory information, corresponding logging curves (mainly including sonic logging curves and density logging curves used for well-seismic calibration), drilling intensity of the target reservoir, and single-well EUR. This application does not focus on the seismic data processing process; therefore, the collected seismic data should be data that has undergone denoising, frequency upscaling, and correction after seismic data acquisition. Otherwise, it will affect the accuracy of this method in predicting unconventional oil and gas recoverable reserves. After all data is collected, well-seismic calibration is performed using logging curves in a synthetic recording manner, and the spatial positional relationship between the logging depth domain information and the seismic data time domain information is matched.

[0101] Step S2: The device's data service module performs pre-stack seismic data processing.

[0102] In this embodiment, pre-stack or post-stack seismic data are first extracted into gathers at different angles, generally divided into three gathers according to reflection angle: near gathers ([0°, 12°), middle gathers ([12°, 24°), and far gathers ([24°, 36°)). The angle boundaries and the number of gathers are adjusted based on the actual seismic data quality acquired in the application area. Generally, the angle interval is between [5, 15], and the number of gathers is between [3, 5]. After the gathers are divided, the data from each gather are superimposed to obtain the post-stack data volume of the corresponding gather.

[0103] Step S3: The device's calculation service module performs seismic frequency division processing.

[0104] In this embodiment, seismic data of different frequencies reflect different amounts of information and resolutions. Before extracting seismic attributes, the seismic data undergoes frequency division processing. Different post-stack data volumes are divided into seismic frequency volumes at fixed intervals of 5Hz or other intervals. The specific frequency division process depends on the quality of the seismic data, and generally, it is divided into no fewer than three frequency bands based on bandwidth. In this embodiment, the purpose of frequency division is to improve the utilization rate of post-stack seismic data and extract seismic information at different scales. After frequency division, post-stack seismic data volumes with different frequency bands are obtained.

[0105] Step S4: The device's calculation service module extracts seismic attributes.

[0106] In this embodiment, after steps S2 and S3 are completed, seismic attribute extraction is performed. The extraction method is based on the seismic interpretation horizon of the target layer, and the distances from the top and bottom of the target layer to the interpretation horizon are converted according to the seismic wave propagation velocity, and time windows t are opened upward and downward respectively. First, around the target layer, the amplitude, frequency, phase, and other attributes of the pre-stack seismic gathering post-stack data and post-stack seismic data are extracted using time window t. For example, the maximum peak amplitude, maximum trough amplitude, mean peak amplitude, root mean square amplitude, total amplitude, mean instantaneous frequency, mean instantaneous phase, mean energy, and total energy are extracted as the middle attributes of the target layer. Then, the time window is shifted upward by t / 2, and the same type of attributes are extracted again. These extracted attributes are used as the top constraint attributes of the target layer. Finally, the time window is shifted downward by t, and the same type of attributes are extracted again. These extracted attributes are used as the bottom constraint attributes of the target layer. In the end, a target layer will have three sets of several types of attributes extracted, namely upper, middle, and lower. All these attributes are used as the final result of the seismic attribute extraction of the target layer.

[0107] In this embodiment, the seismic data is three-dimensional, and after attribute extraction, it becomes two-dimensional. The attributes are two-dimensional data, in the form of x yz, with different attributes equivalent to z1, z2, and z3. The label dataset format is the table mentioned earlier, and a similar format consists of the data in all columns except for the column for the unit horizontal segment length EUR, which serves as the input data. The goal is to predict the unit horizontal segment length EUR.

[0108] Step S5: The device's computational service module performs seismic attribute analysis of unconventional oil and gas recoverable reserves per unit length.

[0109] In this embodiment, the horizontal well section collected in step S1 is projected onto the extracted seismic attribute plane. The seismic data trace spacing is used as the plane sampling interval. The trajectory of each well is projected onto the attribute plane to sample all extracted attributes. After all attributes are sampled, the first and last data points of each attribute for each well are removed. If there are only one or two sampling points, they are not removed. Then, the average value is taken as the final value of the seismic attribute for each well.

[0110] In this embodiment, the single-well EUR collected in step S1 is processed by dividing the single-well EUR (the final recoverable reserve assessed by a single well) by the horizontal segment length to obtain the unit horizontal segment length EUR. The final seismic attribute value of each well is combined with the unit horizontal segment length EUR to form a labeled dataset, as shown in Table 1 below. Various seismic attributes are used as input variables, and recoverable reserves per unit length are used as target variables.

[0111] Table 1. Seismic Attributes After Sorting

[0112]

[0113] In this embodiment, each input variable is normalized and the parameters of the normalization operation are recorded. The endpoint values ​​of each variable during normalization are also used as endpoint values ​​in subsequent normalization of the predicted data. The specific formula is X'=(X-Xmin) / (Xmax-Xmin), where the endpoint values ​​are Xmin and Xmax. The correlation coefficient analysis method is used to analyze the correlation between each input variable and the target variable. The greater the correlation, the higher the importance of the input variable. The input variables are sorted from high to low importance. The specific formula for the distance correlation coefficient is shown in equation (1):

[0114]

[0115] In equation (1), dcor is the distance correlation coefficient, a decimal; dcov(x,y) is the distance covariance between variables x and y; dvar(x,x) is the distance variance of variable x; and dvar(y,y) is the distance variance of variable y.

[0116] The formula for calculating the distance between variables is as follows (taking variable x as an example):

[0117] dist = |x i -x j |,i,j=1,2,3,...,n.

[0118] S6: The device's computing service module establishes a machine learning algorithm model for the unit horizontal segment length EUR.

[0119] In this embodiment, a machine learning algorithm capable of performing regression simulation is selected, such as the Random Forest algorithm, XGBoost algorithm, and SVR algorithm, and the parameter settings in the algorithm model utilize the algorithm's default parameters.

[0120] In one embodiment, the objective function of the Random Forest algorithm is shown in equation (2):

[0121]

[0122] In one embodiment, the objective function of the XGBoost algorithm is shown in equation (3):

[0123]

[0124] In one embodiment, the objective function of the SVR algorithm is shown in equation (4):

[0125]

[0126] In equation (4), w is the weight, C is the regularization function, and Ω is the default parameter for each weak learner.

[0127] In this embodiment, the labeled dataset organized in step S5 is used, with various seismic attributes serving as input variables and the unit horizontal segment length (EUR) as the target variable. First, a machine learning algorithm model is trained and its accuracy is evaluated using all input variables, and the accuracy results are recorded. Then, according to the importance ranking of the input variables in step S5, the input variable with the lowest importance is deleted, and the machine learning algorithm model is trained and its accuracy is evaluated again, recording the accuracy results. This process is repeated until only one input variable is used for training and accuracy evaluation, and the accuracy results are recorded. The input variable with the highest accuracy is selected as the optimal input variable combination. If multiple machine learning methods are used to predict EUR, different machine learning algorithm models can be used, and the above operations can be repeated to obtain the optimal input variable combination for each algorithm.

[0128] In this embodiment, the model evaluation method adopts a unified cross-validation approach, and the cross-validation is performed using the formulas shown in equations (5) and (6):

[0129]

[0130] In this embodiment, the PSO algorithm or the NS algorithm is used for model parameter tuning, wherein the PSO algorithm is shown in equations (7) and (8):

[0131]

[0132] In equations (7) and (8), n is the number of particles, c1 is the individual particle acceleration factor, c2 is the social particle acceleration factor, w is the inertial weight, vi is the particle velocity, xi is the particle position, pbest is the best position traversed by the i-th particle, and gbest is the best position traversed by all particles.

[0133] In this embodiment, the model parameter combination and training result with the highest accuracy in the model evaluation results are selected as the final machine learning model for predicting the EUR (Earnings per Horizontal Segment) length per unit level. If multiple machine learning algorithms are used, the above steps are repeated after replacing different models, and finally, the best model is selected through comparison.

[0134] Step S7: The device's computational service module performs planar prediction of unconventional oil and gas recoverable reserves.

[0135] Seismic attributes extracted from the seismic body are planar data. Each attribute is organized into a format similar to the labeled dataset, using coordinates x, y, or inline, xline plus attribute values, forming the dataset to be predicted. Using the optimal model determined in step S6, the dataset to be predicted is normalized according to the normalization parameters recorded in S5. After processing, the optimal combination of input variables determined in S6 is selected as the input variables and input into the optimal model for prediction, obtaining the unconventional oil and gas unit horizontal segment length (EUR), thus forming the planar distribution of the unconventional oil and gas unit horizontal segment length (EUR). Multiplying the unconventional oil and gas unit horizontal segment length (EUR) obtained from the planar map by the horizontal segment length yields the unconventional oil and gas EUR of the target well.

[0136] In this embodiment, by fully leveraging seismic information and deeply integrating machine learning with seismic attribute analysis, the problem of predicting unconventional oil and gas reservoir EURs in undrilled areas is solved. Previous predictions of unconventional oil and gas EURs relied on drilling and logging data, which were difficult to implement in undrilled areas. Previous researchers used seismic inversion methods for pre-drilling studies in undrilled areas. However, these methods, derived from theoretical models, are complex and only suitable for evaluating and selecting sweet spots in unconventional oil and gas reservoirs. This invention, however, can utilize seismic data in undrilled areas to provide the planar distribution of EURs per unit horizontal segment length of the target layer, offering crucial reference for early decision-making and overall deployment in unconventional oil and gas exploration and development.

[0137] The execution process of unconventional oil and gas recoverable reserves prediction is as follows: Figure 2 As shown in this embodiment, the data collection module acquires and processes pre-stack seismic data, processed post-stack seismic data, seismic interpretation stratigraphic information, wellhead information of mature development areas corresponding to the seismic data, well trajectory information, corresponding logging curves (mainly including sonic logging curves and density logging curves used for well-seismic calibration), drilling intensity of the target reservoir, and single-well EUR. After time-depth matching, a basic data dataset is formed and sent to the data service module. The data service module performs channel set extraction and overlay of the pre-stack seismic data and sends channel set extraction and overlay setting parameters to the storage service module. The data service module also processes the frequency-division data of the overlay seismic data, sends preprocessing setting parameters to the storage service module, and sends preprocessing result data to the computation service module. The computation service module receives the preprocessing result data, performs multi-layer constraint extraction of seismic attributes, analyzes the seismic attributes of recoverable reserves per unit length of shale gas, trains a machine learning model for recoverable reserves per unit length of shale gas, and sends the trained model to the storage service module.

[0138] The process of predicting the planar distribution of recoverable reserves is as follows: Figure 3As shown in this embodiment, the data collection module acquires pre-stack seismic data, post-stack seismic data, and seismic interpretation horizon information of the target layer in the area to be predicted, and sends the data to the data server. The storage server sends the set extraction and stacking settings parameters to the data server. The data server performs set extraction and stacking of the pre-stack seismic data, receives the preprocessing settings parameters sent by the storage server, processes the frequency-divided data of the stacked seismic data, and sends the processed data to the computing server. The computing server extracts seismic attributes through multi-layer constraints, receives the trained model sent by the storage server, inputs the extracted seismic attributes into the trained model, predicts the planar distribution of recoverable reserves per unit length of shale gas, and sends the prediction results data to the storage server.

[0139] See Figure 4 In one embodiment, the recoverable reserves prediction method includes: data collection and processing, extraction and stacking of pre-stack seismic data gathers, frequency-division processing of stacked seismic data, multi-layer constraint extraction of seismic attributes, seismic attribute analysis of recoverable reserves per unit length of shale gas, establishment of a machine learning model for recoverable reserves per unit length of shale gas, and application of the machine learning model for recoverable reserves per unit length of shale gas.

[0140] See Figure 5 The recoverable reserves prediction method is implemented using a seismic prediction device for the planar distribution of shale gas recoverable reserves. The prediction device includes an acquisition module, a data processing module, a training calculation module, a prediction calculation module, and a storage module. The acquisition module acquires training data. The data processing module preprocesses the training data. The training calculation module trains the machine model using the processed training data to obtain the trained model, which is then stored in the storage module. The prediction calculation module retrieves the trained model from the storage module and uses it to predict the recoverable reserves of the data to be tested.

[0141] See Figure 6 An electronic device for performing recoverable reserves prediction methods includes an input device, a controller, a memory, a display, and a monitor. The input device is used to acquire seismic data, the controller is used to control the various execution steps of the training and prediction processes, the memory can be used to store parameters for gather extraction and overlay settings, as well as the trained model, the processor is used to train the model, and the monitor is used to display the shale gas EUR plane prediction map.

[0142] This application can be applied to the prediction of unconventional oil and gas resources such as tight sandstone gas, shale oil, and shale gas extracted using fracturing technology.

[0143] In the planar prediction of recoverable reserves of shale gas reservoirs, the data collection module was first used to collect and organize data such as seismic data, seismic interpretation stratigraphic information, well number and well trajectory information, sonic logging curves, density logging curves, and the length of horizontal well sections of shale gas wells. Well logging curves were used to perform well-seismic calibration in a synthetic recording manner, and the spatial positional relationship between the well logging depth domain information and the time domain information of the seismic data was matched.

[0144] Then, in the data service module, based on the seismic interpretation horizon of the target layer, the seismic data is processed according to steps S2 and S3, and then seismic attributes are extracted. Fourteen seismic attributes are extracted, namely Average energy, Average instantaneous, Average instantaneous phase, Average magnitude, Maximum amplitude, Maximum magnitude, Meanamplitude, Most of, RMS amplitude, Standard deviation of amplitude, Sum of amplitudes, and Sum of energy. Seismic attribute values ​​are extracted along the well trajectory, normalized, and used to form a training dataset. The seismic attributes are used as labels, and the EUR per unit horizontal well section length is used as the target variable, as shown in the table below:

[0145]

[0146]

[0147] In this embodiment, the unit horizontal segment length EUR is the target variable (output variable) Y, and the other variables are the input variables X. The importance of each input variable is calculated and ranked using distance coefficients.

[0148] In this embodiment, a machine learning algorithm model is selected in the computing service module. The XGBoost algorithm model is used here. Input variables are sequentially input into the XGBoost model according to their importance, from most to least important, for training. The relative error index is used to evaluate the impact of the number of input variables on the model's algorithm results each time. The results of this process are plotted as follows: Figure 7 The line graph shown, from Figure 7Find the point with the smallest relative error. The number of variables corresponding to this point is the optimal number of input parameters for the XGBoost algorithm model. In this example, it is 8. Based on the ranking of the importance of the input variables, the optimal combination of input parameters can be determined as Sum of amplitudes, Mean amplitude, Average instantaneous phase, Average magnitude, Average energy, Sum of energy, RMS amplitude, and Time.

[0149] In this embodiment, the trained XGBoost algorithm model is saved to the storage service module for use in shale gas recoverable reserve prediction. Seismic data for the area to be predicted is collected, processed through steps S2 and S3 of the data service module, and subjected to seismic attribute extraction as described in step S4 by the calculation module. The trained XGBoost algorithm model is received from the storage service module, and the calculation service module is used to predict the EUR plane distribution of a single shale gas well, and the results are saved to the storage service module. The trained model is then used to perform plane EUR prediction on the work area, and the results are as follows. Figure 8 As shown.

[0150] In this embodiment, several developed wells exist within the work area, and the predicted EUR values ​​at the corresponding well points can be extracted. The predicted results are compared with the actual values, and the comparison results are as follows: Figure 9 As shown, the overall prediction results are good, with an average relative error of 20.1%, which meets the requirements of the exploration stage.

[0151] In the planar prediction of recoverable reserves of low-permeability sandstone reservoirs, a case study of a low-permeability sandstone reservoir in eastern my country is used for illustration.

[0152] First, we collected and organized data such as seismic data, seismic interpretation stratigraphic information, well logging curves, horizontal well section lengths of oil wells, and single-well recoverable reserves calibration. We then used well logging curves to perform well-seismic calibration in a synthetic recording manner and matched the spatial location relationship between well logging depth domain information and seismic data time domain information.

[0153] Then, in the data service module, based on the seismic interpretation horizon of the target layer, the seismic data is processed according to steps S2 and S3, and then seismic attributes are extracted. Sixteen seismic attributes are extracted, namely Average energy, Average instantaneous, Average instantaneous phase, Average magnitude, Maximum amplitude, Maximum magnitude, Meanamplitude, Most of, RMS amplitude, Standard deviation of amplitude, Sum of amplitudes, and Sum of energy. Seismic attribute values ​​are extracted along the well trajectory, normalized, and used to form a training dataset. Seismic attributes are used as labels, and the EUR per unit horizontal well section length is used as the target variable to form a training data table.

[0154] The unit horizontal segment length EUR is the target variable (output variable) Y, and the remaining variables are the input variables X. The importance of each input variable is calculated and ranked using distance coefficients. The ranking results of the distance coefficients are as follows: Figure 10 As shown, the selected attributes are RMSamplitude, Average energy, Average instantaneous phase, Average magnitude, Sumof energy, Time, Sum of amplitudes, and Mean amplitude.

[0155] In this embodiment, the trained XGBoost algorithm model is saved to the storage service module for use in predicting the recoverable reserves of low-permeability sandstone reservoirs. Seismic data for the area to be predicted is collected, processed in steps S2 and S3 by the data service module, and the seismic attributes described in step S4 are extracted by the calculation module. The trained XGBoost algorithm model is received from the storage service module, and the calculation service module is used to predict the plane distribution of EUR (Earnings Regime) in a single well, and the results are saved to the storage service module. The trained model is then used to predict the plane EUR distribution of the work area, and the prediction results are as follows: Figure 11 As shown.

[0156] In this embodiment, several developed wells exist within the work area, and the predicted EUR values ​​at the corresponding well points can be extracted. The predicted results are compared with the actual values, and the comparison results are as follows: Figure 12 As shown, the overall prediction results are good, with an average relative error of 18.9%, which meets the requirements of the exploration stage.

[0157] In this embodiment, the attribute is the value distribution on a plane. A well segment is a section within an underground wellbore. A trajectory refers to the underground trajectory of the well. A well segment is a section of the well trajectory. Well segments and trajectories exist in three-dimensional space, while attributes exist in planar space. Therefore, the trajectory needs to be projected onto the attribute plane to determine which data to collect on the attribute plane. Sampling involves collecting data at certain intervals on the attribute plane covered by the trajectory projection. The trajectory information of the well can be used to extract the length of the horizontal segment. The horizontal well trajectory is designed with (x, y, h) three-dimensional coordinates, which are used to calculate the length of the horizontal segment.

[0158] In this embodiment, the attribute plane can contain the trajectories of multiple wells. A well segment is a part of the well trajectory, and all projections are corresponding. Different attributes reside on the attribute plane.

[0159] In this embodiment, the final value is used as the input variable, and EUR is used as the target variable. These are then correlated through operations such as wellbore calibration, attribute extraction, projection, and sampling.

[0160] In this embodiment, all data is divided into two categories: a training set and a test set, which are randomly selected. The training set is used to establish the relationship between input and output variables, while the test set is used to verify this relationship. A set of input and target variables constitutes a training object, and many sets together form the training set.

[0161] This application provides a device, electronic equipment, and method for early-stage unconventional oil and gas EUR prediction in unconventional oil and gas reservoir exploration and development based on seismic data. It primarily addresses the difficulty in predicting or quantifying unconventional oil and gas EUR in un-drilled areas. The device enhances seismic information preprocessing based on seismic data and deeply integrates machine learning algorithms with seismic attribute analysis. It includes a data collection module, a data processing module, a training calculation module, a prediction calculation module, and a storage module. It can provide planar distribution prediction results of EUR per unit horizontal segment length in target layers in un-drilled areas using seismic data, providing important reference for well location deployment, production decisions, and overall deployment in unconventional oil and gas exploration and development.

[0162] Example 3:

[0163] The oil and gas reserve prediction method of this embodiment can be applied to electronic devices with communication, computing, and data storage capabilities. Its specific process includes:

[0164] Obtain the target seismic data volume of the target well, and the oil and gas reserve prediction model constructed by the oil and gas reserve prediction model construction method described in any of the above embodiments;

[0165] The target seismic data volume is subjected to attribute extraction to obtain the target measured values ​​of several preset types of attributes;

[0166] Based on the target measured values ​​of each of the preset types of attributes, a target attribute plane is determined, and the trajectory of the target well is projected onto the target attribute plane to obtain the target trajectory projection.

[0167] Sampling is performed along the target trajectory projection to obtain target sample values ​​of several preset types of attributes;

[0168] Based on the oil and gas reserve prediction model, the oil and gas reserve prediction value of the target well is obtained by calculating the target sampled values ​​of several preset types of attributes.

[0169] In this embodiment, the target well is the well whose oil and gas reserve prediction value needs to be estimated. The target seismic data volume can be the post-stack seismic data volume of the target well in different frequency bands after sequential well calibration, gather extraction, and frequency division processing of the pre-stack seismic data of the target well. Attribute extraction is performed on this data to obtain target measured values ​​of several preset types of attributes of the target well. Based on the target measured values ​​of the preset types of attributes, a target attribute plane of the target well is determined. The trajectory of the target well is projected onto the target attribute plane to obtain the target trajectory projection of the target well. Attribute sampling is performed on this projection to obtain target sampled values ​​of several preset types of attributes of the target well. These target sampled values ​​of the preset types of attributes are input into the oil and gas reserve prediction model to calculate the predicted oil and gas reserve value of the target well.

[0170] If the predicted oil and gas reserves are in units of EUR, then the unit length of the target well can be obtained. The unit horizontal segment length can be extracted from the trajectory of the target well. In three-dimensional space, the trajectory of the target well is represented by three-dimensional coordinates (x, y, h), and the horizontal segment length is calculated using these coordinates. Then, the horizontal segment length is multiplied by the oil and gas reserves per unit horizontal segment length to obtain the total oil and gas reserves of the target well. Subsequently, the exploitation value of the target well can be assessed based on the total oil and gas reserves.

[0171] Example 4:

[0172] Another embodiment of this application relates to a device for constructing a prediction model of oil and gas reserves. The implementation details of the device for constructing a prediction model of oil and gas reserves in this embodiment are described in detail below. The following implementation details are provided for ease of understanding and are not necessary for implementing this solution. The device for constructing a prediction model of oil and gas reserves in this embodiment includes a drilling data acquisition module, a post-stack attribute extraction module, a trajectory projection plane module, a projection attribute sampling module, and a machine model training module.

[0173] The drilling data acquisition module is used to acquire the drilling back-stack data volume, drilling oil and gas reserves, and drilling trajectory of several reference wells;

[0174] The post-stack attribute extraction module is used to extract attributes from the post-stack drilling data volume to obtain reference measured values ​​of several preset types of attributes.

[0175] The trajectory projection plane module is used to determine a reference attribute plane based on the reference measured values ​​of the attributes of each preset type, and to project each drilling trajectory onto the reference attribute plane to obtain the drilling trajectory projection.

[0176] The projection attribute sampling module is used to sample along the drilling trajectory projection to obtain reference sample values ​​of several preset types of attributes;

[0177] The machine model training module is used to train a preset machine model using reference sample values ​​of the preset type of attributes and the drilling oil and gas reserves to obtain a prediction model for oil and gas reserves.

[0178] It is worth mentioning that all modules involved in this embodiment are logical modules. In practical applications, a logical unit can be a physical unit, a part of a physical unit, or a combination of multiple physical units. Furthermore, to highlight the innovative aspects of this application, this embodiment does not introduce units that are not closely related to solving the technical problems proposed in this application; however, this does not mean that other units are absent in this embodiment.

[0179] Example 5:

[0180] Another embodiment of this application relates to an oil and gas reserve prediction device. The implementation details of the oil and gas reserve prediction device of this embodiment are described in detail below. The following implementation details are provided for ease of understanding and are not necessary for implementing this solution. The oil and gas reserve prediction device of this embodiment includes a target seismic data module, a target attribute extraction module, a target trajectory projection module, a target trajectory sampling module, and a target reserve prediction module.

[0181] The target seismic data module is used to acquire the target seismic data volume of the target well, and the oil and gas reserve prediction model constructed by the oil and gas reserve prediction model construction method described in any of the above embodiments.

[0182] The target attribute extraction module is used to extract attributes from the target seismic data volume to obtain the target measured values ​​of several preset types of attributes.

[0183] The target trajectory projection module is used to determine the target attribute plane based on the measured values ​​of the target attributes of each preset type, and to project the trajectory of the target drilling onto the target attribute plane to obtain the target trajectory projection.

[0184] The target trajectory sampling module is used to sample along the projection of the target trajectory to obtain target sample values ​​of several preset types of attributes;

[0185] The target reserve prediction module is used to calculate the target well's oil and gas reserve prediction value based on the oil and gas reserve prediction model and the target sample values ​​of several preset types of attributes.

[0186] It is worth mentioning that all modules involved in this embodiment are logical modules. In practical applications, a logical unit can be a physical unit, a part of a physical unit, or a combination of multiple physical units. Furthermore, to highlight the innovative aspects of this application, this embodiment does not introduce units that are not closely related to solving the technical problems proposed in this application; however, this does not mean that other units are absent in this embodiment.

[0187] Example 6:

[0188] Another embodiment of this application relates to an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method for constructing a prediction model for oil and gas reserves in the above embodiments.

[0189] The memory and processor are connected via a bus, which can include any number of interconnecting buses and bridges, connecting various circuits of one or more processors and memories. The bus can also connect various other circuits, such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and will not be described further herein. The bus interface provides an interface between the bus and the transceiver. The transceiver can be a single element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices over a transmission medium. Data processed by the processor is transmitted over the wireless medium via an antenna, which further receives data and transmits it to the processor.

[0190] The processor manages the bus and general processing, and also provides various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory is used to store data used by the processor during operation.

[0191] Example 7:

[0192] Another embodiment of this application relates to a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the method embodiments described above.

[0193] That is, those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. This program is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0194] Those skilled in the art will understand that the above embodiments are specific embodiments for implementing this application, and in practical applications, various changes can be made to them in form and detail without departing from the spirit and scope of this application.

Claims

1. A method for constructing a prediction model for oil and gas reserves, characterized in that, include: Obtain the back-stack data volume, drilling oil and gas reserves, and drilling trajectory of several reference wells; Attribute extraction is performed on the drilling stack data volume to obtain reference measured values ​​of several preset types of attributes; A reference attribute plane is determined based on the reference measured values ​​of the attributes of each of the preset types, and each of the drilling trajectories is projected onto the reference attribute plane to obtain the drilling trajectory projection. Sampling is performed along the projection of the drilling trajectory to obtain reference sample values ​​of several preset types of attributes; The preset machine model is trained using reference sample values ​​of the preset type of attributes and the drilling oil and gas reserves to obtain a prediction model for oil and gas reserves.

2. The construction method according to claim 1, characterized in that, The step of obtaining post-drilling stack data volumes of several reference wells includes: Acquire pre-stack seismic data volumes from several of the aforementioned reference wells; Based on the preset number of traces and the preset angle interval, gather extraction is performed on the pre-stack seismic data volume of each reference well to obtain several gather data of each reference well. For each of the reference wells, the gather data of each well are superimposed to obtain a post-drilling data volume of several reference wells.

3. The construction method according to claim 2, characterized in that, The step of acquiring pre-stack seismic data volumes of the reference wells includes: Acquire pre-stack seismic data and logging information from several of the aforementioned reference wells; For each of the reference wells, the well logging information is used to perform well-seismic calibration on the pre-stack seismic data to obtain a pre-stack seismic data volume of several reference wells.

4. The construction method according to claim 1, characterized in that, The step of extracting attributes from the post-drilling data volume to obtain reference measured values ​​for several preset types of attributes includes: Each of the aforementioned drilling post-stack data volumes is subjected to frequency division processing to obtain frequency-divided post-stack data volumes for several preset frequency band intervals of each of the aforementioned reference drillings; For each reference well, the target layer attributes are extracted from the frequency-divided stacked data volume of each preset frequency band interval to obtain reference measured values ​​of several preset types of attributes of the target layer of each reference well.

5. The construction method according to claim 1, characterized in that, The step of training a preset machine model using reference sample values ​​of the preset type of attributes and the drilling oil and gas reserves to obtain a prediction model for oil and gas reserves includes: Calculate the correlation between the reference sample value of each of the preset types of attributes and the drilling oil and gas reserves, and obtain the correlation coefficient of each of the preset types of attributes; The attributes of each preset category are sorted in descending order of the absolute value of the correlation coefficient; A number of different preset attribute quantities are obtained. For each preset attribute quantity, the reference sample value of the preset type of attribute of the highest preset attribute quantity and the drilling oil and gas reserves are selected to train the preset machine model to obtain a number of initial prediction models. Calculate the accuracy of each initial prediction model, and select the initial prediction model with the highest accuracy as the prediction model for the oil and gas reserves.

6. A method for predicting oil and gas reserves, characterized in that, include: Acquire the target seismic data volume of the target well, and construct the oil and gas reserve prediction model by the method of constructing the oil and gas reserve prediction model according to any one of claims 1 to 5. The target seismic data volume is subjected to attribute extraction to obtain the target measured values ​​of several preset types of attributes; Based on the target measured values ​​of each of the preset types of attributes, a target attribute plane is determined, and the trajectory of the target well is projected onto the target attribute plane to obtain the target trajectory projection. Sampling is performed along the target trajectory projection to obtain target sample values ​​of several preset types of attributes; Based on the oil and gas reserve prediction model, the oil and gas reserve prediction value of the target well is obtained by calculating the target sampled values ​​of several preset types of attributes.

7. An apparatus for constructing a predictive model for oil and gas reserves, characterized in that, include: The drilling data acquisition module is used to acquire the drilling back-stack data volume, drilling oil and gas reserves, and drilling trajectory of several reference wells; The post-stack attribute extraction module is used to extract attributes from the post-stack drilling data volume to obtain reference measured values ​​of several preset types of attributes. The trajectory projection plane module is used to determine a reference attribute plane based on the reference measured values ​​of the attributes of each preset type, and to project each drilling trajectory onto the reference attribute plane to obtain the drilling trajectory projection. The projection attribute sampling module is used to sample along the drilling trajectory projection to obtain reference sample values ​​of several preset types of attributes; The machine model training module is used to train a preset machine model using reference sample values ​​of the preset type of attributes and the drilling oil and gas reserves to obtain a prediction model for oil and gas reserves.

8. A device for predicting oil and gas reserves, characterized in that, include: The target seismic data module is used to acquire the target seismic data volume of the target well and the oil and gas reserve prediction model constructed by the method of constructing the oil and gas reserve prediction model according to any one of claims 1 to 5. The target attribute extraction module is used to extract attributes from the target seismic data volume to obtain the target measured values ​​of several preset types of attributes. The target trajectory projection module is used to determine the target attribute plane based on the measured values ​​of the target attributes of each preset type, and to project the trajectory of the target drilling onto the target attribute plane to obtain the target trajectory projection. The target trajectory sampling module is used to sample along the projection of the target trajectory to obtain target sample values ​​of several preset types of attributes; The target reserve prediction module is used to calculate the target well's oil and gas reserve prediction value based on the oil and gas reserve prediction model and the target sample values ​​of several preset types of attributes.

9. An electronic device, characterized in that, include: At least one processor; as well as, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the method for constructing a prediction model for oil and gas reserves as described in any one of claims 1 to 6, or the method for predicting oil and gas reserves as described in claim 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the method for constructing the oil and gas reserves prediction model according to any one of claims 1 to 6, or the method for predicting oil and gas reserves according to claim 7.

Images