Rock gas-bearing prediction method, device, equipment and medium

Abstract

The invention discloses a rock gas-bearing prediction method, device, equipment and medium, and relates to the technical field of gas reservoir saturation distribution and reserve evaluation, and the method comprises the steps: obtaining initial seismic wave data; performing time-frequency transformation on the initial seismic wave data to obtain a time-frequency diagram, and obtaining a quality factor calculation function expression representing a relationship between a time-frequency attribute and a quality factor based on the time-frequency diagram; the time-frequency attribute comprises an amplitude attribute, a frequency attribute and a time attribute; and constructing a KAN network for calculating the gas saturation based on the quality factor calculation function expression, and training the KAN network by using the target amplitude, the target frequency and the target gas saturation obtained at the target well point to obtain a corrected network comprising the corrected quality factor calculation function expression so as to facilitate the corrected network to predict the gas saturation of the rock. According to the method, the quality factor calculation function expression is corrected through the training process of the KAN network, and the gas content of the rock can be accurately and reliably predicted.

Description

Technical Field

[0001] The invention relates to the technical field of gas reservoir saturation distribution and reserve evaluation, and in particular to a method, device, equipment and medium for predicting gas content in rocks. Background Art

[0002] In the field of oil and gas exploration and production, accurate evaluation of the fluid properties of underground rock formations is the key to improving the success rate of resource exploration and development efficiency. Among them, using seismic data to predict the gas content of formations has become an important research direction. The Q value (quality factor) is a physical quantity that measures the energy attenuation of seismic waves in a medium. It is closely related to the physical properties of rocks and the properties of the fluids therein, and is therefore widely studied for evaluating the gas content of formations.

[0003] However, existing Q-value prediction methods have certain limitations. First, these methods are relatively fixed in form, usually assuming that the attenuation of seismic waves is mainly affected by medium absorption, ignoring the impact of geological structure complexity on attenuation. In addition, existing Q-value calculation formulas often lack flexibility and cannot adapt to the high heterogeneity and changes in formation properties. This results in limited prediction accuracy in practical applications even in theoretically feasible cases. For example, when seismic waves pass through aquifers and gas-bearing layers, it may be difficult to accurately distinguish between the two situations relying solely on traditional Q-value analysis, because the attenuation mechanism is not solely determined by gas content. In addition, factors such as the propagation path of seismic waves, frequency-dependent attenuation characteristics, and the pore structure of rocks may have a significant impact on the Q value, but these factors are often not fully considered in traditional models.

[0004] In addition, the commonly used mineral analysis methods are machine learning or acoustic wave attenuation calculation to regress and calculate the gas content value of the formation. Among them, the machine learning method mainly uses technologies such as random forests to establish a mapping relationship between seismic data and logging response results to realize the calculation of the gas content of the entire reservoir. The acoustic wave attenuation method mainly completes the calculation of gas content indicators by performing a coincidence operation on the attenuation degree of shear waves and longitudinal waves.

[0005] In summary, for gas content prediction, the physical method based on formulas is not accurate enough for predicting gas content, and the gas content prediction method based on machine learning lacks interpretability. First, the results of the gas content prediction method based on acoustic attenuation will be affected by many factors such as formation lithology, physical properties, and fluid properties. In the process of multi-well result correction and fitting, it is impossible to guarantee the complete assimilation and efficient use of existing data, and some well points may not be able to fit, resulting in inaccurate prediction results. Second, the machine learning method has a strong data dependence and does not consider the internal physical driving willingness. Although it can fully fit the observation data represented by well logging, there may be obvious overfitting, and the obtained calculation model cannot be verified for reliability due to black box reasons.

[0006] In summary, how to accurately and reliably predict the gas content of rocks is an urgent problem to be solved. Summary of the invention

[0007] In view of this, the purpose of the present invention is to provide a method, device, equipment and medium for predicting the gas content of rocks, which can accurately and reliably predict the gas content of rocks. The specific scheme is as follows:

[0008] In a first aspect, the present application discloses a method for predicting gas content of rocks, comprising:

[0009] Acquiring initial seismic wave data;

[0010] Performing time-frequency transformation on the initial seismic wave data to obtain a time-frequency graph, and obtaining a quality factor calculation function expression representing the relationship between time-frequency attributes and quality factors based on the time-frequency graph; the time-frequency attributes include amplitude attributes, frequency attributes and time attributes;

[0011] A KAN network is constructed for calculating gas saturation based on the quality factor calculation function expression, and the target amplitude, target frequency and target gas saturation obtained at the target well point are used to train the KAN network to obtain a corrected network including a corrected quality factor calculation function expression, so as to predict rock gas saturation using the corrected network.

[0012] The quality factor calculation function expression that characterizes the relationship between the time-frequency attribute and the quality factor based on the time-frequency diagram includes:

[0013] Extracting the amplitude values ​​corresponding to the first time and the second time at a specific frequency from the time-frequency diagram;

[0014] Extracting the amplitude of the time-domain seismic signal of the target layer corresponding to a specific frequency from the time-frequency diagram; the specific frequency is a frequency determined based on a representative frequency among low-frequency frequencies, a representative frequency among medium-frequency frequencies, and a representative frequency among high-frequency frequencies;

[0015] Determine a quality factor value between each two adjacent time points at the specific frequency according to the amplitude value, and determine a total quality factor value at the specific frequency based on all the quality factor values;

[0016] A quality factor calculation function expression characterizing the relationship between the time-frequency attribute and the quality factor is determined according to the total quality factor value, the specific frequency and the amplitude value.

[0017] Wherein, determining the quality factor value between every two adjacent time points at the specific frequency according to the amplitude value includes:

[0018] Calculate the amplitude logarithmic decay rate between every two adjacent time points at the specific frequency according to the amplitude value;

[0019] Using a target quality factor calculation formula, and according to the amplitude logarithmic decay rate and the specific frequency, calculating a quality factor value between every two adjacent time points at the specific frequency;

[0020] The target quality factor calculation formula is:

[0021] ;

[0022] in, represents the logarithmic decay rate of the amplitude; represents said specific frequency; express and The difference between Indicates the amplitude value; represents the total quality factor value.

[0023] Before performing time-frequency transformation on the initial seismic wave data to obtain a time-frequency graph, the method further includes:

[0024] The seismic data is preprocessed; the preprocessing includes denoising and / or filtering and / or amplitude compensation and / or time difference correction.

[0025] The step of performing time-frequency transformation on the initial seismic wave data to obtain a time-frequency graph includes:

[0026] The initial seismic wave data is subjected to time-frequency transformation by short-time Fourier transform or continuous wavelet transform to obtain a time-frequency diagram.

[0027] The quality factor calculation function expression that characterizes the relationship between the time-frequency attribute and the quality factor based on the time-frequency diagram includes:

[0028] Based on the analysis window corresponding to the seismic wave reflected by the tight sandstone reservoir in the time-frequency diagram, a quality factor calculation function expression representing the relationship between the time-frequency attribute and the quality factor is obtained.

[0029] Wherein, the target amplitude, target frequency and target gas saturation obtained at the target well point are used to train the KAN network to obtain a corrected network including a corrected quality factor calculation function expression, including:

[0030] The mean square error value is used as the loss calculation function, the Adam algorithm is used as the gradient calculation and updating method, and the target amplitude, target frequency and target gas saturation obtained at the target well point are used to train the KAN network to obtain a corrected network including a corrected quality factor calculation function expression.

[0031] In a second aspect, the present application discloses a rock gas content prediction device, comprising:

[0032] A data acquisition module, used for acquiring initial seismic wave data;

[0033] A function expression determination module is used to perform time-frequency transformation on the initial seismic wave data to obtain a time-frequency graph, and obtain a quality factor calculation function expression representing the relationship between the time-frequency attribute and the quality factor based on the time-frequency graph; the time-frequency attribute includes an amplitude attribute, a frequency attribute and a time attribute;

[0034] A network construction module, used for constructing a KAN network for calculating gas saturation based on the quality factor calculation function expression;

[0035] The correction module is used to train the KAN network using the target amplitude, target frequency and target gas saturation obtained at the target well point to obtain a corrected network including a corrected quality factor calculation function expression, so as to predict the rock gas saturation using the corrected network.

[0036] In a third aspect, the present application discloses an electronic device, comprising:

[0037] Memory, used to store computer programs;

[0038] A processor is used to execute the computer program to implement the rock gas content prediction method disclosed above.

[0039] In a fourth aspect, the present application discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, the aforementioned rock gas content prediction method disclosed above is implemented.

[0040] It can be seen that the present application obtains initial seismic wave data; performs time-frequency transformation on the initial seismic wave data to obtain a time-frequency diagram, and obtains a quality factor calculation function expression that characterizes the relationship between time-frequency attributes and quality factors based on the time-frequency diagram; the time-frequency attributes include amplitude attributes, frequency attributes, and time attributes; constructs a KAN network that calculates gas saturation based on the quality factor calculation function expression, and uses the target amplitude, target frequency, and target gas saturation obtained at the target well point to train the KAN network to obtain a corrected network including a corrected quality factor calculation function expression, so as to use the corrected network to predict rock gas saturation. It can be seen that the present application proposes to correct the quality factor calculation function expression through the training process of the KAN network, so that the machine loss of the instruction factor is more accurate, and further the training process of the KAN network also makes the calculation process of gas saturation more accurate; the present application uses the corrected network including the corrected quality factor calculation function expression for prediction, and the corrected quality factor calculation function expression reflects physicality and has reliability; in summary, the present application can accurately and reliably predict rock gas content. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0042] Figure 1 A flow chart of a method for predicting gas content of rocks disclosed in this application;

[0043] Figure 2 A flow chart of a specific rock gas content prediction method disclosed in this application;

[0044] Figure 3 A neuron schematic diagram of a KAN network disclosed in this application;

[0045] Figure 4 A schematic diagram of calculation results of a gas saturation attribute disclosed in this application;

[0046] Figure 5 This is a schematic diagram of the structure of a rock gas content prediction device disclosed in this application;

[0047] Figure 6 This is a structural diagram of an electronic device disclosed in this application. DETAILED DESCRIPTION

[0048] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] In summary, for gas content prediction, the physical method based on formulas is not accurate enough for predicting gas content, and the gas content prediction method based on machine learning lacks interpretability. First, the results of the gas content prediction method based on acoustic attenuation will be affected by many factors such as formation lithology, physical properties, and fluid properties. In the process of multi-well result correction and fitting, it is impossible to guarantee the complete assimilation and efficient use of existing data, and some well points may not be able to fit, resulting in inaccurate prediction results. Second, the machine learning method has a strong data dependence and does not consider the internal physical driving willingness. Although it can fully fit the observation data represented by well logging, there may be obvious overfitting, and the obtained calculation model cannot be verified for reliability due to black box reasons.

[0050] To this end, an embodiment of the present application proposes a rock gas content prediction scheme that can accurately and reliably predict the rock gas content.

[0051] The present application discloses a method for predicting the gas content of rocks. Figure 1 As shown, the method includes:

[0052] Step S11: Acquire initial seismic wave data.

[0053] In this embodiment, the initial seismic wave data should cover the entire area of ​​the target reservoir, including its overlying and underlying strata, so as to conduct a comprehensive analysis. The selected seismic records should have sufficient resolution and signal-to-noise ratio to ensure that the extracted seismic attributes can truly reflect the underground geological conditions. The source of the data can be regional three-dimensional seismic exploration data, or two-dimensional seismic profiles specific to certain well locations.

[0054] Step S12: performing time-frequency transformation on the initial seismic wave data to obtain a time-frequency diagram, and obtaining a quality factor calculation function expression characterizing the relationship between time-frequency attributes and quality factors based on the time-frequency diagram; the time-frequency attributes include amplitude attributes, frequency attributes and time attributes.

[0055] In this embodiment, after the seismic data is selected, preprocessing is performed immediately, which is a key step to ensure data quality, including denoising, filtering and data correction. Specifically, before the time-frequency transformation of the initial seismic wave data to obtain the time-frequency diagram, it also includes: preprocessing the seismic data; the preprocessing includes denoising and / or filtering and / or amplitude compensation and / or time difference correction.

[0056] It should be pointed out that denoising is a very important part of preprocessing, which aims to eliminate background noise and interference in the data, such as random noise or noise caused by surface conditions. This can be achieved through a variety of techniques, such as spectrum analysis, Wiener filtering or adaptive noise suppression technology. Filtering is to remove unnecessary frequency components in the seismic signal through bandpass filters or high-pass / low-pass filters, thereby highlighting signals related to reservoir characteristics. In addition, preprocessing may also include steps such as amplitude compensation and time difference correction to ensure the consistency and comparability of seismic data in vertical and horizontal directions. These preprocessing steps are crucial for subsequent time-frequency analysis and Q value calculation, because only high-quality input data can ensure the accuracy and reliability of the estimation results. After these meticulous preprocessing, the seismic data will provide a solid foundation for the subsequent Q value estimation.

[0057] In this embodiment, the time-frequency transformation of seismic data is performed to convert the signal from the time domain to the time-frequency domain, which enables us to analyze the behavioral characteristics of seismic waves at different times and frequencies. This step is crucial because it provides a detailed view of the change of amplitude over time and frequency, so that the Q value of the formation can be estimated more accurately; specifically, the time-frequency transformation of the initial seismic wave data to obtain the time-frequency diagram includes: using short-time Fourier transform or continuous wavelet transform to transform the initial seismic wave data to obtain the time-frequency diagram.

[0058] It should be pointed out that the short-time Fourier transform (STFT) obtains time-localized frequency information by sliding a fixed-size window function on the signal and performing Fourier transform on the signal in the window. This method can balance the resolution of time and frequency to a certain extent, and is suitable for analyzing the changes of frequency components over time. The continuous wavelet transform (CWT) uses a series of mother wavelets obtained by scaling and translation to analyze the signal, which is particularly suitable for capturing non-fixed frequency features in the signal. It provides the advantage of variable resolution, resulting in higher frequency resolution at low frequencies and higher time resolution at high frequencies.

[0059] In this embodiment, after completing the time-frequency transformation, the next task is to select a suitable analysis window. In the time-frequency diagram, it is necessary to identify and select those time windows and frequency bandwidths that represent the seismic waves reflected from the tight sandstone reservoir. When selecting these windows, it is necessary to ensure that sufficient signal energy is included, while also avoiding including too much irrelevant noise or signals from non-target formations. The selection of these windows is crucial for subsequent Q value estimation, because the wrong window may lead to inaccurate analysis results. Typically, the signal-to-noise ratio (SNR) of each candidate time window and frequency bandwidth is evaluated, and those windows with high signal-to-noise ratios are selected. Specifically, the quality factor calculation function expression that characterizes the relationship between the time-frequency attribute and the quality factor based on the time-frequency diagram includes: based on the analysis window corresponding to the seismic wave reflected by the tight sandstone reservoir in the time-frequency diagram, a quality factor calculation function expression that characterizes the relationship between the time-frequency attribute and the quality factor is obtained.

[0060] Step S13: construct a KAN network for calculating gas saturation based on the quality factor calculation function expression, and train the KAN network using the target amplitude, target frequency and target gas saturation obtained at the target well point to obtain a corrected network including a corrected quality factor calculation function expression, so as to predict rock gas saturation using the corrected network.

[0061] In this embodiment, the target amplitude, target frequency and target gas saturation obtained at the target well point are used to train the KAN network to obtain a corrected network including a corrected quality factor calculation function expression, including: using the mean square error value as the loss calculation function, using the Adam algorithm as the gradient calculation and update method, and using the target amplitude, target frequency and target gas saturation obtained at the target well point to train the KAN network to obtain a corrected network including a corrected quality factor calculation function expression.

[0062] In summary, this application has carried out the design and correction of the Q factor for gas content detection to make it more consistent with the well point interpretation results, so as to obtain more accurate gas saturation prediction results. Specifically, this application makes full use of the KA theory to carry out the inversion of the correction term formula in the Q factor calculation equation to complete the correction; in addition, it should be pointed out that during the KA network design process, this application must keep the original basic framework of Q function calculation unchanged.

[0063] It can be seen that the present application obtains initial seismic wave data; performs time-frequency transformation on the initial seismic wave data to obtain a time-frequency diagram, and obtains a quality factor calculation function expression that characterizes the relationship between time-frequency attributes and quality factors based on the time-frequency diagram; the time-frequency attributes include amplitude attributes, frequency attributes, and time attributes; constructs a KAN network that calculates gas saturation based on the quality factor calculation function expression, and uses the target amplitude, target frequency, and target gas saturation obtained at the target well point to train the KAN network to obtain a corrected network including a corrected quality factor calculation function expression, so as to use the corrected network to predict rock gas saturation. It can be seen that the present application proposes to correct the quality factor calculation function expression through the training process of the KAN network, so that the machine loss of the instruction factor is more accurate, and further the training process of the KAN network also makes the calculation process of gas saturation more accurate; the present application uses the corrected network including the corrected quality factor calculation function expression for prediction, and the corrected quality factor calculation function expression reflects physicality and has reliability; in summary, the present application can accurately and reliably predict rock gas content.

[0064] The embodiment of the present application discloses a specific method for predicting the gas content of rocks. Compared with the previous embodiment, this embodiment further explains and optimizes the technical solution. Figure 2 As shown, specifically including:

[0065] Step S21: Acquire initial seismic wave data.

[0066] Step S22: Perform time-frequency transformation on the initial seismic wave data to obtain a time-frequency diagram, and extract the amplitude of the time-domain seismic signal of the target layer corresponding to a specific frequency from the time-frequency diagram; the specific frequency is a frequency determined based on a representative frequency among low-frequency frequencies, a representative frequency among medium-frequency frequencies, and a representative frequency among high-frequency frequencies; determine the quality factor value between each two adjacent time points at the specific frequency according to the amplitude value, and determine the total quality factor value at the specific frequency based on all the quality factor values; determine a quality factor calculation function expression that characterizes the relationship between the time-frequency attribute and the quality factor according to the total quality factor value, the specific frequency, and the amplitude value; the time-frequency attributes include amplitude attributes, frequency attributes, and time attributes.

[0067] In this embodiment, the quality factor calculation function expression characterizing the relationship between the time-frequency attribute and the quality factor based on the time-frequency diagram includes: obtaining the quality factor calculation function expression characterizing the relationship between the time-frequency attribute and the quality factor based on the analysis window corresponding to the seismic wave reflected by the tight sandstone reservoir in the time-frequency diagram.

[0068] In this embodiment, determining the quality factor value between each two adjacent time points at the specific frequency according to the amplitude value includes: calculating the amplitude logarithmic decay rate between each two adjacent time points at the specific frequency according to the amplitude value; using a target quality factor calculation formula, and calculating the quality factor value between each two adjacent time points at the specific frequency according to the amplitude logarithmic decay rate and the specific frequency; wherein the target quality factor calculation formula is:

[0069] ;

[0070] in, represents the logarithmic decay rate of the amplitude; represents said specific frequency; express and The difference between Indicates the amplitude value; represents the total quality factor value.

[0071] It should be pointed out that after completing the time-frequency transformation and selecting the appropriate analysis window, the next task is to extract the amplitude value of the seismic wave from these windows. Select one representative frequency at the low frequency, medium frequency and high frequency respectively (the specific low frequency, medium frequency and high frequency ranges need to be set according to the actual situation, and the representative frequency is also set according to the actual situation) to fully capture the characteristics of seismic waves of different scales. These three frequencies are used as features to form a specific frequency f. In the time dimension, the amplitudes of all time points are counted, and the logarithmic decay rate of the amplitude between each two adjacent time points is calculated. Specifically, it is necessary to determine the amplitude of each two adjacent time points. and Amplitude and Afterwards, the logarithmic attenuation rate of the amplitude between each two adjacent time points is calculated based on the amplitude at each two adjacent time points. This attenuation rate is a key factor in the calculation of the Q value because it is directly related to the degree of energy attenuation when the seismic wave propagates in the formation.

[0072] Step S23: construct a KAN network for calculating gas saturation based on the quality factor calculation function expression, and train the KAN network using the target amplitude, target frequency and target gas saturation obtained at the target well point to obtain a corrected network including a corrected quality factor calculation function expression, so as to predict rock gas saturation using the corrected network.

[0073] In this embodiment, the seismic signal is a superposition of multiple information, and it is necessary to add modeling of other interference sources on the basis of the Q function to better reflect the gas saturation signal; a deep learning network is constructed using Kolmogorov–Arnold, and the form of the Q function is embedded in it. On the one hand, the calculation process of the amplitude attribute in the Q function is modeled, and at the same time, the formation process of other interference factors is modeled. On the other hand, retaining the calculation process of the Q function can maximize the physical significance of the result and avoid the influence of overfitting. The gas saturation S attribute body calculation based on the calculation result of the Q function is realized.

[0074] It should be pointed out that the multi-layer Kolmogorov–Arnold network constructed according to the Kolmogorov–Arnold (KAN) principle is different from the neural network model in that each node of the network no longer adopts the form of a weight matrix, but a weight equation. At the same time, in order to ensure that the entire formula can always maintain the basic form of the Q function during the fitting process, we restrict and fix most of the weight equations so that the formula form will not change during the entire fitting process.

[0075] Then, for the part that needs to be fitted, use the equation The explicit gas saturation expression relationship is constructed in the form of and the network is used for learning; where G () represents the signal generation model of other factors affecting the seismic signal, E represents the P-wave and S-wave energy, Represents the density of rock and fluid, K represents the bulk modulus of rock. Therefore, this paper establishes a double-layer KAN network structure for the signal generation model G() and the A() function in the Q value calculation process, and performs regression representation in functional form through fitting. The neurons of the KAN network are represented in the form of specific basic equations, such as The neurons in the same layer are additive, and the functions across layers are nested. For example, the second layer is and the first layer is The nested representation is form.

[0076] See also Figure 3 The figure shows a neuron diagram of a KAN network. The figure includes three layers of neurons. The connection lines between each two layers of neurons contain a nested relationship. The transmission of neurons is completed through functions ( Figure 3 In the figure, there is a functional relationship between the neurons in the previous layer and a neuron in the current layer. Since it is a well-known content, it will not be introduced in detail here. It is reflected in the final function expression: the connection between the upper and lower layer neurons is expressed as a nested relationship of functions, and the neurons in the same layer are expressed as an additive relationship. The specific content will not be introduced in detail.

[0077] It should be pointed out that the correction process is as follows: the amplitude, frequency and other attributes of the well points in the seismic data body are used as input, the gas saturation obtained from the interpretation at the well points is used as the label data, the mean square error value is used as the loss calculation function, and the Adam algorithm is used as the gradient calculation and update method to carry out data-based network model training. The training process is completed by iterating 500 times, and the final function form is finally displayed. After the training is completed, some unimportant function parts will be ignored, and the entire fitting process will not change the basic structural form of the fixed part of the G function. By adding multi-layer equations and correction terms, the corrected G function is obtained as the final gas saturation calculation function, as shown in the following formula:

[0078] ;

[0079] ;

[0080] Where b and d are constant terms, is a function of a fixed expression that is fitted. The function subscript represents the function expression for a certain feature. For example, , and They all represent functions for the E feature. , and They all represent functions for K features. , and All of them are aimed at Function of features.

[0081] In summary, this application is based on the existing autoregressive correction method of the Q factor calculation formula. By correcting and improving the calculation process of the Q factor, a more robust and accurate gas saturation prediction formula (trained network) is proposed. Specifically, this application preferably selects the Q factor calculation equation, finds the structure with errors inside the formula, and optimizes the variables into functional expressions through the functionalization method to improve the fitting ability of the formula. Based on Kolmogorov–Arnold, through the verification of well seismic data at the well point, formula regression is carried out on the variables with high uncertainty in the Q factor calculation formula to improve the calculation formula of the Q factor; based on Kolmogorov–Arnold, on the basis of the Q factor calculation results, by adding correction terms and fitting specific formulas, the calculation accuracy of the gas saturation at the well point is improved, thereby obtaining the calculation formula of the gas saturation attribute body.

[0082] In a specific embodiment, when evaluating the gas saturation of a gas field, we first analyze and sort the pre-stack seismic data to extract the relevant amplitude attributes, frequency attributes and time attributes. Then we calculate the Q value and determine the Q function, which is the quality factor calculation function expression, and then establish the gas saturation calculation model. , and establish a KAN network structure corresponding to the Q function. According to the field experience time-depth relationship curve, the well trajectory of the drilled well is calibrated and projected into the seismic data. Then, the seismic attributes at the well trajectory are extracted along the well trajectory and used as input data. At the same time, the interpretation results of the well logging gas saturation on the well trajectory are extracted as label data. The data is input into the KAN network for training, and through continuous iterative learning, the corrected functions G and Q are finally obtained. The seismic gas saturation attributes are calculated using the corrected Q function. The results are as follows Figure 4 As shown, Figure 4 Figure 1 is a schematic diagram of the calculation results of the gas saturation attribute. The white dotted lines are the real gas-water interfaces of multiple small layers. It can be seen from the gas saturation attribute diagram that they are basically consistent with the actual results. Among them, P102-1, P103-1 and P103-2 represent the well numbers.

[0083] It can be seen that the present application obtains initial seismic wave data; performs time-frequency transformation on the initial seismic wave data to obtain a time-frequency diagram, and extracts the amplitude of the time-domain seismic signal of the target layer corresponding to a specific frequency from the time-frequency diagram; the specific frequency is a frequency determined based on a representative frequency in the low-frequency frequency, a representative frequency in the medium-frequency frequency, and a representative frequency in the high-frequency frequency; determines the quality factor value between each two adjacent time points at the specific frequency according to the amplitude value, and determines the total quality factor value at the specific frequency based on all the quality factor values; determines the quality factor calculation function expression characterizing the relationship between the time-frequency attribute and the quality factor according to the total quality factor value, the specific frequency, and the amplitude value; the time-frequency attribute includes amplitude attribute, frequency attribute, and time attribute; constructs a KAN network for calculating gas saturation based on the quality factor calculation function expression, and trains the KAN network using the target amplitude, target frequency, and target gas saturation obtained at the target well point to obtain a corrected network including a corrected quality factor calculation function expression, so as to predict the gas saturation of the rock using the corrected network. It can be seen that the present application discloses the specific content of the quality factor calculation function expression and the specific training process, and predicts the gas saturation of rocks through the trained network to improve the prediction accuracy and reliability.

[0084] Correspondingly, the present application also discloses a rock gas content prediction device, see Figure 5 As shown, the device comprises:

[0085] A data acquisition module 11 is used to acquire initial seismic wave data;

[0086] A function expression determination module 12 is used to perform time-frequency transformation on the initial seismic wave data to obtain a time-frequency graph, and obtain a quality factor calculation function expression representing the relationship between the time-frequency attribute and the quality factor based on the time-frequency graph; the time-frequency attribute includes an amplitude attribute, a frequency attribute and a time attribute;

[0087] A network construction module 13, used to construct a KAN network for calculating gas saturation based on the quality factor calculation function expression;

[0088] The correction module 14 is used to train the KAN network using the target amplitude, target frequency and target gas saturation obtained at the target well point to obtain a corrected network including a corrected quality factor calculation function expression, so as to predict the rock gas saturation using the corrected network.

[0089] Among them, the more specific working process of each of the above modules can refer to the corresponding content disclosed in the above embodiments, which will not be repeated here.

[0090] It can be seen that the present application obtains initial seismic wave data; performs time-frequency transformation on the initial seismic wave data to obtain a time-frequency diagram, and obtains a quality factor calculation function expression that characterizes the relationship between time-frequency attributes and quality factors based on the time-frequency diagram; the time-frequency attributes include amplitude attributes, frequency attributes, and time attributes; constructs a KAN network that calculates gas saturation based on the quality factor calculation function expression, and uses the target amplitude, target frequency, and target gas saturation obtained at the target well point to train the KAN network to obtain a corrected network including a corrected quality factor calculation function expression, so as to use the corrected network to predict rock gas saturation. It can be seen that the present application proposes to correct the quality factor calculation function expression through the training process of the KAN network, so that the machine loss of the instruction factor is more accurate, and further the training process of the KAN network also makes the calculation process of gas saturation more accurate; the present application uses the corrected network including the corrected quality factor calculation function expression for prediction, and the corrected quality factor calculation function expression reflects physicality and has reliability; in summary, the present application can accurately and reliably predict rock gas content.

[0091] Furthermore, an embodiment of the present application also provides an electronic device. Figure 6 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content in the diagram cannot be regarded as any limitation on the scope of use of the present application.

[0092] Figure 6A schematic diagram of the structure of an electronic device 20 provided in an embodiment of the present application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a display screen 23, an input / output interface 24, a communication interface 25, a power supply 26, and a communication bus 27. The memory 22 is used to store a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the rock gas content prediction method disclosed in any of the aforementioned embodiments. In addition, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0093] In this embodiment, the power supply 26 is used to provide working voltage for each hardware device on the electronic device 20; the communication interface 25 can create a data transmission channel between the electronic device 20 and the external device, and the communication protocol it follows is any communication protocol that can be applied to the technical solution of the present application, and is not specifically limited here; the input and output interface 24 is used to obtain external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs and is not specifically limited here.

[0094] In addition, the memory 22, as a carrier for storing resources, may be a read-only memory, a random access memory, a disk or an optical disk, etc., and the resources stored thereon may include a computer program 221, and the storage method may be temporary storage or permanent storage. In addition to the computer program that can be used to complete the rock gas content prediction method executed by the electronic device 20 disclosed in any of the aforementioned embodiments, the computer program 221 may further include a computer program that can be used to complete other specific tasks.

[0095] Furthermore, an embodiment of the present application also discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, the aforementioned rock gas content prediction method disclosed above is implemented.

[0096] The specific steps of the method can refer to the corresponding contents disclosed in the above embodiments, which will not be repeated here.

[0097] The various embodiments in this application are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0098] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0099] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0100] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used 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", "comprise" 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. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0101] The above is a detailed introduction to a rock gas content prediction method, device, equipment, and storage medium provided by the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea; at the same time, for general technical personnel in this field, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A method for predicting gas content of rocks, characterized in that: include: Acquiring initial seismic wave data; Performing time-frequency transformation on the initial seismic wave data to obtain a time-frequency graph, and obtaining a quality factor calculation function expression representing the relationship between time-frequency attributes and quality factors based on the time-frequency graph; the time-frequency attributes include amplitude attributes, frequency attributes and time attributes; A KAN network is constructed for calculating gas saturation based on the quality factor calculation function expression, and the target amplitude, target frequency and target gas saturation obtained at the target well point are used to train the KAN network to obtain a corrected network including a corrected quality factor calculation function expression, so as to predict rock gas saturation using the corrected network.

2. The method for predicting rock gas content according to claim 1, characterized in that: The quality factor calculation function expression characterizing the relationship between the time-frequency attribute and the quality factor is obtained based on the time-frequency graph, including: Extracting the amplitude of the time-domain seismic signal of the target layer corresponding to a specific frequency from the time-frequency diagram; the specific frequency is a frequency determined based on a representative frequency among low-frequency frequencies, a representative frequency among medium-frequency frequencies, and a representative frequency among high-frequency frequencies; Determine a quality factor value between each two adjacent time points at the specific frequency according to the amplitude value, and determine a total quality factor value at the specific frequency based on all the quality factor values; A quality factor calculation function expression characterizing the relationship between the time-frequency attribute and the quality factor is determined according to the total quality factor value, the specific frequency and the amplitude value.

3. The method for predicting rock gas content according to claim 2, characterized in that: The determining, according to the amplitude value, the quality factor value between each two adjacent time points at the specific frequency comprises: Calculate the amplitude logarithmic decay rate between every two adjacent time points at the specific frequency according to the amplitude value; Using a target quality factor calculation formula, and according to the amplitude logarithmic decay rate and the specific frequency, calculating a quality factor value between every two adjacent time points at the specific frequency; The target quality factor calculation formula is: ； in, represents the logarithmic decay rate of the amplitude; represents said specific frequency; express and The difference between Indicates the amplitude value; represents the quality factor value.

4. The method for predicting rock gas content according to claim 1, characterized in that: Before performing time-frequency transformation on the initial seismic wave data to obtain a time-frequency graph, the method further includes: The seismic data is preprocessed; the preprocessing includes denoising and / or filtering and / or amplitude compensation and / or time difference correction.

5. The method for predicting rock gas content according to claim 1, characterized in that: The step of performing time-frequency transformation on the initial seismic wave data to obtain a time-frequency graph comprises: The initial seismic wave data is subjected to time-frequency transformation by short-time Fourier transform or continuous wavelet transform to obtain a time-frequency diagram.

6. The method for predicting rock gas content according to claim 1, characterized in that: The quality factor calculation function expression characterizing the relationship between the time-frequency attribute and the quality factor is obtained based on the time-frequency graph, including: Based on the analysis window corresponding to the seismic wave reflected by the tight sandstone reservoir in the time-frequency diagram, a quality factor calculation function expression representing the relationship between the time-frequency attribute and the quality factor is obtained.

7. The method for predicting rock gas content according to any one of claims 1 to 6, characterized in that: The method of training the KAN network using the target amplitude, target frequency and target gas saturation obtained at the target well point to obtain a modified network including a modified quality factor calculation function expression includes: The mean square error value is used as the loss calculation function, the Adam algorithm is used as the gradient calculation and updating method, and the target amplitude, target frequency and target gas saturation obtained at the target well point are used to train the KAN network to obtain a corrected network including a corrected quality factor calculation function expression.

8. A rock gas content prediction device, characterized in that: include: A data acquisition module, used for acquiring initial seismic wave data; A function expression determination module is used to perform time-frequency transformation on the initial seismic wave data to obtain a time-frequency graph, and obtain a quality factor calculation function expression representing the relationship between the time-frequency attribute and the quality factor based on the time-frequency graph; the time-frequency attribute includes an amplitude attribute, a frequency attribute and a time attribute; A network construction module, used for constructing a KAN network for calculating gas saturation based on the quality factor calculation function expression; The correction module is used to train the KAN network using the target amplitude, target frequency and target gas saturation obtained at the target well point to obtain a corrected network including a corrected quality factor calculation function expression, so as to predict the rock gas saturation using the corrected network.

9. An electronic device, characterized in that: include: Memory, used to store computer programs; A processor is used to execute the computer program to implement the rock gas content prediction method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: Used to store computer programs; wherein, when the computer program is executed by a processor, the rock gas content prediction method according to any one of claims 1 to 7 is implemented.

Images