Sedimentary facies prediction device and method based on three-dimensional seismic attribute fusion

Abstract

The invention belongs to the technical field of petroleum and natural gas exploration and development, and discloses a sedimentary facies prediction device and method based on three-dimensional seismic attribute fusion. The device comprises a determination module, a training module, a prediction module and a construction module. A three-dimensional seismic attribute body is screened through a principal component analysis method, and a three-dimensional convolutional data network is trained through well area data, so that prediction of the sedimentary facies of a well-free area is realized, a prediction result is optimized by synchronously utilizing sequential Gaussian co-location collaborative simulation, a final sedimentary facies geologic model is obtained, and lithologic reservoir body searching is realized. The prediction result is constrained by various data, the accuracy is higher, a real geological mode is better met, meanwhile, the prediction efficiency is improved, and the man-made interference degree is reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of oil and gas exploration and development, and in particular relates to a sedimentary phase prediction device and method based on three-dimensional seismic attribute fusion. Background Art

[0002] After years of exploration, undiscovered structural oil and gas reservoirs are becoming increasingly rare, and lithologic oil and gas reservoirs have gradually become the focus. Widely distributed lithologic oil and gas reservoirs have huge potential production, so the prediction of the distribution of lithologic reservoirs and three-dimensional geological modeling are crucial. From the distribution law, lithologic traps are controlled by sedimentary facies belts, and different sedimentary facies limit the distribution range and reservoir performance of lithologic traps. Therefore, the exploration of lithologic traps has a strong support for the prediction of sedimentary facies distribution.

[0003] In terms of sedimentary facies distribution prediction, related technologies have proposed methods such as sequential indicator simulation and seismic sedimentary facies prediction. These methods have achieved certain results in sedimentary facies identification, but the predicted sedimentary facies distribution is quite different from the actual situation in the study area. Summary of the invention

[0004] In order to overcome the shortcomings of the prior art, the present invention provides a sedimentary phase prediction device and method based on three-dimensional seismic attribute fusion, which determines the principal component three-dimensional seismic attribute body through three-dimensional seismic data and principal component analysis; constructs and trains a three-dimensional convolutional neural network model, utilizes the relationship between wellbore data and the principal component three-dimensional seismic attribute body, utilizes the three-dimensional convolutional data network model to predict the sedimentary phase and constructs a three-dimensional probability data body, and uses sequential Gaussian isotopic collaborative simulation to construct a sedimentary phase geological model, thereby providing a method for finding lithological reservoirs.

[0005] The above object of the present invention is achieved through the following technical solutions:

[0006] Sedimentary facies prediction device based on 3D seismic attribute fusion, including:

[0007] A determination module, used to determine a principal component three-dimensional seismic attribute volume of the target well area based on the three-dimensional seismic data of the target well area and a principal component analysis method;

[0008] A training module, used to construct and train a three-dimensional convolutional neural network model to learn the relationship between the wellbore data of the target well area and the main component three-dimensional seismic attribute volume; the wellbore data includes logging data and sedimentary phase data;

[0009] A prediction module, used to predict the sedimentary phase of the well-free area based on the trained three-dimensional convolutional data network model and construct a three-dimensional probability data volume;

[0010] A construction module is used to construct a sedimentary facies geological model based on the three-dimensional probability data body through sequential Gaussian isotopic collaborative simulation, and the sedimentary facies geological model is used to find lithological reservoirs.

[0011] Another object of the present invention is to protect the prediction method of the above-mentioned sedimentary facies prediction device based on three-dimensional seismic attribute fusion, the steps are:

[0012] 1. Based on the 3D seismic data of the target well area and the principal component analysis method, determine the principal component 3D seismic attribute volume of the target well area;

[0013] 2. Construct and train a three-dimensional convolutional neural network model to learn the relationship between the wellbore data of the target well area and the principal component three-dimensional seismic attribute volume; the wellbore data includes logging data and sedimentary phase data;

[0014] 3. Predict the sedimentary facies in the well-free area based on the trained three-dimensional convolutional data network model and construct a three-dimensional probability data volume;

[0015] 4. Based on the three-dimensional probabilistic data volume, a sedimentary facies geological model is constructed through sequential Gaussian isotopic collaborative simulation, and the sedimentary facies geological model is used to find lithological reservoirs.

[0016] Furthermore, the step 1 is specifically as follows: extracting volume attributes based on the three-dimensional seismic data to obtain the original three-dimensional seismic attribute volume of the target well area; and screening the original three-dimensional seismic attribute volume using principal component analysis to obtain the principal component three-dimensional seismic attribute volume of the target well area.

[0017] Furthermore, the training of the three-dimensional convolutional neural network model in step 2 is specifically as follows: performing data splitting on the main component three-dimensional seismic attribute body, and taking small three-dimensional data bodies of equal size as input of the three-dimensional convolutional neural network model; passing the input data sequentially through the first convolutional layer, the first downsampling layer, the activation function, the second convolutional layer, the second downsampling layer, the activation function, the third convolutional layer, the third downsampling layer, the activation function, the fourth convolutional layer, the fourth downsampling layer, and the activation function, and converting the feature data body obtained by the fourth convolutional layer into column vector data; passing the column vector data sequentially through the first fully connected layer, the second fully connected layer and the Softmax classification layer, and outputting the column vector as the probability of different sedimentary phases.

[0018] Furthermore, the training of the three-dimensional convolutional neural network in step 2 adopts a minimum batch gradient descent optimization method.

[0019] Furthermore, the step 3 is specifically as follows: performing data splitting on the main component three-dimensional seismic attribute body of the well-free area and cutting it into small three-dimensional data bodies of equal size; inputting the small three-dimensional data body into the trained three-dimensional convolutional neural network model, outputting the sedimentary phase probability as the sedimentary phase prediction result of the well-free area and constructing a three-dimensional probability data body.

[0020] Furthermore, the step 4 is specifically as follows: obtaining the variogram of the lithology of the target well area in different directions through single well interpretation of the target well area; constructing a sedimentary facies geological model through sequential Gaussian isotopic collaborative simulation based on the three-dimensional probability data body, the variogram and the sedimentary facies data.

[0021] Compared with the prior art, the present invention has the following beneficial effects: the three-dimensional seismic attribute body is screened by principal component analysis, and the three-dimensional convolutional data network is trained with the data of the well area, so as to realize the prediction of the sedimentary phase in the well-free area, and the prediction result is optimized by sequential Gaussian co-simulation to obtain the final sedimentary phase geological model, so as to realize the search for lithological reservoirs. The prediction result is constrained by multiple data, has higher accuracy, and is more in line with the real geological model. At the same time, the prediction efficiency is improved and the human interference is reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

[0023] Figure 1 This is one of the flow diagrams of the sedimentary facies prediction method based on three-dimensional seismic attribute fusion provided by an embodiment of the present invention;

[0024] Figure 2 It is a schematic diagram of the process of the principal component analysis method provided by an embodiment of the present invention;

[0025] Figure 3 is a schematic diagram of a process for constructing training data provided by an embodiment of the present invention;

[0026] Figure 4 is a schematic diagram of the structure of a three-dimensional convolutional neural network model provided by an embodiment of the present invention;

[0027] Figure 5 is a schematic diagram of the structure of a three-dimensional probability data body provided by an embodiment of the present invention;

[0028] Figure 6 It is a schematic diagram of a process for constructing a sedimentary facies geological model provided by an embodiment of the present invention;

[0029] Figure 7 This is the second flow chart of the sedimentary facies prediction method based on three-dimensional seismic attribute fusion provided by an embodiment of the present invention;

[0030] Figure 8 It is a structural schematic diagram of a sedimentary facies prediction device based on three-dimensional seismic attribute fusion provided by an embodiment of the present invention;

[0031] Fig. 9 It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0032] The present invention is described in detail below by specific examples, but the protection scope of the present invention is not limited. Unless otherwise specified, the experimental methods used in the present invention are all conventional methods, and the experimental equipment, materials, reagents, etc. used can be obtained from commercial channels.

[0033] Example 1

[0034] In the sedimentary facies distribution prediction method, the relevant technologies include:

[0035] ① Sequential Indicator Simulation Method

[0036] The sequential indicator simulation method is used to simulate the sedimentary facies, and the depth sections of different sedimentary facies development are identified in a single well, which are used as hard data for sedimentary facies simulation. Then, based on the interpretation of the single well, the development trend data of different sedimentary facies and the seismic attribute data that can reflect the development of different sedimentary facies are calibrated, which are used as constraint data for sedimentary facies simulation. Finally, the currently very mature sequential indicator simulation algorithm is used to establish a geological model of the sedimentary facies.

[0037] The various sedimentary phases simulated by this method have poor continuity, strong randomness, and chaotic geometric shapes, which are far from the actual situation.

[0038] ②Seismic sedimentary facies prediction method

[0039] The direct seismic attribute truncation method is based on the response characteristics of seismic attributes to different sedimentary phases, combined with the sedimentary phase development section of the wellbore interpreted by the single well, and under the guidance of the sedimentary model, it is manually carved to directly obtain the spatial distribution of the sedimentary phase. Specifically, it includes: first, filtering the seismic attributes, and combining the results under the guidance of the sedimentary phase development model to form a preliminary sedimentary phase model; then, combined with the single well interpretation conclusion, the local sedimentary phase is manually corrected to fill in the missing parts in the model, and the parts that do not conform to the geological laws are modified, such as the diameter of the river channel, the development position of the estuary bar, etc., to obtain the final sedimentary phase geological model.

[0040] On the one hand, this method only applies single seismic attribute data and single well interpretation conclusions, and the degree of integration of multiple types of data is not high, resulting in low accuracy of the results; on the other hand, the implementation process is too subjective and consumes a lot of manpower, resulting in low efficiency.

[0041] In response to the above-mentioned problems existing in the related technologies, an embodiment of the present invention provides a sedimentary phase prediction method and device based on three-dimensional seismic attribute fusion. The prediction result is constrained by multiple data, has higher accuracy, and is more in line with the actual geological model. At the same time, the prediction efficiency is improved and the degree of human interference is reduced.

[0042] The following is an explanation of the professional data that may appear in the present invention:

[0043] (1) Earthquake attribute body

[0044] Seismic attribute bodies refer to the physical and mathematical statistical methods that classify seismic attributes into nine categories, such as amplitude, frequency, phase, and energy, based on the kinematic and dynamic characteristics of seismic waves. Each category contains more than a dozen subcategories of attributes. The physical quantity of the statistical characteristics has a clear physical meaning and represents the attribute transformation in three-dimensional space in a classified manner.

[0045] (2) Principal Component Analysis (PCA)

[0046] Principal component analysis is a multivariate statistical method that examines the correlation between multiple variables. It studies how to reveal the internal structure of multiple variables through a few principal components, that is, to derive a few principal components from the original variables so that they retain as much information as possible from the original variables and are independent of each other. Usually, the mathematical treatment is to linearly combine the original indicators as a new comprehensive indicator.

[0047] (3) Convolutional Neural Networks (CNN)

[0048] Convolutional neural networks are a type of feedforward neural networks (FNN) that include convolutional calculations and have a deep structure. They are one of the representative algorithms of deep learning. Convolutional neural networks have representational learning capabilities and can perform translation-invariant classification of input information according to their hierarchical structure.

[0049] (4) Training data and test data during model training

[0050] Training data refers to the data used to train the data set mining model in the data mining process, from which hidden unknown valuable potential information is extracted. Test data refers to the data used to test the model and evaluate the accuracy of the model.

[0051] (5) Sequential Gaussian co-simulation

[0052] Sequential Gaussian co-simulation is a simulation method based on geostatistics that predicts unknown points according to the categories of known points. It can integrate geological laws into modeling by designing parameters such as the main variable range.

[0053] (6) 3D geological modeling

[0054] Three-dimensional geological modeling refers to the construction of three-dimensional models of various properties of geological bodies on computers, such as lithology models and porosity models.

[0055] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0056] Figure 1 is one of the flow charts of the sedimentary facies prediction method based on three-dimensional seismic attribute fusion provided by an embodiment of the present invention, such as Figure 1 As shown, the method comprises at least the following steps:

[0057] Step 101: Determine a principal component three-dimensional seismic attribute volume of a target well area based on three-dimensional seismic data of the target well area and a principal component analysis method.

[0058] Specifically, sedimentary facies include three different accuracies: sedimentary facies, sedimentary subfacies, and sedimentary microfacies. The accuracies of the three are continuously increasing, and the range is continuously narrowing. Before making a sedimentary facies prediction, it is necessary to obtain relevant data, which mainly include three-dimensional seismic data, well logging data of the wellbore, and sedimentary facies data in the present invention. Compared with data such as wellbores and reservoirs, three-dimensional seismic data has a larger and wider coverage. Three-dimensional seismic data not only includes relevant data of the well area, but also includes relevant data of the well-free area. Fully exploring the various types of information contained in the three-dimensional seismic data body is of great significance for characterizing unknown sedimentary facies and exploring lithological traps.

[0059] The massive attribute characteristics contained in seismic data are essentially probabilistic in their intrinsic response to the location of lithologic traps. The information carried by various seismic attributes (such as average amplitude, average energy, maximum curvature, root mean square amplitude, spectral energy, average instantaneous frequency, average envelope and seismic arc length) may be correlated and have similar responses to lithologic traps, that is, they contain repeated redundant information.

[0060] By using principal component analysis to remove redundant information from the original seismic attributes, a small number (e.g. 3-4) of three-dimensional data volumes are derived to represent the vast majority of valid information in the original seismic data, which are used as principal component three-dimensional seismic attribute volumes, retaining the original variable information as much as possible without being correlated with each other. This greatly reduces the workload of subsequent data calculations, while the prediction accuracy remains basically unchanged.

[0061] Step 102: construct and train a three-dimensional convolutional neural network model to learn the relationship between the wellbore data of the target well area and the principal component three-dimensional seismic attribute volume; the wellbore data includes logging data and sedimentary phase data.

[0062] Specifically, targeting the three-dimensional characteristics of seismic data, a three-dimensional convolutional neural network (CNN-3D) can effectively mine the spatial structural information in seismic data, which is beneficial to the prediction of sedimentary phases.

[0063] The wellbore data and 3D seismic attribute bodies of the target well area are used as training sets and test sets, and a 3D convolutional neural network model is constructed and trained to learn the relationship between the wellbore data and the principal component 3D seismic attribute bodies of the target well area.

[0064] Step 103: predict the sedimentary phase of the well-free area based on the trained three-dimensional convolutional data network model and construct a three-dimensional probability data volume.

[0065] Specifically, after the training of the three-dimensional convolutional neural network model is completed, the sedimentary phase of the well-free area can be predicted using the trained three-dimensional convolutional neural network model. Based on the three-dimensional seismic data of the well-free area and the principal component analysis method, the principal component three-dimensional seismic attribute body of the well-free area is determined, and after data processing, it is used as the input of the three-dimensional convolutional neural network model. The output of the model is used as the sedimentary phase prediction result of the well-free area, and the three-dimensional probability data body is constructed to obtain the probability of different types of sedimentary phases at each point in the entire geological space.

[0066] Step 104: Based on the three-dimensional probability data volume, a sedimentary facies geological model is constructed through sequential Gaussian isotopic collaborative simulation, and the sedimentary facies geological model is used to find lithological reservoirs.

[0067] Specifically, the three-dimensional probability data body constructed according to the above steps is only the result of deep mining of the original seismic data. The three-dimensional probability data body still cannot directly obtain sedimentary phase prediction conclusions with geological knowledge, and the prediction results may be biased. Therefore, the method of sequential Gaussian isotopic collaborative simulation is used to complete the final sedimentary phase prediction work.

[0068] The sedimentary facies prediction method based on three-dimensional seismic attribute fusion provided in the embodiment of the present invention screens the three-dimensional seismic attribute body through the principal component analysis method, and trains the three-dimensional convolution data network with the data of the well area, so as to realize the prediction of the sedimentary facies in the well-free area, and simultaneously optimizes the prediction results by sequential Gaussian co-simulation to obtain the final sedimentary facies geological model and realize the search for lithological reservoirs. The prediction results are constrained by multiple data, have higher accuracy, and are more in line with the real geological model. At the same time, the prediction efficiency is improved and the human interference is reduced.

[0069] Optionally, determining the principal component seismic three-dimensional seismic attribute volume of the target well area includes:

[0070] Extracting volume attributes based on the three-dimensional seismic data to obtain an original three-dimensional seismic attribute volume of the target well area;

[0071] The original three-dimensional seismic attribute volume is screened by using the principal component analysis method to obtain the principal component three-dimensional seismic attribute volume of the target well area.

[0072] Specifically, based on three-dimensional seismic data and principal component analysis, determining the principal component three-dimensional seismic attribute volume of the target well area can include: firstly extracting volume attributes of the three-dimensional seismic data to obtain the original three-dimensional seismic data volume, the original three-dimensional seismic attribute volume contains a variety of different three-dimensional seismic attribute volumes, the information carried by each type of seismic attribute may be correlated, and the response to lithologic traps is similar; then using the principal component analysis method to screen the original three-dimensional seismic attribute volume, eliminate redundant information, and derive a few principal component three-dimensional seismic attribute volumes.

[0073] Figure 2 is a schematic diagram of the process of the principal component analysis method provided by an embodiment of the present invention, such as Figure 2 As shown, three principal component three-dimensional seismic attribute volumes are finally derived.

[0074] The sedimentary facies prediction method based on three-dimensional seismic attribute fusion provided by the embodiment of the present invention screens the three-dimensional seismic attribute body through the principal component analysis method, thereby improving the accuracy of the prediction result.

[0075] Optionally, the training of the three-dimensional convolutional neural network model includes:

[0076] Performing data splitting on the main component three-dimensional seismic attribute volume, and using small three-dimensional data volumes of the same size as inputs of the three-dimensional convolutional neural network model;

[0077] The input data is sequentially passed through the first convolution layer, the first downsampling layer, the activation function, the second convolution layer, the second downsampling layer, the activation function, the third convolution layer, the third downsampling layer, the activation function, the fourth convolution layer, the fourth downsampling layer, and the activation function, and the feature data volume obtained by the fourth convolution layer is converted into column vector data;

[0078] The column vector data is sequentially passed through the first fully connected layer, the second fully connected layer and the Softmax classification layer, and the column vector is output as the probability of different sedimentary phases.

[0079] Specifically, the training of the 3D convolutional neural network is based on the 3D seismic data and wellbore data of the well area.

[0080] Optionally, the method further comprises:

[0081] Based on the wellbore data, the wellbore spatial position positioning and sedimentary phase digitization processing are performed.

[0082] Specifically, the well logging data and sedimentary phase data of the wellbore are used to locate the three-dimensional spatial position of the wellbore and to digitize the sedimentary phase.

[0083] Figure 3 is a flow chart of training data construction provided by an embodiment of the present invention, such as Figure 3 As shown in the figure, the wellbore interpretation sedimentary facies is defined as a matrix form. It is assumed that there are three sedimentary facies in a certain area: distributary channel, estuary bar and inter-distributary channel. For a certain point on the wellbore, when the distributary channel facies is developed at this point, its probability is 1, and the probability of the other two microfacies is 0; the wellbore sedimentary facies are digitized in turn, which corresponds to the output data of the three-dimensional convolutional network during the training process.

[0084] For each point on the wellbore, a 16×16×16 grid around the point is intercepted from the three data volumes output by PCA to obtain three 16×16×16 small data volumes, which are the input data of the three-dimensional convolutional neural network during training.

[0085] In this way, the 3D convolutional network can establish the correspondence between the 3D data volume and the sedimentary phase. All data are prepared as input data-output data according to the above method, and these data pairs are divided into training data sets and test data sets. The former is used to train the 3D convolutional neural network model, and the latter is used to test the prediction effect of the 3D convolutional neural network model.

[0086] Figure 4 is a schematic diagram of the structure of a three-dimensional convolutional neural network model provided by an embodiment of the present invention, such as Figure 4 As shown, an embodiment of the present invention provides a three-dimensional convolutional neural network model, including four convolutional layers and downsampling layers, and two fully connected layers.

[0087] First, the three three-dimensional data volumes obtained by CPA are cut into several small data volumes of 16×16×16 as the input layer; then, four convolutional layers are processed (the convolutional layer is used to extract the characteristics of a certain aspect of the input data, and multiple convolutional layers can extract the characteristics of different levels and aspects of the input data) and the downsampling layer is processed simultaneously to extract the spatially relevant effective information in the input data; then, the feature data volume obtained by the last convolutional layer is converted into column vector data; finally, after calculation by two fully connected neural network layers and one SoftMax layer, a 3×1 column vector is output, representing the probability of different sedimentary phases.

[0088] The training data and the test data are input into the 3D convolutional neural network model to start training, and finally the parameters of the 3D convolutional neural network model can be obtained. After the training is completed, the test data is used to test the accuracy of the model.

[0089] Optionally, the three-dimensional convolutional neural network is trained using a minimum batch gradient descent optimization method.

[0090] Specifically, the minimum batch gradient descent (MBGD) and Adam optimization algorithms are adopted in the training process of the three-dimensional convolutional neural network in the present invention.

[0091] The sedimentary phase prediction method based on three-dimensional seismic attribute fusion provided in the embodiment of the present invention designs a three-dimensional convolutional neural network model and uses the seismic data and wellbore data of the well area for training, digitizes the wellbore sedimentary phase, and grids the seismic data at the edge of the wellbore sedimentary phase interpretation point, so that the sedimentary phase prediction is more accurate and more in line with the actual geological model, while the prediction efficiency is improved and the human interference is reduced.

[0092] Optionally, the three-dimensional convolutional data network model based on the training is used to predict the sedimentary phase of the well-free area and construct a three-dimensional probability data volume, including:

[0093] Splitting the main component three-dimensional seismic attribute volume of the well-free area into small three-dimensional data volumes of the same size;

[0094] The small three-dimensional data volume is input into the trained three-dimensional convolutional neural network model, and the sedimentary phase probability is output as the sedimentary phase prediction result of the well-free area and a three-dimensional probability data volume is constructed.

[0095] Specifically, after completing the training of the three-dimensional convolutional neural network model using the three-dimensional seismic data and wellbore data of the well area, the sedimentary phase of the well-free area can be predicted using the trained three-dimensional convolutional neural network model.

[0096] Similarly, the volume attributes of the three-dimensional seismic data in the well-free area are extracted, and the principal component analysis method is used for data screening to obtain the principal component three-dimensional seismic attribute volume of the well-free area and perform data differentiation, which is then cut into small three-dimensional data volumes of the same size and used as the input of the three-dimensional convolutional neural network model during prediction. The sedimentary phase probability is output, that is, the probability of different types of sedimentary phases at the center of the small three-dimensional data volume.

[0097] By integrating the sedimentary facies data of the well area and the sedimentary facies prediction results of the well-free area, the probability of different types of sedimentary facies at each point in the entire geological space can be obtained.

[0098] Figure 5 is a schematic diagram of the structure of a three-dimensional probability data body provided by an embodiment of the present invention, such as Figure 5 As shown, the three-dimensional space where the geological space is located is divided into multiple small three-dimensional spaces of the same size. The solid small cubes represent the small three-dimensional data bodies cut out from the well area, and the dotted parts are multiple small three-dimensional data bodies cut out from the non-well area. The probability of different types of sedimentary phases at the center of each small three-dimensional data body is represented as a vector.

[0099] The sedimentary phase prediction method based on three-dimensional seismic attribute fusion provided by the embodiment of the present invention uses the three-dimensional seismic data and wellbore data of the well area to train a three-dimensional convolutional neural network model, and then uses the model to predict the sedimentary phase of the well-free area and construct a three-dimensional probability data body to complete the modeling of the sedimentary phase of the entire geological space. The prediction result is constrained by multiple data, has higher accuracy, and is more in line with the actual geological model. At the same time, the prediction efficiency is improved and the degree of human interference is reduced.

[0100] Optionally, constructing a sedimentary facies geological model based on the three-dimensional probability data volume through sequential Gaussian isotopic collaborative simulation includes:

[0101] Obtaining the variogram of the lithofacies in different directions of the target well area through single well interpretation of the target well area;

[0102] Based on the three-dimensional probability data volume, the variation function and the sedimentary phase data, a sedimentary phase geological model is constructed through sequential Gaussian isotopic collaborative simulation.

[0103] Specifically, based on the well logging and sedimentary phase data of the target well area, single well interpretation is performed for different sedimentary phases, the variation function of the lithofacies in the target well area in different directions is obtained, and the three-dimensional change law of the target attributes is quantitatively characterized.

[0104] Among them, the variogram is used to describe the spatial variation of reservoir properties and can quantitatively describe the spatial correlation of regionalized variables. This spatial correlation is anisotropic, so it is necessary to describe the variogram of the lithology of the target well area from different directions. In the present invention, the variogram is mainly used to reflect the length, width and directionality of the river channel.

[0105] According to the three-dimensional probability data volume of sedimentary facies, variogram and sedimentary facies data of wellbore, a sedimentary facies geological model is constructed through sequential Gaussian isotopic collaborative simulation.

[0106] Figure 6 is a schematic diagram of a process for constructing a sedimentary facies geological model provided by an embodiment of the present invention, such as Figure 6 As shown, based on the three-dimensional probability data volume of the sedimentary facies, the variation function and the sedimentary facies data of the wellbore, the construction of the sedimentary facies geological model is realized through sequential Gaussian isotopic collaborative simulation.

[0107] The three-dimensional probability data body constructed by the output results of the three-dimensional convolutional neural network model is only the conclusion after deep mining of the original seismic data. The three-dimensional probability data body still cannot directly obtain the sedimentary phase prediction conclusion with geological knowledge, and the prediction result may be biased.

[0108] Therefore, the present invention adopts the method of sequential Gaussian co-simulation to integrate the sedimentary facies types of the study area, integrates the research results such as provenance direction and river channel aspect ratio into the variation function, uses the probability data body after deep learning as the inter-well constraint data, and the wellbore interpretation sedimentary facies as the hard data point to complete the final sedimentary facies prediction work.

[0109] The sedimentary facies prediction method based on three-dimensional seismic attribute fusion provided in an embodiment of the present invention obtains a principal component three-dimensional seismic attribute body by performing attribute analysis on the three-dimensional seismic data of the target well area, constructs a three-dimensional convolutional neural network model, and completes model training through the principal component three-dimensional seismic attribute body of the wellbore sedimentary facies and the wellbore periphery, so that the trained three-dimensional convolutional neural network can be used to analyze the probability of sedimentary facies development under the current actual geological conditions, thereby providing a theoretical basis for geological research.

[0110] The technical solution provided by the present invention is described below with a specific example.

[0111] Figure 7 FIG. 2 is a flow chart of a sedimentary facies prediction method based on three-dimensional seismic attribute fusion provided by an embodiment of the present invention. Figure 7 As shown in the figure, this method deeply mines the information contained in seismic data, adopts principal component analysis to comprehensively consider multiple types of seismic attributes, and uses a three-dimensional convolutional neural network to learn the relationship between the sedimentary facies at the existing wellbore and the surrounding seismic attributes, so as to achieve accurate prediction of the sedimentary facies probability body in the well-free area. Then, the mature sequential Gaussian co-simulation is applied to calculate the three-dimensional distribution of sedimentary facies.

[0112] Based on the sedimentary facies prediction method based on 3D seismic attribute fusion provided by the embodiment of the present invention, sedimentary facies prediction work was carried out using an oil field in a certain area. The oil field has developed lithologic reservoirs, and the oil field has developed delta front sedimentary subfacies, mainly composed of distributary channels and inter-distributary channel sedimentary facies. Based on the sedimentary facies prediction method based on 3D seismic attribute fusion provided by the embodiment of the present invention, the lithologic trap exploration problem is implemented on the plane distribution for display by using the layered slicing method along the wellbore.

[0113] First, use Figure 4 The three-dimensional convolutional neural network model shown sets a sliding window triggered by the wellbore, uses the main component three-dimensional seismic attribute body of the manually interpreted well area and the corresponding well area periphery from top to bottom as the original training data, and retains 10% of the data as test data.

[0114] Then, the trained 3D convolutional neural network model is used to take the principal component 3D seismic attribute volume of the well-free area as input to obtain the 3D probability data volume of the development of different sedimentary phases under the constraints of seismic attributes.

[0115] Finally, the sedimentary phase data of the wellbore is used as the hard data for modeling, the constructed three-dimensional probability data body is used as the constraint data, and the sequential Gaussian isotopic collaborative simulation method is used to simulate the three-dimensional spatial distribution of the sedimentary phase in the study area, combined with the distribution direction and length and width characteristics of the rivers in the study area.

[0116] The sedimentary facies prediction method based on three-dimensional seismic attribute fusion provided in the embodiment of the present invention makes full use of seismic attributes and geological knowledge, and the predicted sedimentary facies distribution is more consistent with the actual situation in the study area.

[0117] The sedimentary phase prediction device based on three-dimensional seismic attribute fusion provided by the present invention is described below. The sedimentary phase prediction device based on three-dimensional seismic attribute fusion described below and the sedimentary phase prediction method based on three-dimensional seismic attribute fusion described above can be referenced to each other.

[0118] Figure 8 is a schematic diagram of the structure of a sedimentary facies prediction device based on three-dimensional seismic attribute fusion provided by an embodiment of the present invention, such as Figure 8 As shown, the device may include:

[0119] A determination module 801 is used to determine a principal component three-dimensional seismic attribute volume of a target well area based on the three-dimensional seismic data of the target well area and a principal component analysis method;

[0120] A training module 802 is used to construct and train a three-dimensional convolutional neural network model to learn the relationship between the wellbore data of the target well area and the principal component three-dimensional seismic attribute volume; the wellbore data includes logging data and sedimentary phase data;

[0121] A prediction module 803 is used to predict the sedimentary phase of the well-free area based on the trained three-dimensional convolutional data network model and construct a three-dimensional probability data volume;

[0122] The construction module 804 constructs a sedimentary facies geological model based on the three-dimensional probability data volume through sequential Gaussian isotopic collaborative simulation, and the sedimentary facies geological model is used to find lithological reservoirs.

[0123] Optionally, the determining module is further used for:

[0124] Extracting volume attributes based on the three-dimensional seismic data to obtain an original three-dimensional seismic attribute volume of the target well area;

[0125] The original three-dimensional seismic attribute volume is screened by using the principal component analysis method to obtain the principal component three-dimensional seismic attribute volume of the target well area.

[0126] Optionally, the training module is further used for:

[0127] Performing data splitting on the main component three-dimensional seismic attribute volume, and using small three-dimensional data volumes of the same size as inputs of the three-dimensional convolutional neural network model;

[0128] The input data is sequentially passed through the first convolution layer, the first downsampling layer, the activation function, the second convolution layer, the second downsampling layer, the activation function, the third convolution layer, the third downsampling layer, the activation function, the fourth convolution layer, the fourth downsampling layer, and the activation function, and the feature data volume obtained by the fourth convolution layer is converted into column vector data;

[0129] The column vector data is sequentially passed through the first fully connected layer, the second fully connected layer and the Softmax classification layer, and the column vector is output as the probability of different sedimentary phases.

[0130] Optionally, the device further comprises:

[0131] The processing module is used to perform wellbore spatial location positioning and sedimentary phase digital processing based on the wellbore data.

[0132] Optionally, the three-dimensional convolutional neural network is trained using a minimum batch gradient descent optimization method.

[0133] Optionally, the prediction module is further used for:

[0134] Splitting the main component three-dimensional seismic attribute volume of the well-free area into small three-dimensional data volumes of the same size;

[0135] The small three-dimensional data volume is input into the trained three-dimensional convolutional neural network model, and the sedimentary phase probability is output as the sedimentary phase prediction result of the well-free area and a three-dimensional probability data volume is constructed.

[0136] Optionally, the building block is further used to:

[0137] Obtaining the variogram of the lithofacies in different directions of the target well area through single well interpretation of the target well area;

[0138] Based on the three-dimensional probability data volume, the variation function and the sedimentary phase data, a sedimentary phase geological model is constructed through sequential Gaussian isotopic collaborative simulation.

[0139] Fig. 9 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention, such as Fig. 9 As shown, the electronic device may include: a processor (processor) 901, a communication interface (CommunicationsInterface) 902, a memory (memory) 903 and a communication bus 904, wherein the processor 901, the communication interface 902, and the memory 903 communicate with each other through the communication bus 904.

[0140] In addition, the logic instructions in the above-mentioned memory 903 can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the relevant technology or the part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0141] The processor 901 may call the logic instructions in the memory 903 to execute a sedimentary facies prediction method based on three-dimensional seismic attribute fusion, the method comprising:

[0142] Determine the principal component three-dimensional seismic attribute volume of the target well area based on the three-dimensional seismic data of the target well area and the principal component analysis method;

[0143] Constructing and training a three-dimensional convolutional neural network model to learn the relationship between the wellbore data of the target well area and the principal component three-dimensional seismic attribute volume; the wellbore data includes logging data and sedimentary phase data;

[0144] Predicting the sedimentary phase of the well-free area based on the trained three-dimensional convolutional data network model and constructing a three-dimensional probability data volume;

[0145] Based on the three-dimensional probabilistic data volume, a sedimentary facies geological model is constructed through sequential Gaussian isotopic collaborative simulation, and the sedimentary facies geological model is used to find lithological reservoirs.

[0146] Optionally, determining the principal component seismic three-dimensional seismic attribute volume of the target well area includes:

[0147] Extracting volume attributes based on the three-dimensional seismic data to obtain an original three-dimensional seismic attribute volume of the target well area;

[0148] The original three-dimensional seismic attribute volume is screened by using the principal component analysis method to obtain the principal component three-dimensional seismic attribute volume of the target well area.

[0149] Optionally, the training of the three-dimensional convolutional neural network model includes:

[0150] Performing data splitting on the main component three-dimensional seismic attribute volume, and using small three-dimensional data volumes of the same size as inputs of the three-dimensional convolutional neural network model;

[0151] The input data is sequentially passed through the first convolution layer, the first downsampling layer, the activation function, the second convolution layer, the second downsampling layer, the activation function, the third convolution layer, the third downsampling layer, the activation function, the fourth convolution layer, the fourth downsampling layer, and the activation function, and the feature data volume obtained by the fourth convolution layer is converted into column vector data;

[0152] The column vector data is sequentially passed through the first fully connected layer, the second fully connected layer and the Softmax classification layer, and the column vector is output as the probability of different sedimentary phases.

[0153] Optionally, the method further comprises:

[0154] Based on the wellbore data, the wellbore spatial position positioning and sedimentary phase digitization processing are performed.

[0155] Optionally, the three-dimensional convolutional neural network is trained using a minimum batch gradient descent optimization method.

[0156] Optionally, the three-dimensional convolutional data network model based on the training is used to predict the sedimentary phase of the well-free area and construct a three-dimensional probability data volume, including:

[0157] Splitting the main component three-dimensional seismic attribute volume of the well-free area into small three-dimensional data volumes of the same size;

[0158] The small three-dimensional data volume is input into the trained three-dimensional convolutional neural network model, and the sedimentary phase probability is output as the sedimentary phase prediction result of the well-free area and a three-dimensional probability data volume is constructed.

[0159] Optionally, constructing a sedimentary facies geological model based on the three-dimensional probability data volume through sequential Gaussian isotopic collaborative simulation includes:

[0160] Obtaining the variogram of the lithofacies in different directions of the target well area through single well interpretation of the target well area;

[0161] Based on the three-dimensional probability data volume, the variation function and the sedimentary phase data, a sedimentary phase geological model is constructed through sequential Gaussian isotopic collaborative simulation.

[0162] On the other hand, the present invention further provides a computer program product, the computer program product includes a computer program, the computer program can be stored in a non-transitory computer-readable storage medium, when the computer program is executed by a processor, the computer can execute the sedimentary facies prediction method based on three-dimensional seismic attribute fusion provided by the above methods, the method may include:

[0163] Determine the principal component three-dimensional seismic attribute volume of the target well area based on the three-dimensional seismic data of the target well area and the principal component analysis method;

[0164] Constructing and training a three-dimensional convolutional neural network model to learn the relationship between the wellbore data of the target well area and the principal component three-dimensional seismic attribute volume; the wellbore data includes logging data and sedimentary phase data;

[0165] Predicting the sedimentary phase of the well-free area based on the trained three-dimensional convolutional data network model and constructing a three-dimensional probability data volume;

[0166] Based on the three-dimensional probabilistic data volume, a sedimentary facies geological model is constructed through sequential Gaussian isotopic collaborative simulation, and the sedimentary facies geological model is used to find lithological reservoirs.

[0167] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is implemented when the computer program is executed by a processor to execute the sedimentary facies prediction method based on three-dimensional seismic attribute fusion provided by the above methods, and the method may include:

[0168] Determine the principal component three-dimensional seismic attribute volume of the target well area based on the three-dimensional seismic data of the target well area and the principal component analysis method;

[0169] Constructing and training a three-dimensional convolutional neural network model to learn the relationship between the wellbore data of the target well area and the principal component three-dimensional seismic attribute volume; the wellbore data includes logging data and sedimentary phase data;

[0170] Predicting the sedimentary phase of the well-free area based on the trained three-dimensional convolutional data network model and constructing a three-dimensional probability data volume;

[0171] Based on the three-dimensional probabilistic data volume, a sedimentary facies geological model is constructed through sequential Gaussian isotopic collaborative simulation, and the sedimentary facies geological model is used to find lithological reservoirs.

[0172] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0173] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, by hardware. Based on this understanding, the above technical solution can be essentially or in other words, the part that contributes to the relevant technology can be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiment.

[0174] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A sedimentary facies prediction device based on 3D seismic attribute fusion, characterized in that: It includes determination module, training module, prediction module and construction module.

2. The prediction method of the sedimentary facies prediction device based on the three-dimensional seismic attribute fusion according to claim 1 is characterized in that: The steps are: S1. Determine the principal component three-dimensional seismic attribute volume of the target well area based on the three-dimensional seismic data of the target well area and the principal component analysis method; S2. Construct and train a three-dimensional convolutional neural network model to learn the relationship between the wellbore data of the target well area and the main component three-dimensional seismic attribute body; the wellbore data includes logging data and sedimentary phase data; S3. Predicting the sedimentary phase of the well-free area based on the trained three-dimensional convolutional data network model and constructing a three-dimensional probability data volume; S4. Based on the three-dimensional probability data body, a sedimentary facies geological model is constructed through sequential Gaussian isotopic collaborative simulation, and the sedimentary facies geological model is used to find lithological reservoirs.

3. The prediction method of the sedimentary facies prediction device based on 3D seismic attribute fusion according to claim 2, characterized in that: The step S1 specifically comprises: extracting volume attributes based on the 3D seismic data to obtain the original 3D seismic attribute volume of the target well area; and screening the original 3D seismic attribute volume using principal component analysis to obtain the principal component 3D seismic attribute volume of the target well area.

4. The prediction method of the sedimentary facies prediction device based on 3D seismic attribute fusion according to claim 2, characterized in that: The training of the three-dimensional convolutional neural network model in step S2 is specifically as follows: data splitting is performed on the main component three-dimensional seismic attribute volume, and small three-dimensional data volumes of equal size are used as inputs of the three-dimensional convolutional neural network model; the input data is sequentially passed through the first convolutional layer, the first downsampling layer, the activation function, the second convolutional layer, the second downsampling layer, the activation function, the third convolutional layer, the third downsampling layer, the activation function, the fourth convolutional layer, the fourth downsampling layer, and the activation function, and the feature data volume obtained by the fourth convolutional layer is converted into column vector data; the column vector data is sequentially passed through the first fully connected layer, the second fully connected layer and the Softmax classification layer, and the column vector is output as the probability of different sedimentary phases.

5. The prediction method of the sedimentary facies prediction device based on 3D seismic attribute fusion according to claim 4, characterized in that: The training of the three-dimensional convolutional neural network in step S2 adopts the minimum batch gradient descent optimization method.

6. The prediction method of the sedimentary facies prediction device based on 3D seismic attribute fusion according to claim 2, characterized in that: The step S3 specifically comprises: performing data splitting on the main component three-dimensional seismic attribute volume of the well-free area and cutting it into small three-dimensional data volumes of equal size; inputting the small three-dimensional data volume into the trained three-dimensional convolutional neural network model, outputting the sedimentary phase probability as the sedimentary phase prediction result of the well-free area and constructing a three-dimensional probability data volume.

7. The prediction method of the sedimentary facies prediction device based on 3D seismic attribute fusion according to claim 2, characterized in that: The step S4 is specifically as follows: obtaining the variogram of the lithofacies in the target well area in different directions through single well interpretation of the target well area; and constructing a sedimentary facies geological model based on the three-dimensional probability data volume, the variogram and the sedimentary facies data.

8. The prediction method of the sedimentary facies prediction device based on 3D seismic attribute fusion according to claim 7, characterized in that: The sedimentary facies geological model was constructed by sequential Gaussian isotopic co-simulation.

Images