Formation pore pressure prediction method, device and storage medium based on deep learning

Abstract

The present invention relates to a formation pore pressure prediction method, device and storage medium based on deep learning, which are applied to the technical field of formation pore fluid pressure prediction, including: using a small amount of measured formation pore fluid pressure data, using geophysical logging parameter perturbation and Eaton's formula to form a large number of disturbed geophysical logging parameters and pressure parameters, and combining them with a small amount of measured formation pore fluid pressure data, finally forming a sample set composed of a large number of geophysical logging parameters and pressure parameters, and then performing deep learning based on a GRU network to finally complete the formation pore fluid pressure prediction. This method overcomes the difficulty that deep learning requires a large sample, and at the same time uses parameter perturbation to introduce the law represented by Eaton's formula into the data, so as to realize the formation pore fluid pressure prediction with high precision, and provide technical support for the application field of formation pore fluid pressure.

Description

Technical Field

[0001] The present invention relates to the technical field of formation pore fluid pressure prediction, and in particular to a formation pore pressure prediction method, device and storage medium based on deep learning. Background Art

[0002] Formation pore fluid pressure is an important geophysical parameter of underground rock formations. It is a key indicator for predicting reservoir parameters in oil and gas seismic exploration. It can guide the design of drilling engineering plans and reduce drilling engineering risks. It can also evaluate the quality of reservoirs in unconventional oil and gas reservoir development projects. High-precision formation pore fluid pressure prediction methods based on geophysical logging data play an important role.

[0003] At present, a variety of formation pore fluid pressure prediction methods have been developed, including the compaction balance equation method, the equivalent depth formula calculation method, the Eaton method, the Fillippone method, and so on. Among these methods, the Eaton method based on production experience summary has been widely used due to its advantages such as convenience in calculation. Although the empirical formula generally reflects the overall law of some formation measurement parameters and formation pore fluid pressure, the empirical summary method is usually the law summarized in the formations of a specific area. When applied to other regions, there are problems such as difficulty in determining empirical parameters and reduced accuracy. The main reason is that the formations in different regions have certain unique sedimentary environments, which are reflected in the formation pore fluid pressure, that is, data that cannot be predicted by the empirical formula. Deep learning is a successful data-driven method, but because the underground measured pore fluid pressure points are seriously insufficient due to construction and cost reasons, the application of deep learning to formation pore fluid pressure prediction has the difficulty of serious shortage of samples. Summary of the invention

[0004] In view of this, the purpose of the present invention is to provide a formation pore pressure prediction method, device and storage medium based on deep learning, so as to solve the problem of severe shortage of samples in the prior art when applying deep learning to formation pore fluid pressure prediction.

[0005] According to a first aspect of an embodiment of the present invention, a formation pore pressure prediction method based on deep learning is provided, the method comprising:

[0006] Obtain measured fluid pressure data of a small number of sparsely distributed formation pore fluid pressure measurement points in the target area, and obtain existing geophysical logging parameters in the target area;

[0007] Obtaining average empirical parameters based on the measured fluid pressure data of the small number of sparsely distributed formation pore fluid pressure measurement points and geophysical logging parameters;

[0008] Obtain the disturbance parameter of the zth measured pressure point;

[0009] Obtaining the combined pore fluid pressure data of the zth measured pressure point according to the disturbance parameter of the zth measured pressure point; respectively obtaining the combined pore fluid pressure data of a small number of sparsely distributed formation pore fluid pressure measured points, and combining the measured fluid pressure data of the small number of sparsely distributed formation pore fluid pressure measured points to obtain a massive formation pore fluid pressure data set;

[0010] Building a GRU neural network model framework, determining the output parameters of the GRU neural network model framework according to the average empirical parameters and the disturbance parameters, and determining the input parameters of the GRU neural network model framework according to the massive formation pore fluid pressure data set;

[0011] Based on the input parameters and the output parameters, the GRU neural network model architecture is trained through the massive formation pore fluid pressure data set to obtain a formation pore pressure prediction model;

[0012] The formation pore fluid pressure of the target formation pressure point to be predicted is outputted through the formation pore pressure prediction model.

[0013] Preferably,

[0014] The method of obtaining measured fluid pressure data of a small number of sparsely distributed formation pore fluid pressure measurement points in the target area includes:

[0015] Obtain the corresponding fluid pressure, corresponding density and corresponding P-wave velocity of a small number of sparsely distributed formation pore fluid pressure measurement points in the target area;

[0016] The method of obtaining the average empirical parameters based on the measured fluid pressure data of the small number of sparsely distributed formation pore fluid pressure measurement points and geophysical logging parameters includes:

[0017] Obtain existing geophysical logging parameters in the target area, and obtain the overlying formation pressure and hydrostatic pressure of the zth measured pressure point among the small number of sparsely distributed formation pore fluid pressure measurement points according to the existing geophysical logging parameters in the target area, and obtain the P-wave velocity under normal compaction corresponding to the zth measured pressure point;

[0018] The empirical parameters of the zth measured pressure point are obtained according to the corresponding fluid pressure, corresponding P-wave velocity, overlying formation pressure, hydrostatic pressure and P-wave velocity under normal compaction of the zth measured pressure point;

[0019] The empirical parameters of a small number of sparsely distributed formation pore fluid pressure measurement points are obtained respectively, and the average empirical parameters are obtained according to the empirical parameters of a small number of sparsely distributed formation pore fluid pressure measurement points.

[0020] Preferably,

[0021] The step of obtaining the longitudinal wave velocity under normal compaction corresponding to the zth measured pressure point includes:

[0022] Based on the geophysical logging parameters, a normal compaction trend line is obtained by using a least squares fitting method, and a P-wave velocity under normal compaction corresponding to the z-th measured pressure point is obtained according to the normal compaction trend line;

[0023] The method of obtaining the empirical parameters of the zth measured pressure point includes:

[0024] Based on the corresponding fluid pressure, corresponding P-wave velocity, overlying formation pressure, hydrostatic pressure and P-wave velocity under normal compaction of the z-th measured pressure point, the empirical parameters of the z-th measured pressure point are back-calculated based on Eaton's formula.

[0025] Preferably,

[0026] The method of obtaining the disturbance parameter of the zth measured pressure point includes:

[0027] Longitudinally disturbing the overlying formation pressure at the z-th measured pressure point to obtain the disturbed overlying formation pressure at the z-th measured pressure point;

[0028] Perturbing the corresponding density of the zth measured pressure point within its cardinality range to obtain the perturbed density of the zth measured pressure point;

[0029] perturbing the longitudinal wave velocity corresponding to the z-th measured pressure point within the range of its cardinality to obtain the perturbed longitudinal wave velocity of the z-th measured pressure point;

[0030] The disturbed hydrostatic pressure of the zth measured pressure point is obtained according to the disturbed overlying formation pressure, the corresponding density and the disturbed density of the zth measured pressure point.

[0031] Preferably,

[0032] The step of obtaining the combined pore fluid pressure data of the zth measured pressure point according to the disturbance parameter of the zth measured pressure point comprises:

[0033] The disturbed overlying formation pressure, disturbed density and disturbed P-wave velocity of the z-th measured pressure point are linearly combined with the overlying formation pressure, the corresponding density and the corresponding P-wave velocity respectively to obtain a combined overlying formation pressure set, a combined density set and a combined P-wave velocity set. The combined overlying formation pressure set, the combined density set and the combined P-wave velocity set are traversed to obtain the combined pore fluid pressure data of the z-th measured pressure point.

[0034] Preferably,

[0035] Determining the output parameters of the GRU neural network model architecture according to the average empirical parameters and the disturbance parameters includes:

[0036] Substituting the disturbed overlying formation pressure, the disturbed hydrostatic pressure, the disturbed P-wave velocity, the P-wave velocity under normal compaction and the average empirical parameters at the z-th measured pressure point into the Eaton formula to obtain the disturbed formation pore fluid pressure at the z-th measured pressure point;

[0037] Obtaining a combined formation pore fluid pressure at the zth measured pressure point according to the disturbed formation pore pressure set at the zth measured pressure point and the corresponding fluid pressure at the zth measured pressure point;

[0038] The combined formation pore fluid pressure and the corresponding fluid pressure at the zth measured pressure point are used as output parameters of the GRU neural network model architecture.

[0039] Preferably,

[0040] The input parameters of the GRU neural network model architecture are determined according to the massive formation pore fluid pressure data set, including:

[0041] The combined overlying formation pressure and the overlying formation pressure corresponding to the measured point, the combined P-wave velocity and the corresponding P-wave velocity of the measured point, and the combined density and the corresponding density of the measured point in the massive formation pore fluid pressure data set are used as input parameters of the GRU neural network model architecture.

[0042] According to a second aspect of an embodiment of the present invention, a formation pore pressure prediction device based on deep learning is provided, the device comprising:

[0043] Measured parameter acquisition module: used to obtain measured fluid pressure data of a small number of sparsely distributed formation pore fluid pressure measurement points in the target area, and obtain existing geophysical logging parameters in the target area;

[0044] Average empirical parameter acquisition module: used for acquiring average empirical parameters according to the measured fluid pressure data of a small number of sparsely distributed formation pore fluid pressure measurement points and geophysical logging parameters;

[0045] Disturbance parameter acquisition module: used to obtain the disturbance parameter of the zth measured pressure point;

[0046] Data expansion module: used for obtaining the combined pore fluid pressure data of the zth measured pressure point according to the disturbance parameter of the zth measured pressure point; respectively obtaining the combined pore fluid pressure data of a small number of sparsely distributed formation pore fluid pressure measured points, and combining the measured fluid pressure data of the small number of sparsely distributed formation pore fluid pressure measured points to obtain a massive formation pore fluid pressure data set;

[0047] Model parameter determination module: used to build a GRU neural network model framework, determine the output parameters of the GRU neural network model framework according to the average empirical parameters and the disturbance parameters, and determine the input parameters of the GRU neural network model framework according to the massive formation pore fluid pressure data set;

[0048] Model training module: used to train the GRU neural network model architecture based on the input parameters and output parameters through the massive formation pore fluid pressure data set to obtain a formation pore pressure prediction model;

[0049] Prediction output module: used to output the formation pore fluid pressure of the target formation pressure point to be predicted through the formation pore pressure prediction model.

[0050] According to a third aspect of an embodiment of the present invention, there is provided a storage medium, wherein the storage medium stores a computer program, and when the computer program is executed by a main controller, each step in the above method is implemented.

[0051] The technical solution provided by the embodiments of the present invention may have the following beneficial effects:

[0052] This application uses a small amount of measured formation pore fluid pressure data, uses geophysical logging parameter perturbations and Eaton's formula to form a large number of disturbed geophysical logging parameters and pressure parameters, and combines them with a small amount of measured formation pore fluid pressure data to ultimately form a sample set consisting of a large number of geophysical logging parameters and pressure parameters, and then performs deep learning based on a GRU network to ultimately complete the prediction of formation pore fluid pressure. This method overcomes the difficulty that deep learning requires large samples, and at the same time uses parameter perturbations to introduce the laws represented by Eaton's formula into the data, which can achieve high-precision prediction of formation pore fluid pressure, providing technical support for the application field of formation pore fluid pressure.

[0053] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0055] Figure 1 is a flow chart of a formation pore pressure prediction method based on deep learning according to an exemplary embodiment;

[0056] Figure 2 is a schematic diagram of formation pore pressure prediction results according to another exemplary embodiment;

[0057] Figure 3 is a system schematic diagram of a formation pore pressure prediction device based on deep learning according to another exemplary embodiment;

[0058] In the attached figure: 1-measured parameter acquisition module, 2-average empirical parameter acquisition module, 3-disturbance parameter acquisition module, 4-data expansion module, 5-model parameter determination module, 6-model training module, 7-prediction output module. DETAILED DESCRIPTION

[0059] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0060] Embodiment 1

[0061] Figure 1 is a flow chart of a formation pore pressure prediction method based on deep learning according to an exemplary embodiment. Figure 1 As shown, the method includes:

[0062] S1, obtaining measured fluid pressure data of a small number of sparsely distributed formation pore fluid pressure measurement points in the target area, and obtaining existing geophysical logging parameters in the target area;

[0063] S2, obtaining average empirical parameters based on the measured fluid pressure data of the small number of sparsely distributed formation pore fluid pressure measurement points and geophysical logging parameters;

[0064] S3, obtaining the disturbance parameter of the zth measured pressure point;

[0065] S4, obtaining the combined pore fluid pressure data of the zth measured pressure point according to the disturbance parameter of the zth measured pressure point; respectively obtaining the combined pore fluid pressure data of a small number of sparsely distributed formation pore fluid pressure measured points, and combining the measured fluid pressure data of the small number of sparsely distributed formation pore fluid pressure measured points to obtain a massive formation pore fluid pressure data set;

[0066] S5, building a GRU neural network model framework, determining the output parameters of the GRU neural network model framework according to the average empirical parameters and the disturbance parameters, and determining the input parameters of the GRU neural network model framework according to the massive formation pore fluid pressure data set;

[0067] S6, based on the input parameters and the output parameters, the GRU neural network model architecture is trained by the massive formation pore fluid pressure data set to obtain a formation pore pressure prediction model;

[0068] S7, outputting the formation pore fluid pressure of the target formation pressure point to be predicted through the formation pore pressure prediction model;

[0069] It can be understood that the specific implementation of this scheme is as follows:

[0070] (I) Geophysical logging parameters:

[0071] Obtain measured fluid pressure data of a small number of sparsely distributed formation pore fluid pressure measurement points in the target area. Specifically, obtain the measured formation pore fluid pressure P of 12 formation pore fluid pressure measurement points. p and the density ρ in continuous geophysical logging parameters t and the longitudinal wave velocity V p , this embodiment defines the second formation pore pressure measuring point as the zth formation pore pressure measuring point, and the subsequent implementation steps are expressed based on the zth formation pore pressure measuring point;

[0072] Obtain the existing geophysical logging parameters in the target area, and obtain the overlying formation pressure P at the zth measured point based on the existing geophysical logging parameters. ov (z) and hydrostatic pressure P w (z), the formula is as follows:

[0073] P ov (z) = 0.0098 (ρ av H0+ρ b (H(z)-H0))

[0074] P w (z) = ρ w gH(z)

[0075] Where: av is the average density of the overlying strata (existing geophysical logging parameters), ρb is the density of the overlying strata (existing geophysical logging parameters), H0 is the starting depth of the target layer (existing geophysical logging parameters), ρ w is the density of water (usually 1g / cm 3 ), g is the gravitational acceleration (usually 9.8), H(z) is the depth of the target layer (existing geophysical logging parameters);

[0076] Similarly, based on the geophysical logging parameters, the normal compaction trend line is obtained by the least squares fitting method, so as to obtain the P-wave velocity V under normal compaction corresponding to the zth measured pressure point. n (z);

[0077] (ii) Disturbance parameters:

[0078] (1) For the zth measured point, the overlying formation pressure P ov (z) Perform vertical (upward and downward) disturbance to obtain the disturbed overlying formation pressure ΔP corresponding to the measured point z ov (z,y ov ), the formula is as follows:

[0079] ΔP ov (z,y ov )=ω ov y ov P ov (z)

[0080] In the formula, y ov is the overlying formation pressure disturbance variable, an integer ranging from -N ov ≤y ov ≤N ov , N ov is the number of overlying formation pressure disturbances, which is 6; ω ov is the percentage coefficient of the overlying formation pressure at the measured point z, and its range is 0<ω ov <0.3%;

[0081] (2) The measured density is averaged in the target layer, and under the constraints of the measured density and the average density, the measured density ρ corresponding to the measured point z is calculated. t (z) is perturbed within the range of its cardinality to obtain the perturbation density Δρ corresponding to the measured point z t (z,y t ), the formula is as follows:

[0082] Δρ t (z,y t )=ω t y t ρ t (z)

[0083] In the formula, y t is the density perturbation variable, an integer whose range is -N t ≤y t ≤N t , N t is the number of density perturbations, which is 6; ω t is the density percentage coefficient of the measured point z, and its range is 0<ω t <10%;

[0084] (3) The average P-wave velocity is calculated in the target layer, and under the constraints of the measured P-wave velocity and the average P-wave velocity, the measured P-wave velocity V corresponding to the measured point z is calculated. p (z) is disturbed within the range of its cardinality, and the perturbed longitudinal wave velocity ΔV corresponding to point z is obtained p (z,y p ), the formula is as follows:

[0085] ΔV p (z,y p )=ω p y p V p (z)

[0086] In the formula, y p is the disturbance variable of the longitudinal wave velocity, an integer whose range is -N p ≤y p ≤N p , N p is the number of longitudinal wave velocity disturbances, which is 6; ω p is the percentage coefficient of the longitudinal wave velocity at the measured point z, and its range is 0<ω p <12%;

[0087] (4) According to the calculation formula of overlying formation pressure and hydrostatic pressure, the disturbed overlying formation pressure ΔPov(z,yov) corresponding to the measured point z is used to calculate the disturbed hydrostatic pressure ΔP corresponding to the measured point z w (z,y w ), the formula is as follows:

[0088]

[0089] ΔP w (z,y w )=ρ w gΔH(z,y w )

[0090] In the formula, ΔH(z,y w ) is the disturbance depth corresponding to point z, y w is the hydrostatic pressure disturbance variable;

[0091] The disturbed overburden pressure, disturbed density, disturbed P-wave velocity and the overburden pressure, density and P-wave velocity at the measured point z are linearly combined to obtain the combined overburden pressure Combination density and the combined longitudinal wave velocity The formula is as follows:

[0092]

[0093] The combined overlying formation pressure set corresponding to the measured point z Combined density set and the combined longitudinal wave velocity set ΔV p (z,y p ) are 2×6, 2×6 and 2×6 respectively; the combination sets of these three parameters are traversed to obtain 1728 combined pore fluid pressure data of 2×6×2×6×2×6;

[0094] Among the 12 measured pressure points of the input target layer, 5 measured point data are extracted from top to bottom according to depth for disturbance, which are the 2nd point, the 5th point, the 8th point, the 10th point and the 12th point from shallow to deep respectively. The range of the extracted measured points should roughly cover the entire target layer. Since each pore fluid pressure measured point can obtain a combined pore fluid pressure data set of 2×6×2×6×2×6 after disturbance, the 5 pore fluid pressure measured points and their disturbance can obtain a combined pore fluid pressure data set of 5×2×6×2×6×2×6. Adding the 5 pore fluid pressure measured point data can obtain a total of 8645 groups of massive formation pore fluid pressure data sets of 5×(2×6×2×6×2×6+1);

[0095] (III) Using a large set of formation pore fluid pressure data to train the GRU network:

[0096] (1) Dataset division: The massive formation pore fluid pressure dataset obtained above is used as the dataset for training the GRU network, with 80% of the dataset divided into the training set and the remaining 20% ​​into the validation set;

[0097] (2) Determine the network input and output parameters: The combined overburden pressure in the massive formation pore fluid pressure data set And the overlying formation pressure P corresponding to the measured point ov , combined longitudinal wave velocity And the corresponding longitudinal wave velocity V of the measured point p , combined density And the density corresponding to the measured point ρ t As an input parameter of the network, the combined formation pore fluid pressure and measured formation pore fluid pressure P p As the output parameter of the network,

[0098] Among them, the combined formation pore fluid pressure The method of obtaining is as follows:

[0099] ① Based on Eaton's formula, the empirical parameter C(z) of the zth measured point is calculated. The formula is as follows:

[0100]

[0101] Where P ov (z) is the overlying formation pressure at the zth measured point obtained above; P p (z) is the measured formation pore fluid pressure at the zth measured point, P w (z) is the hydrostatic pressure at the zth measured point obtained above, V n (z) is the longitudinal wave velocity under normal compaction corresponding to the zth measured pressure point obtained above, V p (z) is the measured longitudinal wave velocity of the zth measured point obtained above;

[0102] The average empirical parameter C is obtained by averaging the empirical parameters of the 12 pressure measurement points. av , eliminate the impact of outliers;

[0103] ② The disturbed overlying formation pressure ΔP corresponding to the measured point z obtained above is ov (z,y ov ), disturbed hydrostatic pressure ΔP w (z,y w ), disturbance longitudinal wave velocity ΔV p (z,y p ), longitudinal wave velocity V under normal compaction n (z) and the average empirical parameter C av Substituting into Eaton's formula, we can get the disturbed formation pore pressure set ΔP corresponding to the measured point z: p (z,y ov ,y w ,y p ), the formula is as follows:

[0104]

[0105] Combining the sparsely distributed formation pore fluid pressure data points with the above disturbed formation pore fluid pressure set to construct a combined formation pore fluid pressure set The formula is as follows:

[0106]

[0107] Where P p (z) is the measured formation pore fluid pressure at the zth measured point;

[0108] (3) Preprocessing the above divided data sets, normalizing the data using the "Max-Min" method, taking into account the dimensional impact of different geophysical logging parameters, and normalizing the data to a range of 0-1;

[0109] (4) Determine the structure and hyperparameters of the GRU neural network; the constructed GRU network model includes 5 GRU layers and 1 fully connected layer, with the number of neurons being 128, 64, 32, 16, 8 and 1 respectively; the batch size is 64, that is, 64 groups of data are randomly selected for network training each time; the number of iterations is set to 50; the learning rate is set to 0.006; the activation function is the Tanh function; the Adam optimization algorithm is used; the loss function adopts the MSE function;

[0110] (5) Use the normalized data to train the network. The specific training process is as follows:

[0111] ①Perform forward propagation calculation;

[0112] ② Compare the model output with the actual value and reversely calculate the error value, including two directions of back propagation by time and network level;

[0113] ③ Calculate the gradient of each weight according to the corresponding error term;

[0114] ④Use the Adam optimization algorithm to update the weights so that the predicted value gradually approaches the measured value;

[0115] It is worth emphasizing that the model training process is a relatively mature technology in the prior art, including the GRU neural network structure is also a conventional model in the prior art. Therefore, this embodiment is only an exemplary description of the model training process, and should not be used as a limitation to the present application;

[0116] (IV) Using the trained GRU network to predict the continuously distributed formation pore fluid pressure:

[0117] Use the trained network model to predict the test data, use the orthogonal experiment method to adjust various hyperparameters to the optimal level, output a set of prediction results with the highest accuracy, and denormalize the test results to obtain the corresponding prediction values;

[0118] The intermediate parameters of the GRU network for prediction are as follows:

[0119]

[0120] b cal =δ(W b [P cal-1 ,x cal ])

[0121] r cal =δ(W r [P cal-1 ,x cal ])

[0122]

[0123] Among them, x ca l represents the current input, including the combined overburden pressure The overlying formation pressure P corresponding to the measured point ov , combined longitudinal wave velocity The measured point corresponds to the longitudinal wave velocity V p , combined density The density corresponding to the measured point ρ t ; b cal represents the update gate state, δ is the sigmoid function, W b represents the update gate weight matrix, P cal-1 is the output of the network at the previous moment, r cal Indicates the reset gate state, W r Represents resetting the gate weight matrix, represents the sum of the input and past hidden states, The weight matrix representing the current specific output;

[0124] The model finally outputs the predicted formation pore fluid pressure P at the target pressure point. cal :

[0125]

[0126] In addition, this embodiment can also output the formation pore fluid pressure P according to the model cal Calculate the formation pore fluid pressure coefficient Ω cal , as attached Figure 2 As shown, the formula is as follows:

[0127]

[0128] This embodiment first uses a small amount of measured formation pore fluid pressure data, uses geophysical logging parameter perturbation and Eaton's formula to form a large number of perturbed geophysical logging parameters and pressure parameters, and combines them with a small amount of measured formation pore fluid pressure data to finally form a sample set composed of a large number of geophysical logging parameters and pressure parameters, and then performs deep learning based on the GRU network to finally complete the formation pore fluid pressure prediction. This method overcomes the difficulty of deep learning requiring large samples, and at the same time uses parameter perturbation to introduce the law represented by Eaton's formula into the data, which can achieve high-precision prediction of formation pore fluid pressure and provide technical support for the application field of formation pore fluid pressure.

[0129] Embodiment 2:

[0130] Figure 3 is a system schematic diagram of a formation pore pressure prediction device based on deep learning according to another exemplary embodiment, the device comprising:

[0131] Measured parameter acquisition module 1: used to obtain measured fluid pressure data of a small number of sparsely distributed formation pore fluid pressure measurement points in the target area, and obtain existing geophysical logging parameters in the target area;

[0132] Average empirical parameter acquisition module 2: used to acquire average empirical parameters based on the measured fluid pressure data of a small number of sparsely distributed formation pore fluid pressure measurement points and geophysical logging parameters;

[0133] Disturbance parameter acquisition module 3: used to obtain the disturbance parameter of the zth measured pressure point;

[0134] Data expansion module 4: used to obtain the combined pore fluid pressure data of the zth measured pressure point according to the disturbance parameter of the zth measured pressure point; respectively obtain the combined pore fluid pressure data of a small number of sparsely distributed formation pore fluid pressure measured points, and combine the measured fluid pressure data of the small number of sparsely distributed formation pore fluid pressure measured points to obtain a massive formation pore fluid pressure data set;

[0135] Model parameter determination module 5: used to build a GRU neural network model framework, determine the output parameters of the GRU neural network model framework according to the average empirical parameters and the disturbance parameters, and determine the input parameters of the GRU neural network model framework according to the massive formation pore fluid pressure data set;

[0136] Model training module 6: used to train the GRU neural network model architecture based on the input parameters and output parameters through the massive formation pore fluid pressure data set to obtain a formation pore pressure prediction model;

[0137] Prediction output module 7: used to output the formation pore fluid pressure of the target pressure point to be predicted through the formation pore pressure prediction model.

[0138] Embodiment three:

[0139] This embodiment provides a storage medium, wherein the storage medium stores a computer program, and when the computer program is executed by a main controller, each step in the above method is implemented;

[0140] It is understandable that the storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc.

[0141] It can be understood that the same or similar parts of the above embodiments can be referenced to each other, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.

[0142] It should be noted that, in the description of the present invention, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of the present invention, unless otherwise specified, the meaning of "a small number of sparsely distributed" refers to at least two.

[0143] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code that includes one or less sparsely distributed executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present invention belong.

[0144] It should be understood that the various parts of the present invention can be implemented in hardware, software, firmware or a combination thereof. In the above-mentioned embodiment, a small number of sparsely distributed steps or methods can be implemented in software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0145] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0146] In addition, each functional unit in each embodiment of the present invention may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0147] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc.

[0148] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or a small number of sparsely distributed embodiments or examples in a suitable manner.

[0149] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A formation pore pressure prediction method based on deep learning, characterized in that: The method comprises: Obtain measured fluid pressure data of a small number of sparsely distributed formation pore fluid pressure measurement points in the target area, and obtain existing geophysical logging parameters in the target area; Obtaining average empirical parameters based on the measured fluid pressure data of the small number of sparsely distributed formation pore fluid pressure measurement points and geophysical logging parameters; The obtaining of measured fluid pressure data of a small number of sparsely distributed formation pore fluid pressure measurement points in the target area includes: Obtain the corresponding fluid pressure, corresponding density and corresponding P-wave velocity of a small number of sparsely distributed formation pore fluid pressure measurement points in the target area; The method of obtaining the average empirical parameters based on the measured fluid pressure data of the small number of sparsely distributed formation pore fluid pressure measurement points and geophysical logging parameters includes: Obtain existing geophysical logging parameters in the target area, and obtain the overlying formation pressure and hydrostatic pressure of the zth measured pressure point among the small number of sparsely distributed formation pore fluid pressure measurement points according to the existing geophysical logging parameters in the target area, and obtain the P-wave velocity under normal compaction corresponding to the zth measured pressure point; The empirical parameters of the zth measured pressure point are obtained according to the corresponding fluid pressure, the corresponding P-wave velocity, the overlying formation pressure, the hydrostatic pressure and the P-wave velocity under normal compaction at the zth measured pressure point; respectively obtaining empirical parameters of a small number of sparsely distributed formation pore fluid pressure measurement points, and obtaining average empirical parameters based on the empirical parameters of a small number of sparsely distributed formation pore fluid pressure measurement points; Obtain the disturbance parameter of the zth measured pressure point; The method of obtaining the disturbance parameter of the zth measured pressure point includes: Longitudinally disturbing the overlying formation pressure at the z-th measured pressure point to obtain the disturbed overlying formation pressure at the z-th measured pressure point; Perturbing the corresponding density of the zth measured pressure point within its cardinality range to obtain the perturbed density of the zth measured pressure point; perturbing the longitudinal wave velocity corresponding to the z-th measured pressure point within the range of its cardinality to obtain the perturbed longitudinal wave velocity of the z-th measured pressure point; Obtaining the disturbed hydrostatic pressure of the zth measured pressure point according to the disturbed overlying formation pressure, the corresponding density and the disturbed density of the zth measured pressure point; Acquire the combined pore fluid pressure data of the zth measured pressure point according to the disturbance parameter of the zth measured pressure point; The step of obtaining the combined pore fluid pressure data of the zth measured pressure point according to the disturbance parameter of the zth measured pressure point comprises: Linearly combining the disturbed overburden formation pressure, disturbed density and disturbed P-wave velocity of the z-th measured pressure point with the overburden formation pressure, corresponding density and corresponding P-wave velocity, respectively, to obtain a combined overburden formation pressure set, a combined density set and a combined P-wave velocity set, and traversing the combined overburden formation pressure set, the combined density set and the combined P-wave velocity set to obtain the combined pore fluid pressure data of the z-th measured pressure point; Respectively obtaining combined pore fluid pressure data of a small number of sparsely distributed formation pore fluid pressure measurement points, and combining the measured fluid pressure data of the small number of sparsely distributed formation pore fluid pressure measurement points to obtain a massive formation pore fluid pressure data set; Building a GRU neural network model framework, determining the output parameters of the GRU neural network model framework according to the average empirical parameters and the disturbance parameters, and determining the input parameters of the GRU neural network model framework according to the massive formation pore fluid pressure data set; Determining the output parameters of the GRU neural network model architecture according to the average empirical parameters and the disturbance parameters includes: Substituting the disturbed overlying formation pressure, the disturbed hydrostatic pressure, the disturbed P-wave velocity, the P-wave velocity under normal compaction and the average empirical parameters at the z-th measured pressure point into the Eaton formula to obtain the disturbed formation pore fluid pressure at the z-th measured pressure point; Obtaining a combined formation pore fluid pressure at the zth measured pressure point according to the disturbed formation pore pressure set at the zth measured pressure point and the corresponding fluid pressure at the zth measured pressure point; The combined formation pore fluid pressure at the z-th measured pressure point and the measured formation pore fluid pressure are used as output parameters of the GRU neural network model architecture; Based on the input parameters and the output parameters, the GRU neural network model architecture is trained through the massive formation pore fluid pressure data set to obtain a formation pore pressure prediction model; The formation pore fluid pressure of the target formation pressure point to be predicted is outputted through the formation pore pressure prediction model.

2. The method according to claim 1, characterized in that The step of obtaining the longitudinal wave velocity under normal compaction corresponding to the zth measured pressure point includes: Based on the geophysical logging parameters, a normal compaction trend line is obtained by using a least squares fitting method, and a P-wave velocity under normal compaction corresponding to the z-th measured pressure point is obtained according to the normal compaction trend line; The method of obtaining the empirical parameters of the zth measured pressure point includes: Based on the corresponding fluid pressure, corresponding P-wave velocity, overlying formation pressure, hydrostatic pressure and P-wave velocity under normal compaction of the z-th measured pressure point, the empirical parameters of the z-th measured pressure point are back-calculated based on Eaton's formula.

3. The method according to claim 2, characterized in that The input parameters of the GRU neural network model architecture are determined according to the massive formation pore fluid pressure data set, including: The combined overlying formation pressure and the overlying formation pressure corresponding to the measured point, the combined P-wave velocity and the corresponding P-wave velocity of the measured point, and the combined density and the corresponding density of the measured point in the massive formation pore fluid pressure data set are used as input parameters of the GRU neural network model architecture.

4. A formation pore pressure prediction device based on deep learning, characterized in that: The device comprises: Measured parameter acquisition module: used to obtain measured fluid pressure data of a small number of sparsely distributed formation pore fluid pressure measurement points in the target area, and obtain existing geophysical logging parameters in the target area; Average empirical parameter acquisition module: used for acquiring average empirical parameters according to the measured fluid pressure data of a small number of sparsely distributed formation pore fluid pressure measurement points and geophysical logging parameters; The obtaining of measured fluid pressure data of a small number of sparsely distributed formation pore fluid pressure measurement points in the target area includes: Obtain the corresponding fluid pressure, corresponding density and corresponding P-wave velocity of a small number of sparsely distributed formation pore fluid pressure measurement points in the target area; The method of obtaining the average empirical parameters based on the measured fluid pressure data of the small number of sparsely distributed formation pore fluid pressure measurement points and geophysical logging parameters includes: Obtain existing geophysical logging parameters in the target area, and obtain the overlying formation pressure and hydrostatic pressure of the zth measured pressure point among the small number of sparsely distributed formation pore fluid pressure measurement points according to the existing geophysical logging parameters in the target area, and obtain the P-wave velocity under normal compaction corresponding to the zth measured pressure point; The empirical parameters of the zth measured pressure point are obtained according to the corresponding fluid pressure, the corresponding P-wave velocity, the overlying formation pressure, the hydrostatic pressure and the P-wave velocity under normal compaction at the zth measured pressure point; respectively obtaining empirical parameters of a small number of sparsely distributed formation pore fluid pressure measurement points, and obtaining average empirical parameters based on the empirical parameters of a small number of sparsely distributed formation pore fluid pressure measurement points; Disturbance parameter acquisition module: used to obtain the disturbance parameter of the zth measured pressure point; The method of obtaining the disturbance parameter of the zth measured pressure point includes: Longitudinally disturbing the overlying formation pressure at the z-th measured pressure point to obtain the disturbed overlying formation pressure at the z-th measured pressure point; Perturbing the corresponding density of the zth measured pressure point within its cardinality range to obtain the perturbed density of the zth measured pressure point; perturbing the longitudinal wave velocity corresponding to the z-th measured pressure point within the range of its cardinality to obtain the perturbed longitudinal wave velocity of the z-th measured pressure point; Obtaining the disturbed hydrostatic pressure of the zth measured pressure point according to the disturbed overlying formation pressure, the corresponding density and the disturbed density of the zth measured pressure point; Data expansion module: used for obtaining the combined pore fluid pressure data of the zth measured pressure point according to the disturbance parameter of the zth measured pressure point; respectively obtaining the combined pore fluid pressure data of a small number of sparsely distributed formation pore fluid pressure measured points, and combining the measured fluid pressure data of the small number of sparsely distributed formation pore fluid pressure measured points to obtain a massive formation pore fluid pressure data set; The step of obtaining the combined pore fluid pressure data of the zth measured pressure point according to the disturbance parameter of the zth measured pressure point comprises: Linearly combining the disturbed overburden formation pressure, disturbed density and disturbed P-wave velocity of the z-th measured pressure point with the overburden formation pressure, corresponding density and corresponding P-wave velocity, respectively, to obtain a combined overburden formation pressure set, a combined density set and a combined P-wave velocity set, and traversing the combined overburden formation pressure set, the combined density set and the combined P-wave velocity set to obtain the combined pore fluid pressure data of the z-th measured pressure point; Model parameter determination module: used to build a GRU neural network model framework, determine the output parameters of the GRU neural network model framework according to the average empirical parameters and the disturbance parameters, and determine the input parameters of the GRU neural network model framework according to the massive formation pore fluid pressure data set; Determining the output parameters of the GRU neural network model architecture according to the average empirical parameters and the disturbance parameters includes: Substituting the disturbed overlying formation pressure, the disturbed hydrostatic pressure, the disturbed P-wave velocity, the P-wave velocity under normal compaction and the average empirical parameters at the z-th measured pressure point into the Eaton formula to obtain the disturbed formation pore fluid pressure at the z-th measured pressure point; Obtaining a combined formation pore fluid pressure at the zth measured pressure point according to the disturbed formation pore pressure set at the zth measured pressure point and the corresponding fluid pressure at the zth measured pressure point; The combined formation pore fluid pressure at the z-th measured pressure point and the measured formation pore fluid pressure are used as output parameters of the GRU neural network model architecture; Model training module: used to train the GRU neural network model architecture based on the input parameters and output parameters through the massive formation pore fluid pressure data set to obtain a formation pore pressure prediction model; Prediction output module: used to output the formation pore fluid pressure of the target formation pressure point to be predicted through the formation pore pressure prediction model.

5. A storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by the main controller, each step of the formation pore pressure prediction method based on deep learning as described in any one of claims 1 to 3 is implemented.

Images