Oil and gas saturation artificial intelligence prediction system and method based on frequency dispersion attribute

Abstract

The invention discloses an artificial intelligence prediction system for oil and gas saturation based on frequency dispersion attributes. The artificial intelligence prediction system comprises a data acquisition module for acquiring required data and generating a saturation curve belonging to depth domain data according to the acquired data; the time-depth calibration module obtains a synthetic seismic record according to the logging data and the post-stack seismic data, compares the synthetic seismic record with the actual seismic data to obtain a time-depth relationship, and generates a saturation curve belonging to the time domain data in combination with the saturation curve belonging to the depth domain data; a frequency dispersion attribute optimization module obtains frequency dispersion attributes of different angles and different frequencies of sound waves according to the pre-stack gather data, obtains contribution degrees of the frequency dispersion attributes to saturation prediction based on an SHAP method and saturation curve analysis of a time domain, and screens out optimal frequency dispersion attributes according to the contribution degrees; and the saturation prediction module inputs the preferable frequency dispersion attribute to a neural network model for saturation prediction. The method can solve the problems of geological hypothesis simplification, insufficient data representativeness, neglect of reservoir complexity and low calculation efficiency.

Description

Technical Field

[0001] The technical field of the present invention is oil exploration and development, and specifically refers to an artificial intelligence prediction system and method for oil and gas saturation based on dispersion attributes. Background Art

[0002] Oil and gas saturation is a key indicator for measuring the oil content in oil reservoirs, which is directly related to the development efficiency and economic benefits of oil reservoirs. Accurately constructing an oil and gas saturation model can significantly improve the estimation accuracy of the original geological reserves and recoverable reserves of oil reservoirs. This achievement has extremely important guiding value for scientifically formulating oilfield development plans, making reasonable investment decisions, and conducting comprehensive economic evaluations. It is the cornerstone for ensuring efficient development and sustainable development of oil fields.

[0003] However, the traditional method has exposed many defects in practical applications, which seriously restricts its prediction accuracy and practicality. The traditional three-dimensional oil and gas saturation modeling method is usually based on simplified geological assumptions, but the geological conditions of actual oil reservoirs are often complex and diverse. This simplified assumption makes the model unable to accurately reflect the real geological characteristics of the reservoir, thus affecting the accuracy of oil and gas saturation prediction; at the same time, the accuracy of the inter-well interpolation algorithm commonly used in the traditional method is highly dependent on the distribution and quality of the well point data. The well network distribution of offshore oil fields is sparse, and the distance between well points is large, resulting in insufficient representativeness of the inter-well data, further reducing the accuracy of the prediction; in addition, the traditional method ignores the complex structure and geological phenomena inside the reservoir, such as the influence of faults and fractures on fluid flow and saturation distribution, and thus cannot be included in the modeling scope; in the existing reservoir numerical simulation, the traditional method needs to solve the equations of pressure field and saturation field simultaneously, which is not only computationally intensive, but also prone to non-convergence and instability of the solution.

[0004] Therefore, inventing an artificial intelligence prediction system for oil and gas saturation based on dispersion attributes can solve problems such as geological simplification assumptions, uncertainty in inter-well interpolation, neglect of reservoir complexity and low computational efficiency, overcome the shortcomings of traditional oil and gas saturation prediction technology, and provide a more accurate and efficient solution for the efficient development of offshore oil fields, which has become a technical problem that needs to be solved urgently. Summary of the invention

[0005] The purpose of the present invention is to provide an artificial intelligence prediction system for oil and gas saturation based on dispersion attributes. The present invention can solve the problems of simplified geological assumptions, insufficient data representativeness, neglect of reservoir complexity and low computational efficiency in traditional oil and gas saturation prediction methods, improve prediction accuracy and computational efficiency, reduce dependence on data quality and computing resources, and improve the accuracy and reliability of seismic data interpretation.

[0006] To achieve this purpose, the present invention designs an artificial intelligence prediction system for oil and gas saturation based on dispersion attributes, which includes:

[0007] The data acquisition module is used to collect well logging data, post-stack seismic data, pre-stack gather data and actual seismic data, and generate a saturation curve belonging to the depth domain data based on the collected well logging data;

[0008] The time-depth calibration module is used to obtain synthetic seismic records based on well logging data and post-stack seismic data, compare the actual seismic data with the synthetic seismic records, obtain the corresponding relationship between the time of seismic wave propagation and the well logging depth, and generate the saturation curve belonging to the time domain data based on the corresponding relationship between the time of seismic wave propagation and the well logging depth and the saturation curve belonging to the depth domain data;

[0009] The dispersion attribute optimization module is used to obtain the dispersion attributes of acoustic waves at different angles and frequencies based on pre-stack gather data. The dispersion attributes of acoustic waves at different angles and frequencies are analyzed based on the SHAP method (SHapley Additive exPlanations, a model interpretation method based on game theory) and the saturation curve belonging to time domain data. The contribution of the dispersion attributes of acoustic waves at different angles and frequencies to saturation prediction is obtained, and the optimal dispersion attributes are selected according to the contribution to saturation prediction.

[0010] The saturation prediction module is used to input the preferred dispersion attribute into the neural network model trained according to the set target to perform saturation prediction.

[0011] Preferably, the specific method for comparing the actual seismic data with the synthetic seismic record to obtain the corresponding relationship between the time of seismic wave propagation and the logging depth is: obtaining the logging curve according to the logging data, and constructing the initial time-depth relationship through the acoustic wave time difference data in the logging curve; then comparing the synthetic seismic record with the actual seismic data, analyzing the relationship between the synthetic seismic record and the actual seismic data and adjusting them, so as to obtain the corresponding relationship between the time of seismic wave propagation and the logging depth.

[0012] Preferably, a set number of preferred dispersion properties are screened out according to the contribution of saturation prediction, wherein the specific screening requirement is: selecting a set number of dispersion properties with the highest contribution value and positive contribution.

[0013] Preferably, in the neural network model trained according to the set target, the training target of the neural network is to minimize the loss function, and the neural network training method adopted is: obtain the saturation label by logging curve calculation, and then calculate the loss function value between the saturation prediction result and the saturation label according to the cross entropy loss function, calculate the gradient of the loss function for each parameter in the neural network according to the back propagation algorithm, and then use the gradient descent algorithm to update the corresponding parameters in the neural network, and finally update the corresponding parameters in the neural network through repeated iterations until the loss function is minimized.

[0014] Beneficial effects of the present invention:

[0015] The present invention proposes an artificial intelligence prediction system for oil and gas saturation based on dispersion attributes, which combines artificial intelligence algorithms with oil and gas saturation prediction, effectively solving the shortcomings of traditional saturation modeling methods. Through synthetic record time-depth calibration, the logging saturation curve and seismic data are accurately matched to provide an accurate basis for subsequent processing; the SHAP method is used to optimize dispersion attributes, and attributes that are highly correlated with saturation are efficiently screened out, reducing labor costs and improving work efficiency; the neural network prediction model based on the preferred attributes can automatically fit complex nonlinear relationships without simplifying assumptions, significantly improving prediction accuracy and computational efficiency. The present invention combines the advantages of artificial intelligence algorithms, improves prediction accuracy and computational efficiency, reduces dependence on data quality and computing resources, and provides an efficient and reliable solution for seismic data interpretation. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a structural schematic diagram of the present invention;

[0017] Figure 2 It is a schematic diagram of the process of the present invention;

[0018] Figure 3 It is the SHAP dispersion attribute contribution analysis diagram of the present invention;

[0019] Figure 4 Schematic diagram of neural network structure;

[0020] Figure 5 This is a post-stack seismic profile;

[0021] Figure 6 Modeling profiles for conventional saturation;

[0022] Figure 7 This is a saturation prediction result diagram of the present invention; DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents the selected embodiments of the present invention. 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.

[0024] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0025] Example 1

[0026] An artificial intelligence prediction system for oil and gas saturation based on dispersion attributes, such as Figure 1 As shown, it includes:

[0027] The data acquisition module is used to collect well logging data, post-stack seismic data, pre-stack gather data and actual seismic data, and generate a saturation curve belonging to the depth domain data based on the collected well logging data;

[0028] The time-depth calibration module is used to obtain synthetic seismic records based on well logging data and post-stack seismic data, compare the actual seismic data with the synthetic seismic records, obtain the corresponding relationship between the time of seismic wave propagation and the well logging depth, and generate the saturation curve belonging to the time domain data based on the corresponding relationship between the time of seismic wave propagation and the well logging depth and the saturation curve belonging to the depth domain data;

[0029] The dispersion attribute optimization module is used to obtain the dispersion attributes of acoustic waves at different angles and frequencies based on pre-stack gather data, analyze the dispersion attributes of acoustic waves at different angles and frequencies based on the SHAP method and the saturation curve belonging to the time domain data, obtain the contribution of the dispersion attributes of acoustic waves at different angles and frequencies to the saturation prediction, and select the optimal dispersion attributes according to the contribution to the saturation prediction;

[0030] The saturation prediction module is used to input the preferred dispersion attribute into the neural network model trained according to the set target to perform saturation prediction.

[0031] In the above technical solution, post-stack seismic data and well logging data are two different data types, which record different information respectively. Post-stack seismic data is based on time recording, which mainly records the time required for sound waves to reflect back after encountering different rock layer interfaces during underground propagation; well logging data is depth domain data, which records the physical properties at different depths after drilling, such as sound wave velocity, resistivity and saturation.

[0032] In the above technical solution, the pre-stack gather data records the characteristic changes of sound waves propagating underground and being reflected back after encountering the reflection interface at different incident angles. Therefore, the dispersion properties of sound waves at different angles and frequencies can be extracted from the pre-stack seismic data.

[0033] In the above technical solutions, the logging data are all data belonging to the depth domain, such as natural gamma curve (GR), natural potential curve (SP), caliper curve (CAL), deep lateral resistivity curve (DLL), medium lateral resistivity curve (MLL), shallow lateral resistivity curve (SLL), acoustic time difference curve (AC), compensated neutron curve (CNL), compensated density curve (DEN); seismic related data are all data belonging to the time domain, such as post-stack seismic data, pre-stack gather data, etc.

[0034] In the above technical solution, through the collaborative work of various modules in the oil and gas saturation artificial intelligence prediction system, data from different sources are effectively integrated and processed, and artificial intelligence technology is used to improve the accuracy and efficiency of oil and gas saturation prediction, thereby overcoming the limitations of traditional methods.

[0035] In the above technical solution, the specific method for generating synthetic seismic records based on well logging data and post-stack seismic data is:

[0036] First, the reflection coefficient of the formation interface is calculated based on the post-stack seismic data:

[0037]

[0038] Among them, r i Expressed as the reflection coefficient of the i-th layer, Z i+1 Expressed as the longitudinal wave impedance of the i+1th layer, Z i It is expressed as the longitudinal wave impedance of the i-th layer;

[0039] In the processing of logging data and post-stack seismic data, the longitudinal wave impedance reflectivity r i is the relative change of the impedance of adjacent formations, where the specific calculation formula for calculating the longitudinal wave impedance Zp is:

[0040] Zp=Vp*ρ

[0041] Among them, Vp represents the longitudinal wave velocity, ρ represents the rock density;

[0042] The specific formula for the final synthetic seismic record is:

[0043] d(t)=w(t)*r(t)+n(t)

[0044] Where d(t) is the synthetic seismic record, w(t) is the seismic wavelet, r(t) is the reflection coefficient, and n(t) is the noise.

[0045] In the above technical solution, the seismic wavelet of the synthetic seismic record is extracted based on the post-stack seismic data.

[0046] In the above technical solution, the saturation curve in the logging data is converted from the depth domain to the time domain by generating synthetic seismic records, so that it can be accurately matched with the actual seismic data. This can provide a solid foundation for subsequent attribute optimization and neural network prediction, ensure the accuracy and consistency of the data, and thus improve the accuracy and reliability of oil and gas saturation prediction.

[0047] In the above technical solution, the specific method of comparing the actual seismic data with the synthetic seismic record to obtain the corresponding relationship between the propagation time of the seismic wave and the logging depth is:

[0048] The logging curve is obtained according to the logging data, and the initial time-depth relationship is constructed through the acoustic time difference data in the logging curve. This initial relationship is based on the logging data, which converts the logging curve data in the depth domain into data in the time domain. However, since the reflection coefficients of the logging data and the seismic data may not match completely, this initial time-depth relationship may not be accurate enough. In order to further calibrate this time-depth relationship, we need to compare the synthetic seismic records with the actual seismic data, analyze the relationship between the synthetic seismic records and the actual seismic data, and make fine adjustments to improve the accuracy of the time-depth conversion and obtain a more accurate correspondence between the seismic wave propagation time and the logging depth.

[0049] In the above technical solution, the specific adjustment method for analyzing the relationship between the synthetic seismic records and the actual seismic data and making fine adjustments is: adjusting the time-depth relationship by the degree of correlation between the synthetic seismic records and the actual seismic data. When the correlation reaches the highest, it proves that the well-seismic matching is completed and the time-depth relationship adjustment is completed.

[0050] In the above technical solution, the corresponding relationship between the propagation time of the seismic wave and the logging depth is the time-depth relationship.

[0051] In the above technical solution, by constructing the initial time-depth relationship and combining the comparative analysis of synthetic seismic records with actual data, the saturation curve in the depth domain can be more accurately converted into time domain data, thereby improving the matching degree and availability of the data.

[0052] In the above technical solution, the specific method for obtaining dispersion properties at different angles and frequencies based on pre-stack gather data is:

[0053] The P-wave velocity dispersion is calculated using pre-stack gather data. The specific formula is:

[0054]

[0055] Among them, D vp (θ, F) is the dispersion degree of longitudinal wave velocity, which is the dispersion attribute and can be used as a hydrocarbon detection parameter. G is the AVO (Amplitude Variation with Offset) gradient, which characterizes the gradient of the longitudinal wave dispersion with the offset, θ is the incident angle of the seismic wave when it hits the formation interface, F is the seismic wave frequency, F0 is the main frequency of the seismic wave, ΔF is the difference between the seismic wave frequency and the main frequency of the seismic wave, and Rp represents the reflection coefficient.

[0056] In the above technical scheme, by describing the specific method of obtaining dispersion properties at different angles and frequencies based on prestack gather data, data support is provided for the optimization of dispersion properties; this method can make full use of prestack gather data, calculate the dispersion properties under different conditions, enrich the feature set that can be used for analysis, and help to more comprehensively understand the characteristics of underground reservoirs.

[0057] In the above technical solution, the dispersion attribute is analyzed based on the SHAP method and the saturation curve belonging to the time domain data, and the specific analysis method for obtaining the contribution of the dispersion attribute to the saturation prediction is:

[0058] Set the angle to θ1 and the frequency to F1 as the sample X θ1F1 , angle is θk, frequency is Fk, sample is X θkFk , the predicted saturation of the model is y i , the baseline of the entire model is y base , then the SHAP value obeys the following equation:

[0059] y i =y base +f(X θ1F1 )+f(X θ1F2 )+f(X θ1F3 )……+f(X θkFk )

[0060] Where f(X θkFk ) is X θkFk SHAP value, f(X θ1F1 ) is the final predicted value y of the first dispersion attribute in the sample i The contribution value of f(X θ1F1 )>0, it means that the feature improves the prediction value and has a positive contribution to the model; otherwise, it means that the feature reduces the prediction value and has a negative contribution to the model. The SHAP method is used to match the saturation in the time domain and analyze the contribution of dispersion properties and saturation at different angles, such as Figure 3 shown.

[0061] In the above technical solution, the data input into the SHAP method interpretation model is a saturation curve belonging to time domain data.

[0062] In the above technical scheme, the dispersion attributes are analyzed based on the SHAP method and the time domain saturation curve to obtain the specific method of contribution, which not only realizes the effective screening of dispersion attributes, but also utilizes the interpretability of the SHAP method to quantify the contribution of each dispersion attribute to the saturation prediction, thereby quickly and accurately screening out attributes that are highly correlated with saturation, reducing data dimensions, and improving the efficiency and accuracy of the model.

[0063] In the above technical solution, a set number of preferred dispersion attributes are screened out according to the contribution of saturation prediction, wherein the specific screening requirement is: selecting a set number of dispersion attributes with the highest contribution value and positive contribution.

[0064] In the above technical scheme, specific requirements for screening and selecting dispersion attributes based on the contribution of saturation prediction are stipulated, which can ensure the quality of the screened attributes and select the dispersion attributes with the highest contribution value and positive contribution to the model, which can further optimize the data input into the neural network model, improve the prediction performance of the model, and avoid interference of irrelevant or low-contribution attributes on the model.

[0065] In the above technical solution, the specific prediction method of the neural network model for saturation prediction is:

[0066] The neural network consists of multiple layers of neurons, including input layer, hidden layer and output layer. The neural network structure diagram is as follows Figure 4 As shown, where:

[0067] The input layer is responsible for receiving the optimal dispersion properties of different angles and frequencies. Let the input vector be x. The specific formula is:

[0068] x=[D vp (θ1,F1),D vp (θ1,F2),D vp (θ1,F3)…D vp (θk,Fk)]

[0069] The hidden layer is responsible for extracting features and transforming the input vector x. The specific formula is:

[0070] h=f(W (1) x+b (1) )

[0071] Among them, W (1) is the weight matrix of the hidden layer, b (1) is the bias vector of the hidden layer, h is the output of the hidden layer, and f is the activation function;

[0072] The output layer is responsible for outputting the saturation prediction results, and its specific formula is:

[0073] y=g(W (2) h+b (2) )

[0074] Among them, g is the output layer activation function, W (2) is the weight matrix of the output layer, b (2) is the bias vector of the output layer, and y is the saturation prediction result.

[0075] In the above technical scheme, the specific prediction method for saturation prediction using the neural network model is described in detail, and the construction and operation process of the neural network model is clarified, so that the model can effectively learn and fit the complex relationship between the saturation curve and the dispersion properties, and realize efficient and accurate prediction of the three-dimensional data of the entire area.

[0076] In the above technical solution, in the neural network model trained according to the set target, the training target of the neural network is to minimize the loss function, and the neural network training method adopted is:

[0077] The saturation label is obtained by calculating the logging curve, and then the loss function value between the saturation prediction result and the saturation label is calculated according to the cross entropy loss function. The gradient of the loss function to each parameter in the neural network is calculated according to the back propagation algorithm, and then the gradient descent algorithm is used to update the corresponding parameters in the neural network. Finally, the corresponding parameters in the neural network are updated through repeated iterations until the minimized loss function is obtained.

[0078] In the above technical solution, by repeatedly iterating and updating parameters, the neural network will be continuously optimized, and finally the efficient and accurate saturation prediction of the post-optimization dispersion data can be realized. The comparison between the post-stack seismic profile and the conventional saturation modeling profile and the saturation prediction result diagram of the present invention is as follows: Figure 5 , Figure 6 and Figure 7 shown.

[0079] In the above technical solution, the method for training the neural network is not unique, and it only needs to achieve the goal of minimizing the loss function.

[0080] In the above technical scheme, the specific method of training the neural network model according to the set target ensures the optimization and effectiveness of the model; by minimizing the loss function and using an appropriate optimization algorithm, the neural network model can continuously adjust parameters during the training process to improve the fitting ability of the training data, thereby obtaining more accurate saturation prediction results in practical applications.

[0081] In the above technical solution, a three-dimensional saturation body is obtained according to the saturation prediction result, which is usually used in the process of oil and gas exploration and development to display the fluid saturation changes at different locations in the underground reservoir in a visual and quantitative manner.

[0082] Example 2

[0083] An artificial intelligence prediction method for oil and gas saturation based on dispersion attributes, such as Figure 2 As shown, synthetic seismic records are obtained based on logging data and post-stack seismic data, and the actual seismic data are compared with the synthetic seismic records to obtain the time-depth relationship. The saturation curve belonging to the time domain data is generated based on the time-depth relationship and the saturation curve belonging to the depth domain data; the dispersion properties of the sound waves at different angles and frequencies are obtained based on the pre-stack gather data, and the dispersion properties of the sound waves at different angles and frequencies are analyzed based on the SHAP method and the saturation curve belonging to the time domain data to screen out the preferred dispersion properties; the preferred dispersion properties are input into the neural network model trained according to the set target to predict the saturation and obtain a three-dimensional saturated body.

[0084] An artificial intelligence prediction method for oil and gas saturation comprises the following steps:

[0085] Collect well logging data, post-stack seismic data, pre-stack gather data and actual seismic data, and generate a saturation curve belonging to the depth domain data based on the collected well logging data;

[0086] A synthetic seismic record is obtained according to the well logging data and the post-stack seismic data, and the actual seismic data is compared with the synthetic seismic record to obtain the corresponding relationship between the time of seismic wave propagation and the well logging depth, and a saturation curve belonging to the time domain data is generated according to the corresponding relationship between the time of seismic wave propagation and the well logging depth and the saturation curve belonging to the depth domain data;

[0087] The dispersion properties of acoustic waves at different angles and frequencies are obtained based on pre-stack gather data. The dispersion properties of acoustic waves at different angles and frequencies are analyzed based on the SHAP method and the saturation curves belonging to time domain data. The contribution of the dispersion properties of acoustic waves at different angles and frequencies to saturation prediction is obtained, and the preferred dispersion properties are selected according to the contribution to saturation prediction.

[0088] The preferred dispersion properties are input into the neural network model trained according to the set target to predict the saturation.

[0089] Example 3

[0090] A computer program product comprises a computer program, wherein when the computer program is executed by a processor, the steps of the method described in Embodiment 2 are implemented.

[0091] The contents not described in detail in this specification belong to the prior art known to professional and technical personnel in this field.

[0092] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0093] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A system that specifies the functions of a box or boxes.

[0094] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a product including an instruction system, which is implemented in the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0095] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0096] 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 its protection scope. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that after reading the present invention, those skilled in the art can still make various changes, modifications or equivalent substitutions to the specific implementation methods of the invention, but these changes, modifications or equivalent substitutions are all within the protection scope of the pending claims of the invention.

Claims

1. An artificial intelligence prediction system for oil and gas saturation based on dispersion attributes, characterized in that: It includes: The data acquisition module is used to collect well logging data, post-stack seismic data, pre-stack gather data and actual seismic data, and generate a saturation curve belonging to the depth domain data based on the collected well logging data; The time-depth calibration module is used to obtain synthetic seismic records based on well logging data and post-stack seismic data, compare the actual seismic data with the synthetic seismic records, obtain the corresponding relationship between the time of seismic wave propagation and the well logging depth, and generate the saturation curve belonging to the time domain data based on the corresponding relationship between the time of seismic wave propagation and the well logging depth and the saturation curve belonging to the depth domain data; The dispersion attribute optimization module is used to obtain the dispersion attributes of acoustic waves at different angles and frequencies based on pre-stack gather data, analyze the dispersion attributes of acoustic waves at different angles and frequencies based on the SHAP method and the saturation curve belonging to the time domain data, obtain the contribution of the dispersion attributes of acoustic waves at different angles and frequencies to the saturation prediction, and select the optimal dispersion attributes according to the contribution to the saturation prediction; The saturation prediction module is used to input the preferred dispersion attribute into the neural network model trained according to the set target to perform saturation prediction.

2. The oil and gas saturation artificial intelligence prediction system based on dispersion attributes according to claim 1 is characterized by: The specific method for generating synthetic seismic records based on well logging data and post-stack seismic data is as follows: First, the reflection coefficient of the formation interface is calculated based on the post-stack seismic data: Among them, r i Expressed as the reflection coefficient of the i-th layer, Z i+1 Expressed as the longitudinal wave impedance of the i+1th layer, Z i It is expressed as the longitudinal wave impedance of the i-th layer; In the processing of logging data and post-stack seismic data, the longitudinal wave impedance reflectivity r i is the relative change of the impedance of adjacent formations, where the specific calculation formula for calculating the longitudinal wave impedance Zp is: Zp=Vp*ρ Among them, Vp represents the longitudinal wave velocity, ρ represents the rock density; The specific formula for the final synthetic seismic record is: d(t)=w(t)*r(t)+n(t) Where d(t) is the synthetic seismic record, w(t) is the seismic wavelet of the synthetic seismic record, r(t) is the reflection coefficient of the synthetic seismic record, and n(t) is the noise of the synthetic seismic record.

3. The oil and gas saturation artificial intelligence prediction system based on dispersion attributes according to claim 1 is characterized by: The specific method of comparing the actual seismic data with the synthetic seismic records to obtain the corresponding relationship between the propagation time of seismic waves and the logging depth is as follows: The logging curve is obtained based on the logging data, and the initial time-depth relationship is constructed through the acoustic time difference data in the logging curve. The synthetic seismic record is then compared with the actual seismic data, and the relationship between the synthetic seismic record and the actual seismic data is analyzed and adjusted to obtain the corresponding relationship between the propagation time of the seismic wave and the logging depth.

4. The oil and gas saturation artificial intelligence prediction system based on dispersion attributes according to claim 1 is characterized by: The specific method for obtaining dispersion properties at different angles and frequencies based on pre-stack gather data is: The P-wave velocity dispersion is calculated using pre-stack gather data. The specific formula is: Among them, D vp (θ, F) is the dispersion degree of longitudinal wave velocity, which is the dispersion attribute and can be used as a hydrocarbon detection parameter. G is the AVO gradient, which characterizes the gradient of the variation of the longitudinal wave dispersion with the offset, θ is the incident angle of the seismic wave when it hits the formation interface, F is the seismic wave frequency, F0 is the main frequency of the seismic wave, ΔF is the difference between the seismic wave frequency and the main frequency of the seismic wave, and Rp represents the reflection coefficient.

5. The oil and gas saturation artificial intelligence prediction system based on dispersion attributes according to claim 1 is characterized by: Based on the SHAP method and the saturation curve belonging to the time domain data, the dispersion attribute is analyzed, and the specific analysis method of the contribution of the dispersion attribute to the saturation prediction is obtained as follows: Set the angle to θ1 and the frequency to F1 as the sample X θ1F1 , angle is θk, frequency is Fk, sample is X θkFk , the predicted saturation of the model is y i , the baseline of the entire model is y base , then the SHAP value obeys the following equation: y i =y base +f(X θ1F1 )+f(X θ1F2 )+f(X θ1F3 )……+f(X θkFk ) Where f(X θkFk ) is X θkFk SHAP value, f(X θ1F1 ) is the final predicted value y of the first dispersion attribute in the sample i The contribution value of f(X θ1F1 )>0, it means that the feature improves the prediction value and has a positive contribution to the model; otherwise, it means that the feature reduces the prediction value and has a negative contribution to the model. The SHAP method is used to match the saturation in the time domain and analyze the contribution of dispersion properties and saturation at different angles.

6. The oil and gas saturation artificial intelligence prediction system based on dispersion attributes according to claim 1 is characterized by: A set number of preferred dispersion properties are screened out according to the contribution of saturation prediction, wherein the specific screening requirement is: selecting a set number of dispersion properties with the highest contribution value and positive contribution.

7. The oil and gas saturation artificial intelligence prediction system based on dispersion attributes according to claim 1 is characterized by: The specific prediction method for saturation prediction by the neural network model is: A neural network consists of multiple layers of neurons, including input layer, hidden layer and output layer, where: The input layer is responsible for receiving the optimal dispersion properties of different angles and frequencies. Let the input vector be x. The specific formula is: x=[D vp (θ1,F1),D vp (θ1,F2),D vp (θ1,F3)…D vp (θk,Fk)] The hidden layer is responsible for extracting features and transforming the input vector x. The specific formula is: h=f(W (1) x+b (1) ) Among them, W (1) is the weight matrix of the hidden layer, b (1) is the bias vector of the hidden layer, h is the output of the hidden layer, and f is the activation function; The output layer is responsible for outputting the saturation prediction results, and its specific formula is: y=g(W (2) h+b (2) ) Among them, g is the output layer activation function, W (2) is the weight matrix of the output layer, b (2) is the bias vector of the output layer, and y is the saturation prediction result.

8. The oil and gas saturation artificial intelligence prediction system based on dispersion attributes according to claim 7 is characterized by: In the neural network model trained according to the set target, the training target of the neural network is to minimize the loss function, and the neural network training method adopted is: The saturation label is obtained by calculating the logging curve, and then the loss function value between the saturation prediction result and the saturation label is calculated according to the cross entropy loss function. The gradient of the loss function to each parameter in the neural network is calculated according to the back propagation algorithm, and then the gradient descent algorithm is used to update the corresponding parameters in the neural network. Finally, the corresponding parameters in the neural network are updated through repeated iterations until the minimized loss function is obtained.

9. An artificial intelligence prediction method for oil and gas saturation based on dispersion attributes, characterized in that: It includes the following steps: Collect well logging data, post-stack seismic data, pre-stack gather data and actual seismic data, and generate a saturation curve belonging to the depth domain data based on the collected well logging data; A synthetic seismic record is obtained according to the well logging data and the post-stack seismic data, and the actual seismic data is compared with the synthetic seismic record to obtain the corresponding relationship between the time of seismic wave propagation and the well logging depth, and a saturation curve belonging to the time domain data is generated according to the corresponding relationship between the time of seismic wave propagation and the well logging depth and the saturation curve belonging to the depth domain data; The dispersion properties of acoustic waves at different angles and frequencies are obtained based on pre-stack gather data. The dispersion properties of acoustic waves at different angles and frequencies are analyzed based on the SHAP method and the saturation curves belonging to time domain data. The contribution of the dispersion properties of acoustic waves at different angles and frequencies to saturation prediction is obtained, and the preferred dispersion properties are selected according to the contribution to saturation prediction. The preferred dispersion properties are input into the neural network model trained according to the set target to predict the saturation.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method described in claim 9 are implemented.

Images