Density prediction method and device based on neural network, equipment and storage medium

Abstract

The application provides a neural network-based density prediction method and device, equipment and storage medium, including: data preparation: obtaining the depth domain measured depth domain P-wave velocity, depth domain S-wave velocity and density logging data of a study area; sample set construction: normalizing and preprocessing the depth domain measured depth domain P-wave velocity, depth domain S-wave velocity and density logging data, thereby forming a neural network training sample set; model training: constructing a density prediction model based on a depth feedforward neural network, and training the model using the training sample set, thereby obtaining a nonlinear relationship model between the depth domain P-wave velocity, depth domain S-wave velocity and density, and realizing a density prediction function; model application: normalizing and preprocessing the measured depth domain P-wave velocity and depth domain S-wave velocity prediction data, and then inputting the normalized and preprocessed prediction data into the nonlinear relationship model to predict the density.

Description

Technical Field

[0001] This application relates to the field of oil and gas geophysical exploration, and in particular to a density prediction method, apparatus, equipment and storage medium based on neural networks. Background Technology

[0002] In the field of oil and gas exploration, P-wave velocity, S-wave velocity, and density serve as a bridge connecting rock physical properties and seismic wave exploration, playing important roles in AVO analysis of seismic data, pre-stack inversion, and reservoir lithology, physical properties, and fluid identification. However, in actual production, various reasons lead to the absence or incompleteness of density parameters, affecting subsequent exploration work.

[0003] In production applications, acoustic wave (P-wave) velocity is typically used to fit density, with the Gardner empirical formula being the most common method. This formula is a statistical fit to a large amount of rock density data, but when refined to specific regions, it still exhibits significant errors and cannot meet the current needs for high-precision seismic interpretation. Domestic and international scholars have fitted the coefficients of the Gardner empirical formula based on experimental data from different regions, but essentially it remains a power function of P-wave velocity. If systematic or random errors occur in the measured P-wave velocity data, the resulting density will also experience accumulated and amplified errors. Therefore, it is necessary to find a method to improve the accuracy of density prediction.

[0004] To address the above shortcomings, this invention uses P-wave velocity and S-wave velocity as a basis and employs a deep feedforward neural network to fully explore the intrinsic relationship between rock physical parameters, establish a nonlinear relationship model between P-wave and S-wave velocity and density, and obtain a density prediction model based on the deep feedforward neural network, thereby improving the accuracy of density prediction. Summary of the Invention

[0005] To address the aforementioned problems, this application provides a density prediction method, apparatus, device, and storage medium based on neural networks.

[0006] This application provides a density prediction method based on neural networks, including:

[0007] S1: Data preparation: Obtain measured depth domain P-wave velocity, depth domain S-wave velocity and density logging data for the study area;

[0008] S2: Sample set construction: Normalize and preprocess the measured depth domain P-wave velocity, depth domain S-wave velocity and density logging data to form a training sample set for the neural network.

[0009] S3: Model Training: Construct a density prediction model based on a deep feedforward neural network, and train it using the training sample set to obtain a nonlinear relationship model between depth domain P-wave velocity and depth domain S-wave velocity and density, thereby realizing the density prediction function.

[0010] S4: Model Application: Normalize the predicted data of measured depth domain P-wave velocity and depth domain S-wave velocity, and then input the normalized preprocessed predicted data into the nonlinear relationship model to predict the density.

[0011] In some embodiments, the measured depth-domain P-wave velocity, depth-domain S-wave velocity, and density logging data of the study area specifically include:

[0012] The depth-domain P-wave velocity, depth-domain S-wave velocity, and density data obtained from conventional logging and full-wavetrain logging;

[0013] The measured depth domain longitudinal wave velocity and depth domain transverse wave velocity are used as input data for the neural network;

[0014] The density is used as the output data of the neural network.

[0015] In some embodiments, before performing normalization preprocessing on the measured depth-domain P-wave velocity, depth-domain S-wave velocity, and density logging data, the method further includes:

[0016] Outlier removal was performed on the measured depth domain P-wave velocity, depth domain S-wave velocity, and density logging data.

[0017] In some embodiments, the specific method of the normalization preprocessing includes:

[0018]

[0019] In the formula, b(z) and a(z) are the logging values ​​before and after normalization, respectively; b(z) max and b(z) min These represent the maximum and minimum values ​​of the logging data before normalization, respectively.

[0020] In some embodiments, the neural network is a deep feedforward neural network.

[0021] In some embodiments, the structure of the deep feedforward neural network includes:

[0022] Includes input layer h (0) Output layer h (L) And L-1 hidden layers.

[0023] In some embodiments, the deep feedforward neural network has an input layer neuron count of n0 = 2, and an input layer activation function. Choose the ReLU function, and the Sigmoid function as the activation function for both the output and hidden layers. The number of neurons in the output layer is n. L =1.

[0024] This application provides a density prediction device based on a neural network, comprising:

[0025] The module consists of a data preparation module, a sample set construction module, a model training module, and a model application module.

[0026] Data preparation module: Acquire measured depth domain P-wave velocity, depth domain S-wave velocity, and density logging data for the study area;

[0027] Sample set construction module: Normalizes and preprocesses the measured depth domain P-wave velocity, depth domain S-wave velocity and density logging data to form the training sample set for the neural network;

[0028] Model training module: Construct a density prediction model based on a deep feedforward neural network, and train it using the training sample set to obtain a nonlinear relationship model between depth domain P-wave velocity and depth domain S-wave velocity and density, thereby realizing the density prediction function;

[0029] Model application module: Normalizes and preprocesses the predicted data of measured depth domain P-wave velocity and depth domain S-wave velocity, and then inputs the normalized and preprocessed predicted data into the nonlinear relationship model to predict the density.

[0030] This application provides a density prediction device based on a neural network, including a memory and a processor. The memory stores a computer program, which, when executed by the processor, performs any of the aforementioned density prediction methods based on neural networks.

[0031] This application provides a storage medium storing a computer program that can be executed by one or more processors and can be used to implement the neural network-based density prediction method described above.

[0032] This application provides a density prediction method, apparatus, device, and storage medium based on neural networks.

[0033] Replacing conventional Gardner empirical formulas with deep feedforward neural network models can fully explore the intrinsic relationships between rock physical parameters. Deep neural network models only need to be trained once to establish a nonlinear relationship model between P-wave and S-wave velocities and density, yielding highly accurate density parameters. Attached Figure Description

[0034] The present application will be described in more detail below based on embodiments and with reference to the accompanying drawings.

[0035] Figure 1 A flowchart illustrating the implementation of the density prediction method based on neural networks provided in this application embodiment;

[0036] Figure 2 A diagram of a deep feedforward neural network structure provided in an embodiment of this application;

[0037] Figure 3 The target well depth domain logging curves provided in the embodiments of this application;

[0038] Figure 4 The prediction results and errors provided for the embodiments of this application;

[0039] Figure 5 The flowchart illustrates the implementation of the density prediction method based on neural networks provided in this application embodiment.

[0040] In the accompanying drawings, the same parts are referred to by the same reference numerals, and the drawings are not drawn to scale. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0042] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0043] If the application documents contain similar descriptions such as "first, second, third", the following explanation shall be added: In the following description, the terms "first, second, third" are used only to distinguish similar objects and do not represent a specific order of objects. It is understood that "first, second, third" may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0044] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0045] Before introducing the density prediction method based on neural networks provided in the embodiments of this application, the problems existing in related technologies are briefly introduced:

[0046] In the field of oil and gas exploration, P-wave velocity, S-wave velocity, and density serve as a bridge connecting rock physical properties and seismic wave exploration, playing important roles in AVO analysis of seismic data, pre-stack inversion, and reservoir lithology, physical properties, and fluid identification. However, in actual production, various reasons lead to the absence or incompleteness of density parameters, affecting subsequent exploration work.

[0047] In production applications, acoustic wave (P-wave) velocity is typically used to fit density, with the Gardner empirical formula being the most common method. This formula is a statistical fit to a large amount of rock density data, but when refined to specific regions, it still exhibits significant errors and cannot meet the current needs for high-precision seismic interpretation. Domestic and international scholars have fitted the coefficients of the Gardner empirical formula based on experimental data from different regions, but essentially it remains a power function of P-wave velocity. If systematic or random errors occur in the measured P-wave velocity data, the resulting density will also experience accumulated and amplified errors. Therefore, it is necessary to find a method to improve the accuracy of density prediction.

[0048] To address the problems existing in related technologies, this application provides a density prediction method based on neural networks. This method is applied to a density prediction device based on neural networks, which can be an electronic device, such as a computer or mobile terminal. The functionality implemented by the density prediction method based on neural networks provided in this application can be achieved by the processor of the electronic device calling program code, wherein the program code can be stored in a computer storage medium.

[0049] Example 1

[0050] This application provides a density prediction method based on neural networks. Figure 1 This application provides a schematic diagram illustrating the implementation process of a density prediction method based on a neural network, as shown in the embodiments below. Figure 1 As shown, it includes:

[0051] S1: Data preparation: Obtain measured depth domain P-wave velocity, depth domain S-wave velocity and density logging data for the study area;

[0052] S2: Sample set construction: Normalize and preprocess the measured depth domain P-wave velocity, depth domain S-wave velocity and density logging data to form a training sample set for the neural network.

[0053] S3: Model Training: Construct a density prediction model based on a deep feedforward neural network, and train it using the training sample set to obtain a nonlinear relationship model between depth domain P-wave velocity and depth domain S-wave velocity and density, thereby realizing the density prediction function.

[0054] S4: Model Application: Normalize the predicted data of measured depth domain P-wave velocity and depth domain S-wave velocity, and then input the normalized preprocessed predicted data into the nonlinear relationship model to predict the density.

[0055] This application provides a density prediction method based on neural networks. It utilizes a deep feedforward neural network model to replace the conventional Gardner empirical formula, fully exploring the intrinsic relationships between rock physical parameters. The deep neural network model only needs to be trained once to establish a nonlinear relationship model between P-wave and S-wave velocities and density, obtaining highly accurate density parameters.

[0056] Example 2

[0057] Based on the foregoing embodiments, this application further provides a density prediction method based on neural networks, including:

[0058] S21: Data preparation: Obtain measured depth domain P-wave velocity, depth domain S-wave velocity, and density logging data for the study area;

[0059] In some embodiments, the measured depth-domain P-wave velocity, depth-domain S-wave velocity, and density logging data of the study area specifically include:

[0060] The depth-domain P-wave velocity, depth-domain S-wave velocity, and density data obtained from conventional logging and full-wavetrain logging;

[0061] The measured depth domain longitudinal wave velocity and depth domain transverse wave velocity are used as input data for the neural network;

[0062] The density is used as the output data of the neural network;

[0063] In some embodiments, before performing normalization preprocessing on the measured depth-domain P-wave velocity, depth-domain S-wave velocity, and density logging data, the method further includes:

[0064] Outlier removal was performed on the measured depth domain P-wave velocity, depth domain S-wave velocity, and density logging data.

[0065] S22: Sample set construction: Normalize and preprocess the measured depth domain P-wave velocity, depth domain S-wave velocity and density logging data to form a training sample set for the neural network.

[0066] In some embodiments, the specific method of the normalization preprocessing includes:

[0067]

[0068] In the formula, b(z) and a(z) are the logging values ​​before and after normalization, respectively; b(z) max and b(z) min These are the maximum and minimum values ​​of the logging data before normalization.

[0069] The training sample set Set(z) consists of the input data x(z) after normalization and the output data y(z) after unit conversion.

[0070] S23: Model Training: Construct a density prediction model based on a deep feedforward neural network, and train it using the training sample set to obtain a nonlinear relationship model between depth domain P-wave velocity and depth domain S-wave velocity and density, thereby realizing the density prediction function.

[0071] In some embodiments, the neural network is a deep feedforward neural network;

[0072] S24: Model Application: Normalize the predicted data of measured depth domain P-wave velocity and depth domain S-wave velocity, and then input the normalized preprocessed predicted data into the nonlinear relationship model to predict the density.

[0073] This application provides a density prediction method based on neural networks. It utilizes a deep feedforward neural network model to replace the conventional Gardner empirical formula, fully exploring the intrinsic relationships between rock physical parameters. The deep neural network model only needs to be trained once to establish a nonlinear relationship model between P-wave and S-wave velocities and density, obtaining highly accurate density parameters.

[0074] Example 3

[0075] Based on the foregoing embodiments, this application further provides a density prediction method based on neural networks, including:

[0076] S31: Data preparation: Obtain measured depth domain P-wave velocity, depth domain S-wave velocity, and density logging data for the study area;

[0077] In some embodiments, the measured depth-domain P-wave velocity, depth-domain S-wave velocity, and density logging data of the study area specifically include:

[0078] The depth-domain P-wave velocity, depth-domain S-wave velocity, and density data obtained from conventional logging and full-wavetrain logging;

[0079] The measured depth domain longitudinal wave velocity and depth domain transverse wave velocity are used as input data for the neural network.

[0080] The density is used as the output data of the neural network;

[0081] In some embodiments, before performing normalization preprocessing on the measured depth-domain P-wave velocity, depth-domain S-wave velocity, and density logging data, the method further includes:

[0082] Outlier removal was performed on the measured depth domain P-wave velocity, depth domain S-wave velocity, and density logging data.

[0083] S32: Sample set construction: Normalize and preprocess the measured depth domain P-wave velocity, depth domain S-wave velocity and density logging data to form a training sample set for the neural network.

[0084] In some embodiments, the specific method of the normalization preprocessing includes:

[0085]

[0086] In the formula, b(z) and a(z) are the logging values ​​before and after normalization, respectively; b(z) max and b(z) min These are the maximum and minimum values ​​of the logging data before normalization.

[0087] The training sample set Set(z) consists of the input data x(z) after normalization and the output data y(z) after unit conversion.

[0088] S33: Model Training: Construct a density prediction model based on a deep feedforward neural network, and train it using the training sample set to obtain a nonlinear relationship model between depth domain P-wave velocity and depth domain S-wave velocity and density, thereby realizing the density prediction function.

[0089] In some embodiments, the neural network is a deep feedforward neural network;

[0090] S34: Model Application: Normalize the predicted data of measured depth domain P-wave velocity and depth domain S-wave velocity, and then input the normalized preprocessed predicted data into the nonlinear relationship model to predict the density.

[0091] This application provides a density prediction method based on neural networks. It utilizes a deep feedforward neural network model to replace the conventional Gardner empirical formula, fully exploring the intrinsic relationships between rock physical parameters. The deep neural network model only needs to be trained once to establish a nonlinear relationship model between P-wave and S-wave velocities and density, obtaining highly accurate density parameters.

[0092] Example 4

[0093] Based on the foregoing embodiments, this application further provides a density prediction method based on neural networks, including:

[0094] S41: Data preparation: Obtain measured depth domain P-wave velocity, depth domain S-wave velocity, and density logging data for the study area;

[0095] In some embodiments, the measured depth-domain P-wave velocity, depth-domain S-wave velocity, and density logging data of the study area specifically include:

[0096] The depth-domain P-wave velocity, depth-domain S-wave velocity, and density data obtained from conventional logging and full-wavetrain logging;

[0097] The measured depth domain longitudinal wave velocity and depth domain transverse wave velocity are used as input data for the neural network;

[0098] The density is used as the output data of the neural network;

[0099] In some embodiments, before performing normalization preprocessing on the measured depth-domain P-wave velocity, depth-domain S-wave velocity, and density logging data, the method further includes:

[0100] Outlier removal was performed on the measured depth domain P-wave velocity, depth domain S-wave velocity, and density logging data.

[0101] S42: Sample set construction: Normalize and preprocess the measured depth domain P-wave velocity, depth domain S-wave velocity and density logging data to form a training sample set for the neural network.

[0102] In some embodiments, the specific method of the normalization preprocessing includes:

[0103]

[0104] In the formula, b(z) and a(z) are the logging values ​​before and after normalization, respectively; b(z) max and b(z) min These are the maximum and minimum values ​​of the logging data before normalization.

[0105] The training sample set Set(z) consists of the input data x(z) after normalization and the output data y(z) after unit conversion.

[0106] S43: Model Training: Construct a density prediction model based on a deep feedforward neural network, and train it using the training sample set to obtain a nonlinear relationship model between depth domain P-wave velocity and depth domain S-wave velocity and density, thereby realizing the density prediction function.

[0107] In some embodiments, the neural network is a deep feedforward neural network;

[0108] In some embodiments, this network is a fully connected neural network, where neurons in adjacent layers are fully connected to each other, while neurons in the same hidden layer and between adjacent hidden layers are not connected. For the input data x(z)∈R in the training sample set Set(z) 2 Given the output data y(z)∈R, the output of its hidden layer is:

[0109]

[0110] S44: Model Application: Normalize the predicted data of measured depth domain P-wave velocity and depth domain S-wave velocity, and then input the normalized preprocessed predicted data into the nonlinear relationship model to predict the density.

[0111] This application provides a density prediction method based on neural networks. It utilizes a deep feedforward neural network model to replace the conventional Gardner empirical formula, fully exploring the intrinsic relationships between rock physical parameters. The deep neural network model only needs to be trained once to establish a nonlinear relationship model between P-wave and S-wave velocities and density, obtaining highly accurate density parameters.

[0112] Example 5

[0113] Based on the foregoing embodiments, this application further provides a density prediction method based on neural networks, including:

[0114] S51: Data preparation: Obtain measured depth domain P-wave velocity, depth domain S-wave velocity, and density logging data for the study area;

[0115] In some embodiments, the measured depth-domain P-wave velocity, depth-domain S-wave velocity, and density logging data of the study area specifically include:

[0116] The depth-domain P-wave velocity, depth-domain S-wave velocity, and density data obtained from conventional logging and full-wavetrain logging;

[0117] The measured depth domain longitudinal wave velocity and depth domain transverse wave velocity are used as input data for the neural network.

[0118] The density is used as the output data of the neural network;

[0119] In some embodiments, before performing normalization preprocessing on the measured depth-domain P-wave velocity, depth-domain S-wave velocity, and density logging data, the method further includes:

[0120] Outlier removal was performed on the measured depth domain P-wave velocity, depth domain S-wave velocity, and density logging data.

[0121] S52: Sample set construction: Normalize and preprocess the measured depth domain P-wave velocity, depth domain S-wave velocity and density logging data to form a training sample set for the neural network.

[0122] In some embodiments, the specific method of the normalization preprocessing includes:

[0123]

[0124] In the formula, b(z) and a(z) are the logging values ​​before and after normalization, respectively; b(z) max and b(z) min These are the maximum and minimum values ​​of the logging data before normalization.

[0125] The training sample set Set(z) consists of the input data x(z) after normalization and the output data y(z) after unit conversion.

[0126] S53: Model Training: Construct a density prediction model based on a deep feedforward neural network, and train it using the training sample set to obtain a nonlinear relationship model between depth domain P-wave velocity and depth domain S-wave velocity and density, thereby realizing the density prediction function.

[0127] In some embodiments, the neural network is a deep feedforward neural network;

[0128] In some embodiments, this network is a fully connected neural network, where neurons in adjacent layers are fully connected to each other, while neurons in the same hidden layer and between adjacent hidden layers are not connected. For the input data x(z)∈R in the training sample set Set(z) 2 Given the output data y(z)∈R, the output of its hidden layer is:

[0129]

[0130] In some embodiments, the structure of the deep feedforward neural network includes:

[0131] Includes input layer h (0) Output layer h (L) And L-1 hidden layers;

[0132] In some embodiments, the number of network layers, the number of neurons per layer, and the activation function are as follows:

[0133]

[0134] S54: Model Application: Normalize the predicted data of measured depth domain P-wave velocity and depth domain S-wave velocity, and then input the normalized preprocessed predicted data into the nonlinear relationship model to predict the density.

[0135] This application provides a density prediction method based on neural networks. It utilizes a deep feedforward neural network model to replace the conventional Gardner empirical formula, fully exploring the intrinsic relationships between rock physical parameters. The deep neural network model only needs to be trained once to establish a nonlinear relationship model between P-wave and S-wave velocities and density, obtaining highly accurate density parameters.

[0136] Example 6

[0137] Based on the foregoing embodiments, this application further provides a density prediction method based on neural networks, including:

[0138] S61: Data preparation: Obtain measured depth domain P-wave velocity, depth domain S-wave velocity, and density logging data for the study area;

[0139] In some embodiments, the measured depth-domain P-wave velocity, depth-domain S-wave velocity, and density logging data of the study area specifically include:

[0140] The depth-domain P-wave velocity, depth-domain S-wave velocity, and density data obtained from conventional logging and full-wavetrain logging;

[0141] The measured depth domain longitudinal wave velocity and depth domain transverse wave velocity are used as input data for the neural network.

[0142] The density is used as the output data of the neural network;

[0143] In some embodiments, before performing normalization preprocessing on the measured depth-domain P-wave velocity, depth-domain S-wave velocity, and density logging data, the method further includes:

[0144] Outlier removal was performed on the measured depth domain P-wave velocity, depth domain S-wave velocity, and density logging data.

[0145] S62: Sample set construction: Normalize and preprocess the measured depth domain P-wave velocity, depth domain S-wave velocity and density logging data to form a training sample set for the neural network.

[0146] In some embodiments, the specific method of the normalization preprocessing includes:

[0147]

[0148] In the formula, b(z) and a(z) are the logging values ​​before and after normalization, respectively; b(z) max and b(z) min These are the maximum and minimum values ​​of the logging data before normalization.

[0149] The training sample set Set(z) consists of the input data x(z) after normalization and the output data y(z) after unit conversion.

[0150] S63: Model Training: Construct a density prediction model based on a deep feedforward neural network, and train it using the training sample set to obtain a nonlinear relationship model between depth domain P-wave velocity and depth domain S-wave velocity and density, thereby realizing the density prediction function.

[0151] In some embodiments, the neural network is a deep feedforward neural network;

[0152] In some embodiments, this network is a fully connected neural network, where neurons in adjacent layers are fully connected to each other, while neurons in the same hidden layer and between adjacent hidden layers are not connected. For the input data x(z)∈R in the training sample set Set(z) 2 Given the output data y(z)∈R, the output of its hidden layer is:

[0153]

[0154] In some embodiments, the structure of the deep feedforward neural network includes:

[0155] Includes input layer h (0) Output layer h (L) And L-1 hidden layers;

[0156] In some embodiments, the number of network layers, the number of neurons per layer, and the activation function are as follows:

[0157]

[0158] In some embodiments, the deep feedforward neural network has an input layer neuron count of n0 = 2, and an input layer activation function. Choose the ReLU function, and the Sigmoid function as the activation function for both the output and hidden layers. The number of neurons in the output layer is n. L =1;

[0159] The parameters to be learned in the deep feedforward neural network are:

[0160]

[0161] The relationship between the input and output of a deep feedforward neural network is as follows:

[0162]

[0163] Based on the above formula, the density prediction model based on a deep feedforward neural network can be obtained:

[0164] y = f(x, θ);

[0165] S64: Model Application: Normalize the predicted data of measured depth domain P-wave velocity and depth domain S-wave velocity, and then input the normalized preprocessed predicted data into the nonlinear relationship model to predict the density.

[0166] This application provides a density prediction method based on neural networks. It utilizes a deep feedforward neural network model to replace the conventional Gardner empirical formula, fully exploring the intrinsic relationships between rock physical parameters. The deep neural network model only needs to be trained once to establish a nonlinear relationship model between P-wave and S-wave velocities and density, obtaining highly accurate density parameters.

[0167] Example 7

[0168] Based on the method of Embodiment Six, this application provides embodiments based on real data. Figure 1 and Figure 5 A flowchart illustrating a technique for density prediction according to an embodiment of the present invention is shown, including:

[0169] S71: Data preparation: Obtain measured depth domain P-wave velocity, depth domain S-wave velocity, and density logging data for the study area;

[0170] In some embodiments, the measured depth-domain P-wave velocity, depth-domain S-wave velocity, and density logging data of the study area specifically include:

[0171] The depth-domain P-wave velocity, depth-domain S-wave velocity, and density data obtained from conventional logging and full-wavetrain logging; Figure 3 The depth domain logging data of the example well are shown, including P-wave velocity VP, S-wave velocity VS, and density DEN.

[0172] The measured depth domain longitudinal wave velocity and depth domain transverse wave velocity are used as input data for the neural network;

[0173] The density is used as the output data of the neural network;

[0174] In some embodiments, before performing normalization preprocessing on the measured depth-domain P-wave velocity, depth-domain S-wave velocity, and density logging data, the method further includes:

[0175] Outlier removal was performed on the measured depth domain P-wave velocity, depth domain S-wave velocity, and density logging data.

[0176] S72: Sample set construction: Normalize and preprocess the measured depth domain P-wave velocity, depth domain S-wave velocity and density logging data to form a training sample set for the neural network.

[0177] In some embodiments, the specific method of the normalization preprocessing includes:

[0178]

[0179] In the formula, b(z) and a(z) are the logging values ​​before and after normalization, respectively; b(z) max and b(z) min These are the maximum and minimum values ​​of the logging data before normalization.

[0180] The training sample set Set(z) consists of the input data x(z) after normalization and the output data y(z) after unit conversion.

[0181] S73: Model Training: Construct a density prediction model based on a deep feedforward neural network, and train it using the training sample set to obtain a nonlinear relationship model between depth domain P-wave velocity and depth domain S-wave velocity and density, thereby realizing the density prediction function.

[0182] In some embodiments, the neural network is a deep feedforward neural network. Figure 2 The structure diagram of a deep feedforward neural network is shown;

[0183] In some embodiments, this network is a fully connected neural network, where neurons in adjacent layers are fully connected to each other, while neurons in the same hidden layer and between adjacent hidden layers are not connected. For the input data x(z)∈R in the training sample set Set(z) 2 Given the output data y(z)∈R, the output of its hidden layer is:

[0184]

[0185] In some embodiments, the structure of the deep feedforward neural network includes:

[0186] Includes input layer h (0) Output layer h (L) And L-1 hidden layers;

[0187] In some embodiments, the number of network layers, the number of neurons per layer, and the activation function are as follows:

[0188]

[0189] In some embodiments, the deep feedforward neural network has an input layer neuron count of n0 = 2, and an input layer activation function. Choose the ReLU function, and the Sigmoid function as the activation function for both the output and hidden layers. The number of neurons in the output layer is n. L =1;

[0190] The parameters to be learned in the deep feedforward neural network are:

[0191]

[0192] The relationship between the input and output of a deep feedforward neural network is as follows:

[0193]

[0194] Based on the above formula, the density prediction model based on a deep feedforward neural network can be obtained:

[0195] y = f(x, θ);

[0196] In some embodiments, a deep feedforward neural network with 5 hidden layers and 200 neurons in each hidden layer is constructed. The number of training iterations is set to 1000, and the conjugate gradient method is selected as the optimization algorithm. The deep feedforward neural network is trained using the training sample set to obtain a density prediction model based on the deep feedforward neural network.

[0197] S74: Model Application: Normalize the predicted data of measured depth domain P-wave velocity and depth domain S-wave velocity, and then input the normalized preprocessed predicted data into the nonlinear relationship model to predict the density.

[0198] In some embodiments, the P-wave velocity VP and S-wave velocity VS in the depth domain of the well section are selected as input data for the depth feedforward neural network, following step S1. Normalization preprocessing is then performed in step S2 to form the prediction sample set for the depth feedforward neural network. The density prediction model based on the depth feedforward neural network is trained using step S3, and the prediction sample set is processed, with the results as follows: Figure 4 As shown in the figure; DEN represents the original density parameter, DEN_DFNN represents the prediction result, and DEN_ERROR represents the corresponding error; from Figure 4It can be seen that the method proposed in this invention can fully explore the intrinsic relationship between P-wave and S-wave velocity and density, obtain highly accurate density parameters, and have a small prediction error, which can provide strong data support for subsequent oil and gas geophysical exploration in the study area.

[0199] This application provides a density prediction method based on neural networks. It utilizes a deep feedforward neural network model to replace the conventional Gardner empirical formula, fully exploring the intrinsic relationships between rock physical parameters. The deep neural network model only needs to be trained once to establish a nonlinear relationship model between P-wave and S-wave velocities and density, obtaining highly accurate density parameters.

[0200] Example 8

[0201] Based on the foregoing embodiments, this application provides a density prediction device based on a neural network, comprising:

[0202] The module consists of a data preparation module, a sample set construction module, a model training module, and a model application module.

[0203] Data preparation module: Acquire measured depth domain P-wave velocity, depth domain S-wave velocity, and density logging data for the study area;

[0204] Sample set construction module: Normalizes and preprocesses the measured depth domain P-wave velocity, depth domain S-wave velocity and density logging data to form the training sample set for the neural network;

[0205] Model training module: Construct a density prediction model based on a deep feedforward neural network, and train it using the training sample set to obtain a nonlinear relationship model between depth domain P-wave velocity and depth domain S-wave velocity and density, thereby realizing the density prediction function;

[0206] Model application module: Normalizes and preprocesses the predicted data of measured depth domain P-wave velocity and depth domain S-wave velocity, and then inputs the normalized and preprocessed predicted data into the nonlinear relationship model to predict the density.

[0207] Example 9

[0208] Based on the foregoing embodiments, this application provides a density prediction device based on a neural network, comprising:

[0209] The module consists of a data preparation module, a sample set construction module, a model training module, and a model application module.

[0210] Data preparation module: Acquire measured depth domain P-wave velocity, depth domain S-wave velocity, and density logging data for the study area;

[0211] In some embodiments, the measured depth-domain P-wave velocity, depth-domain S-wave velocity, and density logging data of the study area specifically include:

[0212] The depth-domain P-wave velocity, depth-domain S-wave velocity, and density data obtained from conventional logging and full-wavetrain logging;

[0213] The measured depth domain longitudinal wave velocity and depth domain transverse wave velocity are used as input data for the neural network.

[0214] The density is used as the output data of the neural network;

[0215] In some embodiments, before performing normalization preprocessing on the measured depth-domain P-wave velocity, depth-domain S-wave velocity, and density logging data, the method further includes:

[0216] Outlier removal was performed on the measured depth domain P-wave velocity, depth domain S-wave velocity, and density logging data.

[0217] Sample set construction module: Normalizes and preprocesses the measured depth domain P-wave velocity, depth domain S-wave velocity and density logging data to form the training sample set for the neural network;

[0218] In some embodiments, the specific method of the normalization preprocessing includes:

[0219]

[0220] In the formula, b(z) and a(z) are the logging values ​​before and after normalization, respectively; b(z) max and b(z) min These are the maximum and minimum values ​​of the logging data before normalization.

[0221] The training sample set Set(z) consists of the input data x(z) after normalization and the output data y(z) after unit conversion.

[0222] Model training module: Construct a density prediction model based on a deep feedforward neural network, and train it using the training sample set to obtain a nonlinear relationship model between depth domain P-wave velocity and depth domain S-wave velocity and density, thereby realizing the density prediction function;

[0223] In some embodiments, the neural network is a deep feedforward neural network;

[0224] In some embodiments, this network is a fully connected neural network, where neurons in adjacent layers are fully connected to each other, while neurons in the same hidden layer and between adjacent hidden layers are not connected. For the input data x(z)∈R in the training sample set Set(z) 2Given the output data y(z)∈R, the output of its hidden layer is:

[0225]

[0226] In some embodiments, the structure of the deep feedforward neural network includes:

[0227] Includes input layer h (0) Output layer h (L) And L-1 hidden layers;

[0228] In some embodiments, the number of network layers, the number of neurons per layer, and the activation function are as follows:

[0229]

[0230] In some embodiments, the deep feedforward neural network has an input layer neuron count of n0 = 2, and an input layer activation function. Choose the ReLU function, and the Sigmoid function as the activation function for both the output and hidden layers. The number of neurons in the output layer is n. L =1;

[0231] The parameters to be learned in the deep feedforward neural network are:

[0232]

[0233] The relationship between the input and output of a deep feedforward neural network is as follows:

[0234]

[0235] Based on the above formula, the density prediction model based on a deep feedforward neural network can be obtained:

[0236] y = f(x, θ);

[0237] Model application module: Normalizes and preprocesses the predicted data of measured depth domain P-wave velocity and depth domain S-wave velocity, and then inputs the normalized and preprocessed predicted data into the nonlinear relationship model to predict the density.

[0238] It should be noted that, in the embodiments of this application, if the above-described density prediction method based on neural networks is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiments of this application, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), magnetic disks, or optical disks. Thus, the embodiments of this application are not limited to any specific hardware and software combination.

[0239] Accordingly, this application provides a storage medium storing a computer program thereon, characterized in that the computer program, when executed by a processor, implements the steps in the density prediction method based on neural networks provided in the above embodiments.

[0240] Example 10

[0241] This application provides a density prediction device memory and processor based on a neural network. The memory stores a computer program. When the processor executes the computer program, the processor is configured to execute a program for a multi-scale electromagnetic field component denoising method stored in the memory, so as to implement the steps in the multi-scale electromagnetic field component denoising method provided in the above embodiment.

[0242] The descriptions of the display device and storage medium embodiments above are similar to those of the method embodiments above, and have similar beneficial effects. For technical details not disclosed in the computer device and storage medium embodiments of this application, please refer to the descriptions of the method embodiments of this application for understanding.

[0243] It should be noted that the descriptions of the storage medium and device embodiments above are similar to those of the method embodiments above, and have similar beneficial effects. For technical details not disclosed in the storage medium and device embodiments of this application, please refer to the descriptions of the method embodiments of this application for understanding.

[0244] It should be understood that the phrase "one embodiment" or "an embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of this application. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. It should be understood that in the various embodiments of this application, the sequence numbers of the above-described processes do not imply a sequential order of execution; the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application. The sequence numbers of the above-described embodiments are merely descriptive and do not represent the superiority or inferiority of the embodiments.

[0245] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0246] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0247] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0248] In addition, each functional unit in the various embodiments of this application can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0249] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media that can store program code, such as mobile storage devices, read-only memory (ROM), magnetic disks, or optical disks.

[0250] Alternatively, if the integrated units described above are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a controller to execute all or part of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROMs, magnetic disks, or optical disks.

[0251] The above description is merely an embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A density prediction method based on neural networks, characterized in that, include: S1: Data preparation: Obtain measured depth domain P-wave velocity, depth domain S-wave velocity, and density logging data for the study area; S2: Sample set construction: Normalize and preprocess the measured depth domain P-wave velocity, depth domain S-wave velocity and density logging data to form a training sample set for the neural network. S3: Model Training: The conjugate gradient method is selected as the optimization algorithm for the neural network. A density prediction model based on a deep feedforward neural network is constructed, and the training sample set is used to train it to obtain a nonlinear relationship model between depth domain P-wave velocity and depth domain S-wave velocity and density, thereby realizing the density prediction function. S4: Model Application: Normalize the predicted data of measured depth domain P-wave velocity and depth domain S-wave velocity, and then input the normalized preprocessed predicted data into the nonlinear relationship model to predict the density. The neural network is a deep feedforward neural network, and the structure of the deep feedforward neural network includes: an input layer. Output layer And L-1 hidden layers; the number of neurons in the input layer of the deep feedforward neural network. Input layer activation function Choose the ReLU function, and the Sigmoid function as the activation function for both the output and hidden layers. The number of neurons in the output layer... ; The deep feedforward neural network is a fully connected neural network, where neurons in adjacent layers are fully connected to each other, while neurons in the same hidden layer and between adjacent hidden layers are not connected to each other; for the input data x(z)∈ in the training sample set Set(z) Given the output data y(z)∈R, the output of its hidden layer is: The parameters to be learned in the deep feedforward neural network are: The measured depth-domain P-wave velocity, depth-domain S-wave velocity, and density logging data for the study area specifically include: The depth-domain P-wave velocity, depth-domain S-wave velocity, and density data obtained from conventional logging and full-wavetrain logging; The measured depth domain longitudinal wave velocity and depth domain transverse wave velocity are used as input data for the neural network; The density is used as the output data of the neural network; Before performing normalization preprocessing on the measured depth domain P-wave velocity, depth domain S-wave velocity, and density logging data, the method further includes: Outliers were removed from the measured depth domain P-wave velocity, depth domain S-wave velocity, and density logging data. The specific methods for the normalization preprocessing include: In the formula, b(z) and a(z) are the logging values ​​before and after normalization, respectively; and These represent the maximum and minimum values ​​of the logging data before normalization, respectively.

2. A neural network-based density prediction device for implementing the neural network-based density prediction method of claim 1, characterized in that, include: The module consists of a data preparation module, a sample set construction module, a model training module, and a model application module. Data preparation module: Acquire measured depth domain P-wave velocity, depth domain S-wave velocity, and density logging data for the study area; Sample set construction module: Normalizes and preprocesses the measured depth domain P-wave velocity, depth domain S-wave velocity and density logging data to form the training sample set for the neural network; Model training module: Constructs a density prediction model based on a deep feedforward neural network, and trains it using the training sample set to obtain a nonlinear relationship model between depth domain P-wave velocity and depth domain S-wave velocity and density, thereby realizing the density prediction function. Model application module: Normalizes and preprocesses the predicted data of measured depth domain P-wave velocity and depth domain S-wave velocity, and then inputs the normalized and preprocessed predicted data into the nonlinear relationship model to predict the density.

3. A density prediction device based on a neural network, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program that, when executed by the processor, performs the density prediction method based on a neural network as described in claim 1.

4. A storage medium, characterized in that, The computer program stored in the storage medium can be executed by one or more processors and can be used to implement the neural network-based density prediction method as described in claim 1.

Images