Well-seismic combined reservoir fracture prediction method based on artificial neural network

Abstract

The invention relates to a well-seismic combined reservoir fracture prediction method based on an artificial neural network, and the method comprises the steps: determining a plurality of fracture parameters representing the fracture development degree based on the obtained logging information of a research region; determining a comprehensive crack prediction index based on the plurality of crack parameters; seismic attributes are extracted by using the acquired seismic data, and target seismic attributes sensitive to the change of the comprehensive fracture prediction index are optimized; based on the seismic attribute data set corresponding to the target seismic attribute, training a pre-established artificial neural network to obtain a comprehensive fracture prediction model; and reservoir fracture prediction is carried out based on the comprehensive fracture prediction model, and a fracture prediction result is obtained. According to the method, fracture information reflected by the seismic data and the logging data is fused through machine learning, so that more comprehensive underground fracture development information is obtained. According to the method, the precision of the logging data and the breadth of the seismic data are integrated, fracture prediction is carried out on the spatial scale, and reservoir fractures can be effectively identified and predicted.

Description

Technical Field

[0001] The invention relates to a method for predicting reservoir fractures based on well-seismic combination with artificial neural network. Background Art

[0002] Fractured reservoirs are an important type of oil and gas reservoirs. The fracture development mechanism is complex and the heterogeneity is strong. Fracture development seriously restricts the development and accumulation of oil and gas. Predicting the development of reservoir fractures is of great significance to oil and gas exploration and development. Existing methods for identifying reservoir fractures are mainly through core observation, well logging and seismic methods. Summary of the invention

[0003] In order to enrich the reservoir fracture prediction methods in the prior art and realize the effective identification and prediction of reservoir fractures, the present invention proposes a reservoir fracture prediction method based on well seismic combined with artificial neural network. The technical solution proposed by the present invention is as follows:

[0004] In a first aspect, the present invention provides a reservoir fracture prediction method based on well seismic data combined with artificial neural network, comprising:

[0005] Obtain seismic attribute data of the target area, perform reservoir fracture prediction based on the comprehensive fracture prediction model, and obtain fracture prediction results;

[0006] The comprehensive crack prediction model is trained in the following way:

[0007] Based on the well logging data obtained in the study area, multiple fracture parameters characterizing the degree of fracture development are determined;

[0008] Determining a comprehensive fracture prediction index based on the plurality of fracture parameters;

[0009] Extracting seismic attributes using the acquired seismic data, and selecting target seismic attributes that are sensitive to changes in the comprehensive fracture prediction index;

[0010] Based on the seismic attribute data set corresponding to the target seismic attribute, a pre-established artificial neural network is trained to establish a functional relationship between a comprehensive fracture prediction index and the target seismic attribute to obtain a comprehensive fracture prediction model.

[0011] In one or some embodiments, the seismic data includes a post-stack seismic data volume;

[0012] The method of extracting seismic attributes using the acquired seismic data and selecting target seismic attributes that are sensitive to changes in the comprehensive fracture prediction index includes:

[0013] Acquire a post-stack seismic data volume, and extract multiple seismic attributes of a target layer from the post-stack seismic data volume; wherein the multiple seismic attributes include instantaneous frequency attributes, instantaneous phase attributes, root mean square amplitude attributes, maximum amplitude attributes, curvature attributes, relative wave impedance attributes, coherence volume attributes, and multi-scale edge detection attributes;

[0014] The correlation between each seismic attribute and the comprehensive fracture prediction index is calculated respectively, and a target seismic attribute having a correlation greater than a preset correlation threshold is screened out from the multiple seismic attributes.

[0015] In one or some embodiments, the method of training a pre-established artificial neural network based on a seismic attribute data set corresponding to the target seismic attribute, establishing a functional relationship between a comprehensive fracture prediction index and the target seismic attribute, and obtaining a comprehensive fracture prediction model includes:

[0016] Standardizing and normalizing the seismic attribute data set corresponding to the target seismic attribute to obtain a processed seismic attribute data set;

[0017] Dividing the processed earthquake attribute data set into a training set and a test validation set according to a preset ratio;

[0018] Using the training set as input data of the artificial neural network, optimizing model parameters through a back propagation algorithm, and obtaining a trained model;

[0019] The trained model is verified based on the obtained fracture observation results, core data, drilling data and oil test data of the study area, and the verification set, and the training process is repeated until the iteration stop condition is reached to obtain the comprehensive fracture prediction model.

[0020] In one or some embodiments, the trained model is verified by:

[0021] Inputting the validation set into the trained model to obtain a model prediction result;

[0022] Obtain fracture observation results, core data, drilling data and oil test data in the study area, and align the fracture observation results, the core data, the drilling data and the oil test data according to the geographical location and depth position of the well;

[0023] The model prediction results are verified based on the fracture observation results, core data, drilling data and oil test data to determine whether the error rate of the trained model is lower than a preset error threshold.

[0024] In one or some embodiments, the model prediction results are verified by:

[0025] Comparing the crack observation results with the model prediction results to determine whether the trained model can accurately predict cracks at corresponding locations;

[0026] Comparing the core data and drilling data with the model prediction results to determine whether the fracture locations predicted by the trained model are consistent with the actual fracture distribution;

[0027] The oil test data is compared with the model prediction results to determine whether there are underground features in the model prediction results that are consistent with the degree of fracture development and the oil test production.

[0028] In one or some embodiments, the crack parameter includes a change rate of the acoustic wave curve, and the change rate of the acoustic wave curve is calculated by the following formula:

[0029]

[0030] In the formula, Δa i is the rate of change of the sound wave curve, a i is the sampling point of the sound wave curve.

[0031] In one or some embodiments, the fracture parameters include a formation factor ratio; the formation factor ratio is calculated by the following formula:

[0032]

[0033]

[0034]

[0035] Where b is the formation factor ratio, b 1 is the formation factor calculated from resistivity logging, b 2 is the formation factor calculated from the porosity, R 0 is the deep lateral resistivity, R W is the formation water resistivity, a is the lithology coefficient, φ is the porosity, and m is the rock pore structure index.

[0036] In one or some embodiments, the fracture parameter includes a resistivity correction difference ratio; the resistivity correction difference ratio is calculated by the following formula:

[0037]

[0038] Where c is the resistivity correction difference ratio, R is the true resistivity of the formation, and RLLS is the shallow lateral resistivity.

[0039] In one or some embodiments, the fracture parameter includes a saturation ratio; the saturation ratio is calculated by the following formula:

[0040]

[0041] In the formula, d is the saturation ratio, R W is the formation water resistivity, R XO is the resistivity of the flushing zone, R mf is the mud resistivity, R is the true resistivity of the formation, and m is the rock pore structure index.

[0042] In one or some embodiments, before determining the comprehensive fracture prediction index based on the multiple fracture parameters, the method further includes:

[0043] The fracture parameters are normalized by the following formula to obtain the normalized fracture parameters:

[0044]

[0045] In the formula, x is the crack parameter after normalization, y is the original crack parameter, and λ is the transformation parameter;

[0046] The normalized crack parameters are standardized by the following formula to obtain the standardized crack parameters:

[0047]

[0048] In the formula, is the crack parameter after normalization, x max and x min are the maximum and minimum values ​​of the crack parameters after normalization.

[0049] In one or some embodiments, determining a comprehensive fracture prediction index based on the multiple fracture parameters includes:

[0050] Based on the standardized crack parameters, the comprehensive crack prediction index is calculated by the following formula:

[0051]

[0052] Where f is the comprehensive crack prediction index, is the crack parameter after normalization, k i is the parameter weight;

[0053] in,

[0054]

[0055] Where ti is the fracture parameter statistically obtained from the coring section of the well, which reflects the thickness of the fracture development, and T is the total thickness of the coring section.

[0056] In a second aspect, the present invention provides a comprehensive crack prediction model training method, comprising:

[0057] Based on the well logging data obtained in the study area, multiple fracture parameters characterizing the degree of fracture development are determined;

[0058] Determining a comprehensive fracture prediction index based on the plurality of fracture parameters;

[0059] Extracting seismic attributes using the acquired seismic data, and selecting target seismic attributes that are sensitive to changes in the comprehensive fracture prediction index;

[0060] Based on the seismic attribute data set corresponding to the target seismic attribute, a pre-established artificial neural network is trained to establish a functional relationship between a comprehensive fracture prediction index and the target seismic attribute to obtain a comprehensive fracture prediction model.

[0061] In a third aspect, the present invention provides a reservoir fracture prediction device based on well seismic data and artificial neural network, comprising:

[0062] The first calculation module is used to determine a plurality of fracture parameters characterizing the degree of fracture development based on the acquired well logging data of the study area;

[0063] A second calculation module, configured to determine a comprehensive fracture prediction index based on the plurality of fracture parameters;

[0064] A screening module, used to extract seismic attributes using the acquired seismic data, and select target seismic attributes that are sensitive to changes in the comprehensive fracture prediction index;

[0065] A training module, for establishing a functional relationship between the comprehensive fracture prediction index and the target seismic attribute based on a pre-established artificial neural network, and obtaining a comprehensive fracture prediction model;

[0066] The prediction module is used to predict reservoir fractures based on the comprehensive fracture prediction model to obtain fracture prediction results.

[0067] In a fourth aspect, the present invention provides a comprehensive crack prediction model training device, comprising:

[0068] The first calculation module is used to determine a plurality of fracture parameters characterizing the degree of fracture development based on the acquired well logging data of the study area;

[0069] A second calculation module, configured to determine a comprehensive fracture prediction index based on the plurality of fracture parameters;

[0070] A screening module, used to extract seismic attributes using the acquired seismic data, and select target seismic attributes that are sensitive to changes in the comprehensive fracture prediction index;

[0071] The training module is used to train a pre-established artificial neural network based on a seismic attribute data set corresponding to the target seismic attribute, establish a functional relationship between a comprehensive fracture prediction index and the target seismic attribute, and obtain a comprehensive fracture prediction model.

[0072] In a fifth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for predicting reservoir fractures by combining wellbore seismic data with an artificial neural network as described in the first aspect, or the method for training a comprehensive fracture prediction model as described in the second aspect.

[0073] In a sixth aspect, the present invention provides an electronic device, comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus;

[0074] Memory, used to store computer programs;

[0075] The processor is used to implement the well-seismic combined reservoir fracture prediction method based on artificial neural network as described in the first aspect, or the comprehensive fracture prediction model training method as described in the second aspect when executing the program stored in the memory.

[0076] Based on the above technical solution, the beneficial effects of the present invention compared with the prior art are as follows:

[0077] The well-seismic combined reservoir fracture prediction method based on artificial neural network provided by the present invention determines multiple fracture parameters characterizing the degree of fracture development based on the well logging data of the study area obtained; determines the comprehensive fracture prediction index based on the determined multiple fracture parameters; extracts seismic attributes using seismic data, and selects the seismic attributes that are more sensitive to the changes of the above-mentioned comprehensive fracture prediction index through qualitative and quantitative comparative analysis to obtain the target seismic attributes; then, based on the pre-established artificial neural network, establishes a functional relationship between the comprehensive fracture prediction index and the target seismic attribute to obtain a comprehensive fracture prediction model, and obtains the comprehensive fracture prediction model through multiple seismic attributes that are more sensitive to the changes of the comprehensive fracture prediction index through seismic multi-attribute inversion, thereby realizing the seismic multi-attribute inversion of reservoir fracture development zones. The present invention fuses the fracture information reflected by seismic data and well logging data through machine learning, thereby obtaining more comprehensive underground fracture development information. The present invention combines the accuracy of well logging data and the breadth of seismic data to predict fractures on a spatial scale, and well predicts the development of fracture zones in actual block applications.

[0078] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0079] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0081] Figure 1 It is a flow chart of a method for predicting reservoir fractures based on well seismic combined with artificial neural network provided by an embodiment of the present invention;

[0082] Figure 2 is another schematic flow chart of a method for predicting reservoir fractures based on well seismic and artificial neural network provided by an embodiment of the present invention;

[0083] Figure 3 This is a prediction result diagram of an application example of the reservoir fracture prediction method based on well seismic combined with artificial neural network provided in an embodiment of the present invention;

[0084] Figure 4 It is a schematic diagram of the structure of a well-seismic combined reservoir fracture prediction device based on an artificial neural network provided by an embodiment of the present invention;

[0085] Figure 5 It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0086] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0087] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0088] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0089] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0090] The inventors found in their work that the existing means of identifying reservoir fractures are mainly through core observation, well logging and seismic methods. Field, core, microscopic observation and other methods can visually observe the development of fractures, but the samples are limited and it is difficult to grasp the overall fracture development; the well logging method can identify vertical reservoir fractures through the changes in the logging curve, and this method is limited to the development characteristics of vertical fractures. Seismic fracture identification methods mainly include two types of methods: post-stack longitudinal wave seismic attribute method and pre-stack longitudinal wave azimuthal anisotropy, shear wave prediction, conversion wave and multi-wave exploration. The latter is limited by the purchase cost and has a low actual application value. The post-stack method includes coherence, dip azimuth, absorption attenuation coefficient, etc. The seismic data prediction range is large, but the accuracy is not enough and cannot meet the inventor's expectations. The inventor made the present invention after further research and development.

[0091] Embodiment 1

[0092] Conventional logging data is limited by the number of wells drilled and has limited spatial resolution. Conventional logging data is point measurement data, which is not applicable to the prediction of fracture development in a large range. It is necessary to combine other technologies to fill the blank areas. The existing technology is limited to a single logging method or seismic method, and fails to explore the correlation between well data and seismic data that can reflect fracture development; and the existing reservoir fracture prediction technology fails to or rarely involves artificial intelligence learning; in addition, the existing reservoir fracture prediction technology fails to or rarely establishes a fracture development model on a spatial scale. Limited by the accuracy of seismic data, the seismic method can only roughly predict the location of fracture development. The combination of seismic and logging methods can overcome some defects of conventional logging methods in predicting fracture development characteristics to a certain extent. Since the logging curve has a response characteristic to reservoir fractures and a high vertical resolution, the present invention combines logging data and seismic data to predict fracture development zones. Therefore, the present invention is prepared to calculate a comprehensive fracture prediction index through logging curves, extract seismic attributes that are sensitive to reservoir fractures through seismic data, establish a correlation between the comprehensive fracture prediction index and seismic attributes, and use seismic attributes to invert to obtain a comprehensive fracture prediction model to predict fracture development zones.

[0093] The reservoir fracture prediction method based on well seismic combined with artificial neural network provided in the embodiment of the present invention is referred to Figure 1 and Figure 2 As shown in the figure, reservoir core data, conventional well logging curves, seismic data and other data are collected to predict the fracture development zone through the following steps. First, the comprehensive fracture prediction index is calculated through the fracture response characteristics of conventional well logging curves, and the vertical development of fractures is analyzed in combination with laboratory core observations; at the same time, seismic attributes are extracted using post-stack seismic data, and seismic attributes that are more sensitive to changes in the comprehensive fracture prediction index are selected through qualitative and quantitative comparative analysis. For the sake of distinction, they are called target seismic attributes here; finally, the functional relationship between the comprehensive fracture prediction index and the target seismic attribute is established through the artificial neural network method, realizing the seismic multi-attribute inversion of reservoir fracture development zones.

[0094] The specific steps include the following:

[0095] S101. Determine multiple fracture parameters characterizing the degree of fracture development based on the acquired well logging data of the study area;

[0096] The conventional logging data of the study area are obtained, the fracture response characteristics corresponding to the logging curves are analyzed, and the quantitative parameters characterizing the degree of fracture development are defined to obtain the above-mentioned multiple fracture parameters.

[0097] S102, determining a comprehensive crack prediction index based on the multiple crack parameters;

[0098] The collected core data, i.e. the core photos of the drilling, were analyzed to observe the characteristics of the cracks, pores, mineral composition, etc., mark the depth section of the crack development, and compare with each fracture parameter reflecting the degree of fracture development in S101 to verify the core fracture development corresponding to the depth section with large fluctuations in each fracture parameter. At the same time, the oil test and dynamic production data of the drilling, including production, pressure, fluid type, etc., were analyzed to observe the correlation between the oil test and production data and multiple fracture parameters reflecting the degree of fracture development in S101. Based on the prior knowledge that the location with good fracture development corresponds to a high oil test yield, multiple fracture parameters are compared with the oil test and production data. If the fracture parameter has a depth section with large fluctuations and the oil test yield is high, it means that the fracture parameter can reflect the degree of fracture development.

[0099] Different weights are assigned to different fracture parameters according to the comparison results. Specifically, the ratio of the thickness of the depth section where each fracture parameter has a large fluctuation to the total thickness of the fracture development depth section of the core is determined to obtain the weight corresponding to the fracture parameter. According to each fracture parameter and its corresponding weight, a comprehensive fracture prediction index calculation function is constructed. Based on the multiple fracture parameters and the comprehensive fracture prediction index calculation function, a weighted average algorithm is used to calculate the comprehensive fracture prediction index.

[0100] S103, extracting seismic attributes using the acquired seismic data, and selecting target seismic attributes that are sensitive to changes in the comprehensive fracture prediction index;

[0101] Seismic data of the study area were obtained. Based on previous knowledge, attributes such as frequency, phase, amplitude, and waveform can reveal the development characteristics of fractures. In the experimental case, professional geological research software was used to extract various seismic attributes such as instantaneous frequency, instantaneous phase, root mean square amplitude, maximum amplitude, curvature, relative wave impedance, and coherence volume. The seismic attribute data was normalized and dimensioned, and then the intersection diagram of the seismic attribute and the comprehensive fracture prediction index of the previous step S102 was drawn. Finally, seismic attributes with obvious correlation with the comprehensive fracture prediction index calculated by S101 were selected to obtain the target seismic attributes. The preferred target seismic attributes are prepared to provide input data for the establishment of the next step comprehensive fracture prediction model.

[0102] S104, based on the seismic attribute data set corresponding to the target seismic attribute, training a pre-established artificial neural network, establishing a functional relationship between a comprehensive fracture prediction index and the target seismic attribute, and obtaining a comprehensive fracture prediction model;

[0103] The correlation between the comprehensive fracture prediction index and the target seismic attribute is established, and the comprehensive fracture prediction model is trained through an unsupervised artificial neural network. The unsupervised artificial neural network performs clustering, dimensionality reduction and feature extraction on seismic attribute data. Neural networks are used as a data processing tool to discover patterns, structures and relationships in data. It includes an input layer, a hidden layer and an output layer, where the hidden layer has multiple nodes to capture features in the data. The input data passes through each layer of the neural network, is processed by weights and activation functions, and finally the output is obtained. In a neural network, each node performs some mathematical operations, such as weighted sum, activation function, etc. Set the output representation of the node as:

[0104]

[0105] Among them, y is the output of the node, f is the activation function, and w i is the weight, x i is the input data, and b is the bias.

[0106] The seismic attribute data corresponding to the target seismic attribute is used as the data input of the artificial neural network, and the hyperparameters of the neural network such as the number of layers, the number of nodes, and the activation function are set; the training parameters such as the number of training iterations and the learning rate are defined, and the neural network performs a back propagation algorithm based on the data to gradually optimize the weights and biases to reduce the error between the predicted results and the actual data; after the training is completed, check whether the model can reasonably reflect the geological understanding; if not, adjust the training parameters until the model outputs a comprehensive fracture prediction model that can reasonably predict and conform to the actual geological understanding of the experimental area.

[0107] S105, acquiring seismic attribute data of the target area, performing reservoir fracture prediction based on the comprehensive fracture prediction model, and obtaining fracture prediction results.

[0108] The seismic attribute data of the target area is data corresponding to the target seismic attributes. The seismic attribute data of the target area is input into the comprehensive fracture prediction model to predict reservoir fractures, and fracture prediction results can be obtained, so as to realize seismic multi-attribute inversion of reservoir fracture development zones, effectively identify and predict reservoir fractures, and provide a reliable basis for reservoir evaluation in oil and gas exploration.

[0109] The present invention belongs to the field of sedimentary reservoir evaluation in petroleum geological exploration, and relates to a reservoir fracture prediction method. The present invention aims to utilize well logging technology, seismic fracture prediction methods, and artificial neural network technology to effectively identify and predict reservoir fractures, and provide a reliable basis for reservoir evaluation in oil and gas exploration. The present invention is applicable to reservoirs with developed fractures, such as tight sandstone and carbonate rock. The present invention integrates the accuracy of well logging curves in identifying the development of longitudinal fractures and the breadth of seismic data, effectively combines and innovates the existing technologies, and applies them to the new field of artificial intelligence, making fracture prediction more intelligent. The data and tools required by the present invention are all conventional geological basic data. The data are easy to obtain and low in cost. The technical process is easy to operate. The invention is highly practical and can comprehensively, comprehensively, accurately, and creatively predict the reservoir space fracture development zone.

[0110] The method for predicting reservoir fractures based on well-seismic combined with artificial neural network provided by the embodiment of the present invention determines multiple fracture parameters characterizing the degree of fracture development based on the well logging data of the study area obtained; determines the comprehensive fracture prediction index based on the determined multiple fracture parameters; extracts seismic attributes using seismic data, and selects the seismic attributes that are more sensitive to the changes of the above-mentioned comprehensive fracture prediction index through qualitative and quantitative comparative analysis to obtain the target seismic attributes; then, based on the pre-established artificial neural network, establishes a functional relationship between the comprehensive fracture prediction index and the target seismic attribute to obtain a comprehensive fracture prediction model, and obtains the comprehensive fracture prediction model through multiple seismic attributes that are more sensitive to the changes of the comprehensive fracture prediction index through seismic multi-attribute inversion, thereby realizing the seismic multi-attribute inversion of reservoir fracture development zones. The present invention integrates the fracture information reflected by seismic data and well logging data through machine learning, thereby obtaining more comprehensive underground fracture development information. The present invention combines the accuracy of well logging data and the breadth of seismic data to predict fractures on a spatial scale, and well predicts the development of fracture zones in actual block applications.

[0111] The present invention acquires conventional logging data of the study area, analyzes the fracture response characteristics corresponding to the logging curve, and defines quantitative parameters characterizing the degree of fracture development, namely, the above-mentioned multiple fracture parameters, namely: logging curve change rate, formation factor ratio, resistivity correction difference ratio, saturation ratio. The above-mentioned logging curve change rate is specifically the acoustic wave curve change rate.

[0112] The development of fractures will cause the acoustic time difference to jump periodically, so the data change rate of the acoustic curve will increase abnormally. The present invention obtains conventional logging curve data and calculates the acoustic curve change rate. In an optional embodiment, the fracture parameters include the acoustic curve change rate, and the acoustic curve change rate is calculated by the following formula:

[0113]

[0114] In the formula, Δai is the rate of change of the sound wave curve, a i is the sampling point of the sound wave curve.

[0115] In an optional embodiment, the fracture parameter includes a formation factor ratio; the formation factor ratio is calculated by the following formula:

[0116]

[0117]

[0118]

[0119] Where b is the formation factor ratio, b 1 is the formation factor calculated from resistivity logging, b 2 is the formation factor calculated from the porosity, R 0 is the deep lateral resistivity, R W is the formation water resistivity, a is the lithology coefficient, φ is the porosity, and m is the rock pore structure index.

[0120] Among them, a is the lithology coefficient. For pure sandstone with good physical properties, a is generally taken as 0.6-1.5, and can be set according to the actual situation of the study area; m is the rock pore structure index. For pure sandstone with good physical properties, m is generally taken as 1.5-3, and can be set according to the actual situation of the study area.

[0121] The present invention uses the ratio of the formation factor calculated by resistivity logging and the formation factor obtained by porosity logging to characterize the fracture development. The calculation process of the formation factor calculated by resistivity logging and the formation factor obtained by porosity logging can be described in the prior art and will not be repeated here.

[0122] In an optional embodiment, the crack parameter includes a resistivity correction difference ratio; the resistivity correction difference ratio is calculated by the following formula:

[0123]

[0124] Where c is the resistivity correction difference ratio, R is the true resistivity of the formation, and RLLS is the shallow lateral resistivity.

[0125] In the formation fracture development zone, mud and filtrate invade the formation, with strong invasion near the wellbore wall and weak invasion far from the wellbore wall, resulting in the water saturation of the flushing zone being greater than the water saturation. In an optional embodiment, the fracture parameter includes a saturation ratio; the saturation ratio is calculated by the following formula:

[0126]

[0127] Where d is the saturation ratio, S W is water saturation; S XO is the water saturation of the flushing zone, R W is the formation water resistivity, R XO is the resistivity of the flushing zone, R mf is the mud resistivity, R is the true resistivity of the formation, and m is the rock pore structure index.

[0128] The present invention performs mathematical processing transformations such as normalization and standardization on the above-mentioned crack parameters to unify the dimensions and magnitudes of the crack parameters, so as to facilitate comparative analysis. In an optional embodiment, before determining the comprehensive crack prediction index based on the multiple crack parameters in the above-mentioned step S102, the method further includes:

[0129] S1011, normalizing the crack parameters using the following formula to obtain the normalized crack parameters:

[0130]

[0131] In the formula, x is the crack parameter after normalization, and y is the original crack parameter (i.e., Δa i , b, c, d), where λ is the transformation parameter; the method to determine the transformation parameter λ is to try a series of different values, ranging from -5 to 5. Plot a histogram or probability density function of the transformed data to observe whether the data is close to a normal distribution. Based on visual judgment, choose the λ value that makes the data closest to a normal distribution.

[0132] The normalized crack parameters are standardized by the following formula to obtain the standardized crack parameters:

[0133]

[0134] In the formula, is the crack parameter after normalization, x max and x min are the maximum and minimum values ​​of the normalized crack parameters. The original crack parameters are mapped to the interval [0,1] through normalization.

[0135] In an optional embodiment, the step S102 described above of determining a comprehensive crack prediction index based on the multiple crack parameters includes:

[0136] S1021. Based on the standardized crack parameters, a comprehensive crack prediction index is calculated by the following formula:

[0137]

[0138] Where f is the comprehensive crack prediction index, is the crack parameter after normalization, k i is the parameter weight;

[0139] in,

[0140] k i =t i / T

[0141] Where, t i The fracture parameters of the coring section of the well reflect the thickness of the fracture development, and T is the total thickness of the coring section. The larger the comprehensive fracture prediction index, the more developed the fractures.

[0142] The embodiment of the present invention assigns different parameter weights according to the strength of the crack identification ability of different crack parameters. The above four crack parameters reflecting the degree of crack development (including the acoustic wave curve change rate Δa i , formation factor ratio b, resistivity correction difference ratio c, saturation ratio d) parameter weight k i Determined by:

[0143] The collected core data, i.e. the core photos of the drilling, were analyzed to observe the characteristics of the cracks, pores, mineral composition, etc., mark the depth section of the crack development, and compare with each fracture parameter of S101 reflecting the degree of fracture development to verify the core fracture development corresponding to the depth section where the above four fracture parameters have large fluctuations. At the same time, the oil test and dynamic production data of the drilling, including production, pressure, fluid type, etc., were analyzed to observe the correlation between the oil test and production data and the four fracture parameters of S101 reflecting the degree of fracture development. Based on the prior knowledge that the location with good fracture development corresponds to a high oil test yield, the above four fracture parameters are compared with the oil test and production data. If the fracture parameter has a depth section with large fluctuations and the oil test yield is high, it means that the fracture parameter can reflect the degree of fracture development.

[0144] Different weights are assigned to different fracture parameters according to the comparison results. Specifically, the ratio of the thickness of the depth section where each fracture parameter has a large fluctuation to the total thickness of the fracture development depth section of the core is determined to obtain the weight corresponding to the fracture parameter. According to each fracture parameter and its corresponding weight, a comprehensive fracture prediction index calculation function is constructed. Based on the multiple fracture parameters and the comprehensive fracture prediction index calculation function, a weighted average algorithm is used to calculate the comprehensive fracture prediction index, which is ready to establish a correlation with the seismic attributes in the next step.

[0145] In an optional embodiment, the seismic data includes a post-stack seismic data volume;

[0146] The step S103 described above of extracting seismic attributes using the acquired seismic data to select target seismic attributes that are sensitive to changes in the comprehensive fracture prediction index includes:

[0147] S1031, acquiring a post-stack seismic data volume, and extracting a plurality of seismic attributes of a target layer from the post-stack seismic data volume; wherein the plurality of seismic attributes include an instantaneous frequency attribute, an instantaneous phase attribute, a root mean square amplitude attribute, a maximum amplitude attribute, a curvature attribute, a relative wave impedance attribute, a coherence volume attribute, and a multi-scale edge detection attribute;

[0148] The post-stack seismic data volume of the study area is obtained. Based on the prior knowledge of predecessors, the attributes such as frequency, phase, amplitude, and waveform can reveal the development characteristics of fractures. In the experimental example, professional geological research software is used to extract various seismic attributes such as instantaneous frequency, instantaneous phase, root mean square amplitude, maximum amplitude, curvature, relative wave impedance, coherence volume, and multi-scale edge detection by setting reasonable iteration parameters. The extraction methods of the above attributes can refer to the description in the prior art and will not be repeated here.

[0149] S1032. Calculate the correlation between each seismic attribute and the comprehensive fracture prediction index respectively, and select target seismic attributes whose correlation is greater than a preset correlation threshold from the multiple seismic attributes.

[0150] Normalize the seismic attribute data, unify the dimensions, and then draw the cross-plot of seismic attributes and the comprehensive fracture prediction index in the previous step. The above preset correlation threshold can be set according to actual needs, refer to Figure 2 As shown, the present invention analyzes the correlation between each seismic attribute and the comprehensive crack prediction index calculated in step S102, and preferably selects the seismic attributes with a correlation greater than 0.85 as the input data of the next step S104. In this embodiment, the coherence volume attribute, curvature attribute, relative wave impedance attribute, and root mean square amplitude attribute have a good correlation with the comprehensive crack prediction index, and the correlation coefficients are all greater than 0.85. Finally, the coherence volume attribute, curvature attribute, relative wave impedance attribute, and root mean square amplitude attribute that have a significant correlation with the comprehensive crack prediction index calculated in the first step are selected to obtain the above-mentioned target seismic attributes. The preferred target seismic attributes are prepared to provide input data for the establishment of the comprehensive crack prediction model in the next step S104.

[0151] Before extracting seismic attributes using the acquired seismic data and selecting target seismic attributes that are sensitive to changes in the comprehensive fracture prediction index in S103, the method further includes: converting the seismic data into a depth domain by time-to-depth conversion.

[0152] The logging data, i.e. the above-mentioned logging curve, is in the depth domain. The seismic data is first converted into the depth domain, and then the seismic attributes are extracted. The correlation between the comprehensive fracture prediction index and the seismic attributes is analyzed in the depth domain. The present invention matches and corrects the logging data and seismic data through time-depth conversion, which is one of the key steps to ensure that the logging curve and the post-stack seismic data volume are spatially and physically consistent. The steps are as follows:

[0153] Before starting the time-depth conversion, well logging data and seismic data need to be prepared. Make sure that the data set contains well logging curves and post-stack seismic data volumes. In order to perform time-depth conversion, a reference data set needs to be established, usually well logging data. These data must have spatial coordinates (longitude and latitude) and time-depth relationships. Use well logging data to establish a time-depth relationship model for underground strata. This model associates time and depth for subsequent conversion of seismic data. According to the established time-depth relationship model, the seismic data is interpolated and extrapolated to map the seismic data to depth coordinates. The converted data is quality controlled to check whether there are outliers or errors, and corrections or data cleaning are performed if necessary. Use seismic data that has been converted through time-depth conversion to verify and compare with well logging data. This step can help ensure the spatial and physical consistency of the data. If the conversion effect is not good, the time-depth relationship model needs to be reviewed and modified. The present invention collects well logging and geophysical data, including well logging curves, post-stack seismic data volumes, etc., matches and corrects different data to ensure spatial and physical consistency.

[0154] The present invention uses artificial neural network technology, takes the seismic attribute data set corresponding to the preferred target seismic attribute as input data, trains the functional relationship between the comprehensive crack prediction index and the seismic attribute, adjusts reasonable parameters, and outputs a comprehensive crack prediction model. In an optional embodiment, the above step S104, based on the seismic attribute data set corresponding to the target seismic attribute, trains a pre-established artificial neural network, establishes the functional relationship between the comprehensive crack prediction index and the target seismic attribute, and obtains the comprehensive crack prediction model, including:

[0155] S1041, standardizing and normalizing the seismic attribute data set corresponding to the target seismic attribute to obtain a processed seismic attribute data set;

[0156] The seismic attribute data corresponding to the target seismic attribute selected in step S103 are standardized and normalized to ensure the uniformity of the data dimension for neural network training.

[0157] S1042, dividing the processed earthquake attribute data set into a training set and a test verification set according to a preset ratio;

[0158] The above preset ratio can be set according to actual needs, such as 3:2.

[0159] S1043, using the training set as input data of the artificial neural network, optimizing model parameters through a back propagation algorithm, and obtaining a trained model;

[0160] The processed seismic attribute data set is used as the data input of the artificial neural network, and the hyperparameters of the neural network, such as the number of layers, the number of nodes, and the activation function, are set; the training parameters, such as the number of training iterations and the learning rate, are defined. The artificial neural network gradually optimizes the weights and biases of the artificial neural network through the back propagation algorithm to reduce the error between the predicted results and the actual fracture development corresponding to the input data.

[0161] The architecture of the artificial neural network used in the embodiment of the present invention is as follows:

[0162] import tensorflow as tf

[0163] #Create a neural network model

[0164] model = tf.keras.Sequential()

[0165] #Input layer (four earthquake attributes)

[0166] model.add(tf.keras.layers.Input(shape=(4,)))#4 earthquake attribute features

[0167] #Hidden layer 1

[0168] model.add(tf.keras.layers.Dense(units=64,activation='relu'))

[0169] #Hidden layer 2

[0170] model.add(tf.keras.layers.Dense(units=32,activation='relu'))

[0171] #Output layer (comprehensive crack prediction index, single numerical output)

[0172] model.add(tf.keras.layers.Dense(units=1,activation='linear'))

[0173] # Compile the model

[0174] model.compile(optimizer=tf.keras.optimizers.Adam(learning_rate=0.001),

[0175] loss = 'mean_squared_error')

[0176] #View model architecture

[0177] model.summary()

[0178] The artificial neural network comprises an input layer, two hidden layers and an output layer.

[0179] Input layer: The input layer has four nodes, each corresponding to an earthquake attribute. These four nodes receive and pass the information of earthquake attributes to the neural network.

[0180] Hidden layer 1 and hidden layer 2: These two hidden layers are the middle layers of the neural network, which are used to learn the complex relationship between the input features and the comprehensive crack prediction index. Hidden layer 1 contains 64 neurons and hidden layer 2 contains 32 neurons. The neurons in these hidden layers use activation functions (ReLU, Rectified Linear Unit) to introduce nonlinear properties so that the network can capture nonlinear relationships in the data.

[0181] Output layer: The output layer is the last layer of the neural network, which generates the prediction results of the comprehensive crack prediction index. The output layer of the present invention has only one node because a single continuous value is to be predicted. The activation function is set to linear ('linear') so that the output layer can generate real-valued prediction results.

[0182] Compile the model: When compiling the model, the optimizer (Adam) and loss function (mean squared error, 'mean_squared_error') are selected. The optimizer is responsible for adjusting the weights and biases of the neural network to minimize the loss function. The mean squared error is used to measure the difference between the predicted value and the true value.

[0183] Training process: During the training process, the neural network will use the four target seismic attributes in the training data set as input, pass information through the hidden layer, and finally produce a predicted value of the comprehensive fracture prediction index. Then, the mean square error between the model prediction result and the true comprehensive fracture prediction index is calculated, and the back propagation algorithm is used to adjust the weights and biases of the network to reduce the loss. This process will be repeated many times, determined by the number of iterations, until the model converges to the best prediction result.

[0184] Model evaluation: After training, use an independent validation set to evaluate the performance of the model. Use mean squared error (MSE), mean absolute error (MAE), etc. The model needs to be adjusted and optimized multiple times to obtain the best results.

[0185] S1044. Verify the trained model based on the obtained fracture observation results, core data, drilling data and oil test data of the study area, and the verification set, and repeat the training process until the iteration stop condition is reached to obtain the comprehensive fracture prediction model.

[0186] After the training is completed, check whether the trained model can reasonably reflect the geological knowledge. If not, repeat the above process of S1043 and continue to adjust the training parameters until a comprehensive fracture prediction model that can reasonably predict and conform to the actual geological knowledge of the experimental area is output. The above iteration stop condition is to reach the maximum number of iterations. In the embodiment of the present invention, the maximum number of iterations is 5000.

[0187] The present invention uses geological research results, core, drilling, oil testing and other data in the study area to conduct a comparative test on the comprehensive fracture prediction model, verify the reliability of the fracture prediction results, and finally complete the fracture prediction work in the reservoir area.

[0188] In an optional embodiment, the trained model is verified by:

[0189] S201, inputting the verification set into the trained model to obtain a model prediction result;

[0190] S202, obtaining fracture observation results, core data, drilling data and oil test data in the study area, and aligning the fracture observation results, the core data, the drilling data and the oil test data according to the geographical location and depth of the well;

[0191] Collect geological research results, cores, drilling, oil testing and other data in the study area, including fracture observation results, core data, drilling data, oil testing data, etc. Align these data by the geographical location and depth position of the wells to compare with the seismic data and the results of the fracture prediction model.

[0192] S203, verifying the model prediction results based on the fracture observation results, core data, drilling data and oil test data, and determining whether the error rate of the trained model is lower than a preset error threshold.

[0193] According to the fracture observation results, core data, drilling data and oil test data, the model prediction results predicted by the trained model are compared to determine whether the error rate of the trained model is lower than the preset error threshold. If not, it means that the fracture development characteristics predicted by the model cannot match the geological data and the model reliability is low.

[0194] In an optional embodiment, the model prediction result is verified by the following method:

[0195] Comparing the crack observation results with the model prediction results to determine whether the trained model can accurately predict cracks at corresponding locations;

[0196] Comparing the core data and drilling data with the model prediction results to determine whether the fracture locations predicted by the trained model are consistent with the actual fracture distribution;

[0197] Compare the oil test data with the model prediction results to determine whether there are underground features in the model prediction results that are consistent with the degree of fracture development and the oil test production. Observe whether there are underground features that are consistent with the degree of fracture development and the oil test production, that is, locations with good fracture development have high oil test production. By comparing the model prediction results with the actual data, the model prediction compliance rate is calculated. If the compliance rate is greater than 80%, or the model error rate is calculated. If the model error rate is lower than the above-mentioned preset error threshold, it means that the model can accurately predict the fracture development characteristics and is consistent with the geological data, indicating that the model is reliable. The above-mentioned preset error threshold can be set according to actual needs, such as 10%.

[0198] The fracture information reflected by seismic data and well logging data is integrated through machine learning to obtain more comprehensive underground fracture development information. The present invention combines the accuracy of well logging data and the breadth of seismic data to predict fractures on a spatial scale, and can well predict the development of fracture zones in actual block applications.

[0199] In order to verify the accuracy of the well-seismic combined reservoir fracture prediction method based on artificial neural network provided in the embodiment of the present invention, the present invention applies the well-seismic combined reservoir fracture prediction method based on artificial neural network to Nangang I District and Xiaoji Oilfield. The present invention has achieved good application effect in the prediction of tight sandstone reservoir fractures in Nangang I District and Xiaoji Oilfield. Figure 3 As shown, it can be seen from the prediction map that the cracks are mainly developed near the fault zone. The cracks in the area between fault zone 1 and fault zone 2 have a large development range, mainly concentrated in the area near wells 1, 2, 3, and 4. The core scanning electron microscope photos, oil test data, production data and other information on the wells are used for verification, and it is found that the prediction compliance rate reaches 80%. The invention can be applied and promoted in the reservoir evaluation stage of oil and gas exploration, and has reference significance for the evaluation of other similar mineral reservoirs.

[0200] Embodiment 2

[0201] An embodiment of the present invention provides a comprehensive crack prediction model training method, comprising:

[0202] Based on the well logging data obtained in the study area, multiple fracture parameters characterizing the degree of fracture development are determined;

[0203] Determining a comprehensive fracture prediction index based on the plurality of fracture parameters;

[0204] Extracting seismic attributes using the acquired seismic data, and selecting target seismic attributes that are sensitive to changes in the comprehensive fracture prediction index;

[0205] Based on the seismic attribute data set corresponding to the target seismic attribute, a pre-established artificial neural network is trained to establish a functional relationship between a comprehensive fracture prediction index and the target seismic attribute to obtain a comprehensive fracture prediction model.

[0206] The specific implementation of the comprehensive fracture prediction model training method provided in the embodiment of the present invention can refer to the detailed description of the well-seismic combined reservoir fracture prediction method based on artificial neural network in the above embodiment, and the repeated parts will not be repeated.

[0207] Embodiment 3

[0208] The embodiment of the present invention provides a reservoir fracture prediction device based on well seismic and artificial neural network, referring to Figure 4 As shown, including:

[0209] The first calculation module 301 is used to determine a plurality of fracture parameters characterizing the degree of fracture development based on the acquired well logging data of the study area;

[0210] A second calculation module 302 is used to determine a comprehensive fracture prediction index based on the multiple fracture parameters;

[0211] A screening module 303 is used to extract seismic attributes using the acquired seismic data, and select target seismic attributes that are sensitive to changes in the comprehensive fracture prediction index;

[0212] A training module 304 is used to establish a functional relationship between the comprehensive fracture prediction index and the target seismic attribute based on a pre-established artificial neural network to obtain a comprehensive fracture prediction model;

[0213] The prediction module 305 is used to predict reservoir fractures based on the comprehensive fracture prediction model to obtain fracture prediction results.

[0214] The implementation principle and technical effect of the well-seismic combined reservoir fracture prediction device based on artificial neural network provided in the embodiment of the present invention are similar to any of the aforementioned method embodiments, which will not be repeated here.

[0215] Embodiment 4

[0216] An embodiment of the present invention provides a comprehensive crack prediction model training device, comprising:

[0217] The first calculation module is used to determine a plurality of fracture parameters characterizing the degree of fracture development based on the acquired well logging data of the study area;

[0218] A second calculation module, configured to determine a comprehensive fracture prediction index based on the plurality of fracture parameters;

[0219] A screening module, used to extract seismic attributes using the acquired seismic data, and select target seismic attributes that are sensitive to changes in the comprehensive fracture prediction index;

[0220] The training module is used to train a pre-established artificial neural network based on a seismic attribute data set corresponding to the target seismic attribute, establish a functional relationship between a comprehensive fracture prediction index and the target seismic attribute, and obtain a comprehensive fracture prediction model.

[0221] The implementation principle and technical effect of the comprehensive crack prediction model training device provided in the embodiment of the present invention are similar to those of any of the aforementioned method embodiments, and will not be repeated here.

[0222] Embodiment 5

[0223] An embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the method for predicting reservoir fractures by combining well-seismic analysis with an artificial neural network as described in any of the aforementioned method embodiments, or the method for training a comprehensive fracture prediction model as described in any of the aforementioned method embodiments, is implemented.

[0224] The computer-readable storage medium may be included in the device / apparatus described in the above embodiment; or it may exist independently without being assembled into the device / apparatus. The above computer-readable storage medium carries one or more programs, and when the above one or more programs are executed, the method according to the embodiment of the present invention is implemented.

[0225] According to an embodiment of the present invention, the computer-readable storage medium may be a non-volatile computer-readable storage medium, for example, may include but is not limited to: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, the computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, apparatus, or device.

[0226] Embodiment 6

[0227] An embodiment of the present invention provides an electronic device, referring to Figure 5 As shown, it includes a processor 111, a communication interface 112, a memory 113 and a communication bus 114, wherein the processor 111, the communication interface 112 and the memory 113 communicate with each other through the communication bus 114.

[0228] Memory 113, used for storing computer programs;

[0229] The processor 111 is used to execute the program stored in the memory 113 to implement the well-seismic combined reservoir fracture prediction method based on artificial neural network as described in any of the above method embodiments, or the comprehensive fracture prediction model training method as described in any of the above method embodiments.

[0230] The implementation principle and technical effect of the electronic device provided by the embodiment of the present invention are similar to those of any of the aforementioned method embodiments, and will not be repeated here.

[0231] The above-mentioned memory 113 can be an electronic memory such as a flash memory, an EEPROM (electrically erasable programmable read-only memory), an EPROM, a hard disk or a ROM. The memory 113 has a storage space for program codes for executing any method steps in the above-mentioned method. For example, the storage space for program codes may include various program codes for implementing various steps in the above method respectively. These program codes can be read from or written into one or more computer program products. These computer program products include program code carriers such as hard disks, compact disks (CDs), memory cards or floppy disks. Such computer program products are usually portable or fixed storage units. The storage unit may have a storage segment or storage space arranged similarly to the memory 113 in the above-mentioned electronic device. The program code can be compressed, for example, in an appropriate form. Generally, the storage unit includes a program for executing the method steps according to an embodiment of the present invention, that is, a code that can be read by, for example, a processor 111, and when these codes are run by an electronic device, the electronic device performs the various steps in the method described above.

[0232] In this document, relational terms such as first and second, etc. are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprises" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. The orientation or position relationship indicated by the terms "upper", "lower", etc. is based on the orientation or position relationship shown in the accompanying drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0233] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. It should be noted that the embodiments of the present invention and the features in the embodiments can be combined with each other without conflict. The present invention is not limited to any single aspect, any single embodiment, or any combination and / or permutation of these aspects and / or embodiments. Each aspect and / or embodiment of the present invention can be used alone, or in combination with one or more other aspects and / or other embodiments.

[0234] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A reservoir fracture prediction method based on well-seismic data combined with artificial neural network, characterized in that: include: Obtain seismic attribute data of the target area, perform reservoir fracture prediction based on the comprehensive fracture prediction model, and obtain fracture prediction results; The comprehensive crack prediction model is trained in the following way: Based on the well logging data obtained in the study area, multiple fracture parameters characterizing the degree of fracture development are determined; Determining a comprehensive fracture prediction index based on the plurality of fracture parameters; Extracting seismic attributes using the acquired seismic data, and selecting target seismic attributes that are sensitive to changes in the comprehensive fracture prediction index; Based on the seismic attribute data set corresponding to the target seismic attribute, a pre-established artificial neural network is trained to establish a functional relationship between a comprehensive fracture prediction index and the target seismic attribute to obtain a comprehensive fracture prediction model.

2. The method for predicting reservoir fractures based on well-seismic data combined with artificial neural network according to claim 1 is characterized in that: The seismic data includes a post-stack seismic data volume; The method of extracting seismic attributes using the acquired seismic data and selecting target seismic attributes that are sensitive to changes in the comprehensive fracture prediction index includes: Acquire a post-stack seismic data volume, and extract multiple seismic attributes of a target layer from the post-stack seismic data volume; wherein the multiple seismic attributes include instantaneous frequency attributes, instantaneous phase attributes, root mean square amplitude attributes, maximum amplitude attributes, curvature attributes, relative wave impedance attributes, coherence volume attributes, and multi-scale edge detection attributes; The correlation between each seismic attribute and the comprehensive fracture prediction index is calculated respectively, and a target seismic attribute having a correlation greater than a preset correlation threshold is screened out from the multiple seismic attributes.

3. The method for predicting reservoir fractures based on well-seismic data combined with artificial neural network according to claim 1 is characterized in that: The method of training a pre-established artificial neural network based on a seismic attribute data set corresponding to the target seismic attribute, establishing a functional relationship between a comprehensive crack prediction index and the target seismic attribute, and obtaining a comprehensive crack prediction model includes: Standardizing and normalizing the seismic attribute data set corresponding to the target seismic attribute to obtain a processed seismic attribute data set; Dividing the processed earthquake attribute data set into a training set and a test validation set according to a preset ratio; Using the training set as input data of the artificial neural network, optimizing model parameters through a back propagation algorithm, and obtaining a trained model; The trained model is verified based on the obtained fracture observation results, core data, drilling data and oil test data of the study area, and the verification set, and the training process is repeated until the iteration stop condition is reached to obtain the comprehensive fracture prediction model.

4. The method for predicting reservoir fractures based on well-seismic data combined with artificial neural network according to claim 3 is characterized in that: The trained model is verified by the following method: Inputting the validation set into the trained model to obtain a model prediction result; Obtain fracture observation results, core data, drilling data and oil test data in the study area, and align the fracture observation results, the core data, the drilling data and the oil test data according to the geographical location and depth position of the well; The model prediction results are verified based on the fracture observation results, core data, drilling data and oil test data to determine whether the error rate of the trained model is lower than a preset error threshold.

5. The method for predicting reservoir fractures based on well-seismic data combined with artificial neural network according to claim 4 is characterized in that: The prediction results of the model are verified by the following methods: Comparing the crack observation results with the model prediction results to determine whether the trained model can accurately predict cracks at corresponding locations; Comparing the core data and drilling data with the model prediction results to determine whether the fracture locations predicted by the trained model are consistent with the actual fracture distribution; The oil test data is compared with the model prediction results to determine whether there are underground features in the model prediction results that are consistent with the degree of fracture development and the oil test production.

6. The method for predicting reservoir fractures based on well-seismic data combined with artificial neural network according to claim 1, characterized in that: The crack parameters include the acoustic wave curve change rate, which is calculated by the following formula: In the formula, Δa i is the rate of change of the sound wave curve, a i is the sampling point of the sound wave curve.

7. The method for predicting reservoir fractures based on well-seismic data combined with artificial neural network according to claim 1, characterized in that: The fracture parameters include a formation factor ratio; the formation factor ratio is calculated by the following formula: Where b is the formation factor ratio, b1 is the formation factor calculated by resistivity logging, b2 is the formation factor calculated by porosity, R0 is the deep lateral resistivity, R W is the formation water resistivity, a is the lithology coefficient, φ is the porosity, and m is the rock pore structure index.

8. The method for predicting reservoir fractures based on well-seismic data combined with artificial neural network according to claim 1, characterized in that: The crack parameters include a resistivity correction difference ratio; the resistivity correction difference ratio is calculated by the following formula: Where c is the resistivity correction difference ratio, R is the true resistivity of the formation, and RLLS is the shallow lateral resistivity.

9. The method for predicting reservoir fractures based on well-seismic data combined with artificial neural network according to claim 1, characterized in that: The fracture parameters include a saturation ratio; the saturation ratio is calculated by the following formula: In the formula, d is the saturation ratio, R W is the formation water resistivity, R XO is the resistivity of the flushing zone, R mf is the mud resistivity, R is the true resistivity of the formation, and m is the rock pore structure index.

10. The method for predicting reservoir fractures based on well-seismic data combined with artificial neural network according to claim 1, characterized in that: Before determining the comprehensive fracture prediction index based on the multiple fracture parameters, the method further includes: The fracture parameters are normalized by the following formula to obtain the normalized fracture parameters: In the formula, x is the crack parameter after normalization, y is the original crack parameter, and λ is the transformation parameter; The normalized crack parameters are standardized by the following formula to obtain the standardized crack parameters: In the formula, is the crack parameter after normalization, x max and x min are the maximum and minimum values ​​of the crack parameters after normalization.

11. The method for predicting reservoir fractures based on well-seismic data combined with artificial neural network according to claim 10, characterized in that: The step of determining a comprehensive crack prediction index based on the plurality of crack parameters comprises: Based on the standardized crack parameters, the comprehensive crack prediction index is calculated by the following formula: Where f is the comprehensive crack prediction index, is the crack parameter after normalization, k i is the parameter weight; in, k i =t i / T In the formula, t i The fracture parameters counted in the coring section of the drilling reflect the thickness of the fracture development, and T is the total thickness of the coring section.

12. A comprehensive crack prediction model training method, characterized in that: include: Based on the well logging data obtained in the study area, multiple fracture parameters characterizing the degree of fracture development are determined; Determining a comprehensive fracture prediction index based on the plurality of fracture parameters; Extracting seismic attributes using the acquired seismic data, and selecting target seismic attributes that are sensitive to changes in the comprehensive fracture prediction index; Based on the seismic attribute data set corresponding to the target seismic attribute, a pre-established artificial neural network is trained to establish a functional relationship between a comprehensive fracture prediction index and the target seismic attribute to obtain a comprehensive fracture prediction model.

13. A reservoir fracture prediction device based on well seismic data and artificial neural network, characterized in that: include: The first calculation module is used to determine a plurality of fracture parameters characterizing the degree of fracture development based on the acquired well logging data of the study area; A second calculation module, configured to determine a comprehensive fracture prediction index based on the plurality of fracture parameters; A screening module, used to extract seismic attributes using the acquired seismic data, and select target seismic attributes that are sensitive to changes in the comprehensive fracture prediction index; A training module, for establishing a functional relationship between the comprehensive fracture prediction index and the target seismic attribute based on a pre-established artificial neural network, and obtaining a comprehensive fracture prediction model; The prediction module is used to predict reservoir fractures based on the comprehensive fracture prediction model to obtain fracture prediction results.

14. A comprehensive crack prediction model training device, characterized in that: include: The first calculation module is used to determine a plurality of fracture parameters characterizing the degree of fracture development based on the acquired well logging data of the study area; A second calculation module, configured to determine a comprehensive fracture prediction index based on the plurality of fracture parameters; A screening module, used to extract seismic attributes using the acquired seismic data, and select target seismic attributes that are sensitive to changes in the comprehensive fracture prediction index; The training module is used to train a pre-established artificial neural network based on a seismic attribute data set corresponding to the target seismic attribute, establish a functional relationship between a comprehensive fracture prediction index and the target seismic attribute, and obtain a comprehensive fracture prediction model.

15. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method for predicting reservoir fractures based on well-seismic combined with artificial neural network as described in any one of claims 1 to 11, or the method for training a comprehensive fracture prediction model as described in claim 12 is implemented.

16. An electronic device, characterized in that: It includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus; Memory, used to store computer programs; The processor is used to implement the well-seismic combined reservoir fracture prediction method based on artificial neural network as described in any one of claims 1 to 11, or the comprehensive fracture prediction model training method as described in claim 12 when executing the program stored in the memory.

Images