Oil reservoir parameter prediction method, device and equipment based on petrophysics

By combining petrophysics and deep learning techniques in reservoir parameter prediction, using random forest and long and short-term memory network models to process core and log data, the problems of poor prediction accuracy and low efficiency in the existing technology are solved, and the problem of achieving the goal of achieving More accurate and efficient assessment of reservoir parameters.

CN118939955AActive Publication Date: 2024-11-12CHINA UNIV OF PETROLEUM (BEIJING)

Patent Information

Application Number
CN202410795809.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-06-19
Publication Date
2024-11-12
Estimated Expiration
2044-06-19

AI Technical Summary

Technical Problem

In the prior art, the prediction of reservoir parameters is poor and the working efficiency is low. The direct determination method is subject to economic and technical constraints, and the indirect interpretation method is prone to overcomplexity and overfitting problems.

Method used

Using a method based on petrophysics, by obtaining core data and logging data, after initialization, the logging data is input to the random forest model to calculate the feature weight, and a long and short-term memory network model is constructed, and the target prediction is constructed based on petrophysics knowledge. Model, used to predict reservoir parameters.

Benefits of technology

It improves the accuracy and working efficiency of reservoir reservoir parameter prediction, reduces prediction errors, and enhances the accuracy of reservoir characteristics evaluation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118939955A_ABST
    Figure CN118939955A_ABST
Patent Text Reader

Abstract

The invention provides an oil reservoir parameter prediction method, device and equipment based on petrophysics. The method is applied to the technical field of oil reservoir engineering. Performing initialization processing on the core data and the logging data to obtain initialized core data and logging data; inputting the initialized logging data into a random forest model, and calculating weights of a plurality of features of the logging data to oil reservoir parameters to obtain target features of the logging data; constructing a long-short-term memory network model according to target features of the initialized core data and logging data, constructing a target prediction model according to the long-short-term memory network model and rock physics knowledge, training the target prediction model, and predicting parameters of an oil reservoir based on rock physics by the target prediction model. Therefore, the prediction accuracy is improved, and the working efficiency is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application belongs to the field of oil reservoir engineering technology, and in particular, relates to a method, device, and equipment for predicting oil reservoir parameters based on rock physics. Background Art

[0002] Reservoir engineering is a branch of petroleum engineering that focuses on the study and development of underground oil and gas reservoirs to effectively extract oil and gas resources. During the exploration phase, reservoir engineers use geological exploration techniques such as seismic exploration, geological exploration, and geophysical exploration to find potential oil and gas reservoirs. Through reserve assessment, the size, nature, and recoverable reserves of oil and gas reservoirs are determined to provide a basis for subsequent development.

[0003] Reservoir parameter prediction is an important basis for detailed reservoir evaluation and oil and gas development. There are currently two main methods for determining reservoir parameters: direct measurement method and indirect interpretation method. The direct measurement method obtains physical parameters by coring the core and analyzing it through rock physics knowledge; the indirect interpretation method predicts and estimates relevant reservoir parameters based on logging data. The main idea is to use the current logging data to create a prediction model that links reservoir parameters with logging curves to predict reservoir parameters.

[0004] However, although the data obtained by the direct measurement method is accurate, due to economic and technical constraints, the limited data collected from rock samples makes it inefficient to accurately estimate the reservoir parameters of the entire work area; the indirect interpretation method is prone to problems such as over-complexity and overfitting, so the predicted results have large errors in accuracy. Summary of the invention

[0005] The present application provides a method, device and equipment for predicting oil reservoir parameters based on petrophysics, so as to solve the defects of poor prediction accuracy and low working efficiency in the prior art.

[0006] In a first aspect, the present application provides a method for predicting reservoir parameters based on petrophysics, the method comprising:

[0007] Obtain core data and logging data of the area to be tested;

[0008] Initializing the core data and the well logging data to obtain initialized core data and well logging data;

[0009] Inputting the initialized logging data into a random forest model, calculating the weights of multiple features of the logging data to reservoir parameters, and obtaining target features of the logging data;

[0010] According to the initialized core data and the target features of the logging data, a long short-term memory network model is constructed, wherein the long short-term memory network model is used to determine the correlation relationship between the core data and the target features of the logging data;

[0011] A target prediction model is constructed according to the long short-term memory network model and rock physics knowledge, and the target prediction model is trained. The target prediction model predicts parameters of oil reservoirs based on rock physics.

[0012] Optionally, the reservoir parameters include porosity, permeability, and water saturation. The initialized logging data is input into a random forest model, and the weights of multiple features of the logging data on the reservoir parameters are calculated to obtain target features of the logging data, including:

[0013] Inputting the initialized logging data into a random forest model, respectively calculating the influence weight values ​​of the multiple features on the porosity, permeability, and water saturation, and obtaining a porosity weight value, a permeability weight value, and a water saturation weight value;

[0014] The porosity weight value, the permeability weight value, and the water saturation weight value are sorted from large to small;

[0015] The features corresponding to the first preset number of porosity weight values ​​after sorting are used as the target features of porosity;

[0016] The features corresponding to the first preset number of permeability weight values ​​after sorting are used as the target features of the permeability;

[0017] The features corresponding to the first preset number of water saturation weight values ​​after sorting are used as target features of water saturation.

[0018] Optionally, before sorting the porosity weight values, the permeability weight values, and the water saturation weight values ​​from large to small weight values, the method further includes:

[0019] Determine whether there is a first associated feature with a correlation coefficient greater than a threshold value among the multiple features;

[0020] If so, compare the weight values ​​of at least two features in the first associated features, the first associated features with smaller weight values ​​in the first associated features do not participate in the weight value sorting process, and the first associated features with larger weight values ​​in the first associated features participate in the weight value sorting process;

[0021] If not, continue to sort the multiple weight values ​​in descending order of weight value.

[0022] Optionally, the determining whether there is a first associated feature having a correlation coefficient greater than a threshold value among the multiple features includes:

[0023] The correlation coefficient between the multiple features is calculated using the following formula:

[0024]

[0025] Among them, γ is the correlation coefficient, is the mean value of the logging curve, x i ,y i Represents the logging curve value corresponding to the i-th sample.

[0026] The absolute value of the correlation coefficient γ is calculated, and it is determined whether |γ| is greater than a threshold.

[0027] Optionally, constructing a long short-term memory network model according to the initialized core data and the target features of the logging data, wherein the long short-term memory network model is used to determine the association relationship between the core data and the target features of the logging data, includes:

[0028] Constructing a long short-term memory network model according to the initialized core data and the target features of the logging data;

[0029] According to the target features of the initialized core data and the logging data, a long short-term memory network model is trained to obtain a trained long short-term memory network model;

[0030] The initialized core data is used as input data of the trained long short-term memory network model, and the target features of the well logging data are used as output data of the trained long short-term memory network model. The long short-term memory network model is controlled to perform correlation analysis on the target features of the core data and the well logging data to obtain the correlation relationship between the core data and the target features of the well logging data.

[0031] Optionally, the initialized core data includes multiple features, and constructing the long short-term memory network model according to the initialized core data and the target features of the logging data includes:

[0032] According to the number of features of the initialized core data, a long short-term memory network model input layer with a dimension corresponding to the number of features is constructed;

[0033] According to the number of features of the target features of the well logging data, a long short-term memory network model output layer with a dimension corresponding to the number of features is constructed;

[0034] The number of units and the number of layers of the long short-term memory network model are constructed according to the target features of the initialized core data and the logging data.

[0035] Optionally, the training of the target prediction model includes:

[0036] Using the correlation between the core data and the target features of the logging data and petrophysical knowledge as input data of the target prediction model, and controlling the target prediction model to perform iterative processing;

[0037] Obtaining a total loss function of the target prediction model, wherein the total loss function includes a first loss function of a long short-term memory network model and a second loss function of petrophysical knowledge;

[0038] When it is determined that the total loss function converges, the target prediction model is controlled to stop iterative processing, and it is determined that the training of the target prediction model is completed.

[0039] In a second aspect, the present application provides a device for predicting oil reservoir parameters based on petrophysics, the device comprising:

[0040] An acquisition module is used to acquire core data and logging data of the area to be tested;

[0041] A processing module, used for initializing the core data and the well logging data to obtain initialized core data and well logging data;

[0042] A calculation module, used for inputting the initialized logging data into a random forest model, calculating the weights of multiple features of the logging data to reservoir parameters, and obtaining target features of the logging data;

[0043] The processing module is further used to construct a long short-term memory network model according to the initialized core data and the target features of the logging data, and the long short-term memory network model is used to determine the correlation relationship between the core data and the target features of the logging data;

[0044] The processing module is also used to construct a target prediction model based on the long short-term memory network model and rock physics knowledge, and train the target prediction model, wherein the target prediction model predicts parameters of oil reservoirs based on rock physics.

[0045] Optionally, the petrophysics-based reservoir parameter prediction device further includes: a determination module;

[0046] The calculation module is further used to input the initialized logging data into a random forest model, and respectively calculate the influence weight values ​​of the multiple features on the porosity, permeability, and water saturation to obtain a porosity weight value, a permeability weight value, and a water saturation weight value;

[0047] The processing module is further used to sort the porosity weight value, the permeability weight value, and the water saturation weight value from large to small weight values;

[0048] The determination module is used to use the features corresponding to the first preset number of porosity weight values ​​after sorting as the target features of porosity;

[0049] The determination module is further used to use the features corresponding to the first preset number of permeability weight values ​​after sorting as the target features of the permeability;

[0050] The determination module is further configured to use the features corresponding to the first preset number of water saturation weight values ​​after sorting as target features of water saturation.

[0051] Optionally, the petrophysics-based reservoir parameter prediction device further includes: a judgment module;

[0052] The judging module is used to judge whether there is a first associated feature with a correlation coefficient greater than a threshold value among the multiple features;

[0053] The processing module is further configured to compare the weight values ​​of at least two features among the plurality of features when there is a first associated feature whose correlation coefficient is greater than a threshold value, so that the first associated feature with a smaller weight value among the first associated features does not participate in the weight value sorting process, and the first associated feature with a larger weight value among the first associated features participates in the weight value sorting process;

[0054] The processing module is further configured to, when there is no first associated feature with a correlation coefficient greater than a threshold value among the multiple features, continue to sort the multiple weight values ​​in descending order of weight value.

[0055] Optionally, the calculation module is further configured to calculate the correlation coefficient between the multiple features using the following formula:

[0056]

[0057] Among them, γ is the correlation coefficient, is the mean value of the logging curve, x i ,y i Represents the logging curve value corresponding to the i-th sample;

[0058] The calculation module is further used to calculate the absolute value of the correlation coefficient γ;

[0059] The judgment module is further used to judge whether |γ| is greater than a threshold.

[0060] Optionally, the processing module is further used to construct a long short-term memory network model according to the target features of the initialized core data and the logging data;

[0061] The processing module is further used to train the long short-term memory network model according to the target features of the initialized core data and the logging data to obtain a trained long short-term memory network model;

[0062] The processing module is also used to use the initialized core data as input data of the trained long short-term memory network model, and the target features of the logging data as output data of the trained long short-term memory network model, and control the long short-term memory network model to perform correlation analysis and processing on the target features of the core data and the logging data, so as to obtain the correlation relationship between the core data and the target features of the logging data.

[0063] Optionally, the processing module is further used to construct a long short-term memory network model input layer with a dimension corresponding to the number of features according to the number of features of the initialized core data;

[0064] The processing module is further used to construct a long short-term memory network model output layer with a dimension corresponding to the number of features according to the number of features of the target features of the well logging data;

[0065] The processing module is also used to construct the number of units and layers of the long short-term memory network model according to the target features of the initialized core data and the logging data.

[0066] Optionally, the processing module is further used to use the correlation between the core data and the target characteristics of the logging data and petrophysical knowledge as input data of the target prediction model, and control the target prediction model to perform iterative processing;

[0067] The acquisition module is further used to acquire the total loss function of the target prediction model, wherein the total loss function includes a first loss function of a long short-term memory network model and a second loss function of petrophysical knowledge;

[0068] The processing module is also used to control the target prediction model to stop iterative processing when it is determined that the total loss function has converged, and determine that the training of the target prediction model is completed.

[0069] In a third aspect, the present application provides a petrophysics-based reservoir parameter prediction device, comprising:

[0070] Memory;

[0071] processor;

[0072] Wherein, the memory stores computer-executable instructions;

[0073] The processor executes the computer-executable instructions stored in the memory to implement the petrophysics-based reservoir parameter prediction method as described in the above-mentioned first aspect and various possible implementations of the first aspect.

[0074] In a fourth aspect, the present application provides a computer storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the petrophysics-based reservoir parameter prediction method as described in the first aspect and various possible implementations of the first aspect.

[0075] The present application provides a method, device and equipment for predicting oil reservoir parameters based on rock physics. The method obtains core data and well logging data of the area to be tested; initializes the core data and well logging data to obtain initialized core data and well logging data; inputs the initialized well logging data into a random forest model, calculates the weights of multiple features of the well logging data on the oil reservoir parameters, and obtains the target features of the well logging data; constructs a long short-term memory network model based on the target features of the initialized core data and well logging data, constructs a target prediction model based on the long short-term memory network model and rock physics knowledge, and trains the target prediction model, which predicts the parameters of the oil reservoir based on rock physics, thereby improving the accuracy of the prediction and improving work efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0077] Figure 1 Schematic diagram of the process of predicting reservoir parameters based on rock physics provided in this application Figure 1 ;

[0078] Figure 2 Schematic diagram of the process of predicting reservoir parameters based on rock physics provided in this application Figure 2 ;

[0079] Figure 3 Schematic diagram of different degrees of correlation coefficients between each well logging feature provided for this application;

[0080] Figure 4 A schematic diagram showing the ranking of the weight values ​​of the impact of multiple features of the well logging data provided in this application on porosity;

[0081] Figure 5 A schematic diagram showing the ranking of weight values ​​of multiple features of well logging data provided in this application that affect permeability;

[0082] Figure 6 A schematic diagram showing the ranking of weight values ​​of multiple features of well logging data provided in this application that affect water saturation;

[0083] Figure 7 A petrophysics-based reservoir parameter prediction model provided for this application;

[0084] Figure 8 A schematic diagram of the structure of the oil reservoir parameter prediction device based on rock physics provided in this application;

[0085] Fig. 9 A schematic diagram of the structure of the oil reservoir parameter prediction device based on rock physics provided in this application.

[0086] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0087] In order to make the purpose, technical solutions and advantages of this application clearer, the technical solutions in this application will be clearly and completely described below in conjunction with the drawings in this application. Obviously, the described embodiments are part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0088] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in sequences other than those illustrated or described herein.

[0089] In the embodiments of the present application, the words "exemplary" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.

[0090] First, the terms involved in this application are explained.

[0091] Long Short-Term Memory Network Model: (LSTM, Long Short-Term Memory) is a special type of recurrent neural network (RNN) that can learn long-term dependent information. LSTM solves the gradient vanishing or exploding problem encountered by traditional RNN when processing long sequence data by introducing three gating mechanisms. The three gates are: input gate, forget gate and output gate. The input gate determines which information should be added to the cell state; the forget gate determines which information should be forgotten, that is, removed from the cell state; the output gate determines which information should be included in the final output. This structure enables LSTM to perform well in processing time series prediction, language models, machine translation and other tasks, and can capture long-term dependencies in the data.

[0092] Random Forest Model: Random forest is an ensemble learning method consisting of multiple decision trees for classification and regression problems. It improves the accuracy and robustness of the model by building multiple decision trees and aggregating their predictions. Each decision tree uses a randomly selected subset of samples (bootstrap sampling) during training, and only considers a random subset of features when splitting nodes. This randomness increases the diversity of the model and reduces the risk of overfitting. The advantages of random forest include easy implementation, fast training speed, and good tolerance to noise and outliers in the data. In addition, it can provide an assessment of feature importance to help understand the decision-making process of the model. However, random forest models may take up a lot of memory and the model is not as interpretable as a single decision tree.

[0093] Reservoir engineering is a subfield of petroleum engineering that focuses on the study and development of underground oil and gas reservoirs to efficiently extract oil and gas resources. During the exploration phase, reservoir engineers use a variety of geological survey techniques, such as seismic surveys, geological surveys, and geophysical surveys, to identify and evaluate potential oil and gas reservoirs. These techniques help engineers understand the physical properties, structure, and fluid distribution of underground rocks. Once an oil and gas reservoir is identified, reservoir engineers perform reserve assessment, which involves estimating the volume of the reservoir, the quality of the oil and gas, and the amount of recoverable reserves.

[0094] Reservoir parameter prediction is crucial for the detailed evaluation of reservoirs and the effective development of oil and gas. It involves an accurate assessment of the physical characteristics of the reservoir and the properties of the fluid. The direct measurement method relies on core samples obtained from the reservoir. Through laboratory tests such as core analysis, porosity and permeability tests, the physical properties of the rock can be directly and accurately measured. However, the direct measurement method requires drilling to obtain core samples from underground reservoirs, resulting in high costs; due to the limited number of core samples, it is impossible to fully represent the geological characteristics of the entire reservoir, and the heterogeneity of the reservoir may cause deviations between the sample data and the actual reservoir conditions; the collection, transportation and analysis of cores takes time, which greatly affects the timeliness of development decisions.

[0095] The indirect interpretation method mainly relies on logging data. Well logging is a non-invasive technology that can evaluate underground reservoirs through the wellbore without removing the core. It collects the response data of underground rocks and fluids by using different types of logging tools, establishes a mathematical model between logging response and reservoir parameters, and applies the established model to predict the physical and fluid properties of the reservoir. The advantage of the indirect interpretation method is that it can provide a continuous distribution map of reservoir parameters without relying on limited core samples. However, the accuracy of this method is limited by the quality of logging data and the reliability of the interpretation model.

[0096] In response to the above problems, the present application proposes a method, device and equipment for predicting oil reservoir parameters based on rock physics. The method obtains core data and logging data of the area to be tested; initializes the core data and logging data to obtain initialized core data and logging data; inputs the initialized logging data into a random forest model, calculates the weights of multiple features of the logging data on the oil reservoir parameters, and obtains the target features of the logging data; constructs a long short-term memory network model based on the target features of the initialized core data and logging data, constructs a target prediction model based on the long short-term memory network model and rock physics knowledge, and trains the target prediction model. The target prediction model predicts the parameters of the oil reservoir based on rock physics, thereby improving the accuracy of the prediction and improving work efficiency.

[0097] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0098] Figure 1 Schematic diagram of the process of predicting oil reservoir parameters based on rock physics provided in the application example Figure 1 .like Figure 1As shown, the method for predicting reservoir parameters based on petrophysics provided in this embodiment includes:

[0099] S101, obtaining core data and logging data of the area to be tested.

[0100] The core is a cylindrical sample of underground rock taken from an oil and gas well. Through the coring barrel of special equipment in the drilling process, a physical sample of the underground rock formation can be obtained. The core data of the area to be tested is obtained, specifically the porosity, permeability, and water saturation of the core of the area to be tested.

[0101] Porosity: Indicates the volume of pore space in rock as a percentage of the total volume, and is an important parameter for evaluating reservoir storage capacity. Permeability: Describes the ability of fluid to flow in rock pores, usually in Darcy units, and is a key parameter that determines oil and gas flow and reservoir production capacity. Water Saturation: The volume percentage of water phase in a reservoir, which is related to porosity and fluid properties.

[0102] Well logging is a method of measuring the physical properties of the rocks around the well wall by inserting special instruments into the well. Well logging technology does not require the removal of rock samples, so it can provide continuous underground information. Specifically, the natural gamma, lithology density, acoustic wave time difference, compensated neutron, deep lateral resistivity, shallow lateral resistivity, natural potential, and caliper logging data of the well logging in the area to be tested are obtained.

[0103] Natural Gamma: The content of radioactive elements in rocks, often used to identify rock types, such as mudstone and sandstone. Density Log: The density of rocks is estimated by measuring the absorption of gamma rays by rocks, which helps to identify rock types and porous fluids. Acoustic Log: The propagation time of sound waves in rocks can be used to estimate the porosity of rocks and the type of fluids. Compensated Neutron Log: Porosity is estimated by measuring the scattering of neutrons by rocks, and has different responses to the porosity of water and oil. Deep Lateral Resistivity: Measuring the resistivity of formations away from the wellbore wall helps to identify the oil-water interface and determine the reservoir boundary. Shallow Lateral Resistivity: Measuring the resistivity of formations close to the wellbore wall can provide information on the lateral changes in the resistivity of the formation. Spontaneous Potential (SP): The potential difference formed due to the adsorption of fluids on the rock surface can be used to identify oil and water layers. Caliper Log: It measures the diameter of the wellbore and can detect irregularities in the wellbore wall, such as shrinkage and collapse.

[0104] Obtaining core data and logging data of the area to be tested is a very important step in reservoir engineering. These data are crucial for evaluating the properties and characteristics of underground reservoirs. The comprehensive use of core data and logging data can help reservoir engineers better understand the characteristics of underground reservoirs, guide oil and gas exploration and development, and improve the efficiency and success rate of resource extraction.

[0105] S102: Initialize the core data and the well logging data to obtain initialized core data and well logging data.

[0106] Among them, the core data are initialized. Specifically, the core data are cleaned to remove outliers, missing values ​​and noise data to ensure the accuracy and completeness of the data; the core data are standardized so that data with different attributes have the same scale and range to facilitate subsequent data analysis and comparison; useful features are extracted from the core data, such as porosity, permeability, etc., for subsequent geological modeling and reservoir evaluation.

[0107] The logging data is initialized. Specifically, the logging data is cleaned to remove outliers, missing values ​​and noise data to ensure the accuracy and completeness of the data; the logging data is corrected, taking into account the response characteristics of the logging tool and environmental factors to obtain more accurate formation parameters; the logging curve is interpreted, the logging curve is compared and matched with the geological characteristics, and the formation parameters and characteristics are extracted.

[0108] Align and match core data and well logging data to ensure consistency and comparability between the two. By initializing the core data and well logging data, cleaned, standardized and interpreted data can be obtained to prepare for subsequent geological modeling, reservoir evaluation and resource development.

[0109] S103, inputting the initialized logging data into the random forest model, calculating the weights of multiple features of the logging data to the reservoir parameters, and obtaining the target features of the logging data.

[0110] Among them, the initialized logging data is input into the random forest model, and the parameters of the random forest model, such as the number of trees, the size of the feature subset considered during splitting, etc., are set to initialize the model; the initialized logging data is used to train the random forest model. Specifically, each decision tree randomly extracts samples from the entire data set during training, and randomly selects features for splitting at each splitting node. The random forest model can evaluate the importance of each feature to the prediction result. Through this evaluation, the logging data features that have the greatest impact on the reservoir parameters can be identified, and the logging data features that have the greatest impact on the reservoir parameters can be used as the target features of the logging data.

[0111] S104. Construct a long short-term memory network model based on the target features of the initialized core data and logging data.

[0112] Among them, according to the target features of the initialized core data and logging data, a long short-term memory network model is constructed. The long short-term memory network model is used to determine the correlation relationship between the core data and the target features of the logging data.

[0113] Specifically, according to the target features of the initialized core data and well logging data, define the architecture of the long short-term memory network, including the number of layers, the number of units in each layer, etc.; select an appropriate activation function, such as tanh or ReLU; determine the loss function and optimizer for the training process. Input the organized core data and well logging data into the long short-term memory model for model training. During the training process, the model will learn the correlation between the features of the core data and the well logging data. The trained long short-term memory network model can show the correlation between the target features of the core data and the well logging data.

[0114] S105. Construct a target prediction model based on the long short-term memory network model and rock physics knowledge, and train the target prediction model. The target prediction model predicts parameters of the oil reservoir based on rock physics.

[0115] Before building the target prediction model, we first integrate rock physics knowledge, including but not limited to understanding the relationship between different rock properties and reservoir parameters; determine which rock properties have an important impact on the prediction of reservoir parameters, and incorporate this knowledge into the model design. Combine the long short-term memory network model and rock physics knowledge to build a comprehensive target prediction model for predicting reservoir parameters.

[0116] The data is input into the comprehensive model to train the target prediction model. During the training process, the model will learn the complex relationship between rock properties and reservoir parameters and optimize the parameters. The model is optimized and adjusted according to the evaluation results to improve the accuracy and generalization ability of the model. The trained target prediction model is obtained, which predicts the parameters of the reservoir based on rock physics.

[0117] This embodiment proposes a method for predicting oil reservoir parameters based on rock physics, which obtains core data and logging data of the area to be tested; initializes the core data and logging data to obtain initialized core data and logging data; inputs the initialized logging data into a random forest model, calculates the weights of multiple features of the logging data on the oil reservoir parameters, and obtains target features of the logging data; constructs a long short-term memory network model based on the target features of the initialized core data and logging data, constructs a target prediction model based on the long short-term memory network model and rock physics knowledge, and trains the target prediction model, which predicts the parameters of the oil reservoir based on rock physics, thereby improving the accuracy of the prediction and improving work efficiency.

[0118] Figure 2 The process diagram of the method for predicting oil reservoir parameters based on rock physics provided in the embodiment of the present application is as follows: Figure 2 This embodiment is in Figure 1 Based on the embodiments, the method for predicting oil reservoir parameters based on rock physics is described in detail. Figure 2 As shown, the method for predicting reservoir parameters based on petrophysics provided in this embodiment includes:

[0119] S201, obtaining core data and logging data of the area to be tested.

[0120] Step S201 is the same as the above-mentioned step S101 and will not be described again here.

[0121] S202: Initialize the core data and the logging data to obtain initialized core data and logging data.

[0122] Step S202 is the same as the above-mentioned step S102 and will not be described again.

[0123] S203. Oil reservoir parameters include porosity, permeability, and water saturation. The initialized logging data is input into the random forest model, and the weight values ​​of the influence of multiple features on porosity, permeability, and water saturation are calculated respectively to obtain the porosity weight value, permeability weight value, and water saturation weight value.

[0124] Among them, a random forest model is constructed for each reservoir parameter: porosity, permeability, and water saturation. The well logging data is input into the random forest model for training, and the parameters of the random forest, such as the number of trees and the randomness of feature selection, are adjusted to obtain the best performance. The random forest model can provide the weight value of the influence of each feature of the well logging data on the reservoir parameters of porosity, permeability, and water saturation, and obtain the porosity weight value, permeability weight value, and water saturation weight value.

[0125] The Random Forest algorithm provides an effective mechanism to evaluate the importance of features. It measures the importance of features by calculating the average contribution of each feature across all decision trees in the Random Forest. Random Forest is an ensemble learning method that builds multiple decision trees and uses bootstrap sampling to generate each training subset during the construction of each decision tree. Due to the use of bootstrap sampling, some data may be missed in a single sampling. These data sets that are not selected for building a specific decision tree are called Out of Bag (OOB) data. Random Forest uses this OOB data to evaluate the performance of the model, providing a built-in evaluation mechanism without cross-validation.

[0126] Specifically, the steps for evaluating feature importance in random forests are usually as follows:

[0127] For each tree, the importance of the feature is calculated by comparing the difference in model performance when including and excluding the feature when predicting with OOB data; for each feature, this performance difference is accumulated and averaged over all trees to obtain the average importance score of the feature; by sorting the average importance scores of all features, the most important features and their contribution to the model's predictive ability can be determined. This feature importance assessment method based on random forests can not only provide guidance for feature selection, but also because it includes the randomness of the model and OOB estimates, it is generally considered to be a feature importance estimation method with lower bias.

[0128] In this study, random forest was used for feature selection, and the total importance of feature variables was normalized to 1. The results were arranged in descending order according to the importance of feature variables.

[0129] S204, determining whether there is a first associated feature with a correlation coefficient greater than a threshold value among the multiple features, if so, executing step S205; if not, executing step S206.

[0130] Among them, a formula is used to calculate whether there is a first associated feature with a correlation coefficient greater than a threshold among multiple features of the logging data. If there is a first associated feature with a correlation coefficient greater than the threshold, step S205 is executed; if there is no first associated feature with a correlation coefficient greater than the threshold, step S206 is executed.

[0131] Specifically, the correlation coefficients between the multiple features are calculated using the following formula:

[0132]

[0133] Among them, γ is the correlation coefficient, is the mean value of the logging curve, x i ,yi Represents the logging curve value corresponding to the i-th sample.

[0134] The absolute value of the correlation coefficient γ is calculated, and it is determined whether |γ| is greater than a threshold.

[0135] The correlation coefficient takes values ​​between [-1, 1], and different values ​​indicate different degrees of correlation.

[0136] Through correlation analysis, different degrees of correlation coefficients between each logging feature were obtained. The correlation coefficients were visualized on the heat map, such as Figure 3 shown.

[0137] Among them, according to Figure 3 The visualization of the correlation coefficient in the heat map determines whether there is a first associated feature with a correlation coefficient greater than a threshold among multiple features.

[0138] Specifically, Figure 3 The data presented in show that the correlation coefficient between the deep and shallow resistivity curves is 0.934, indicating that there is a strong correlation between them. This shows that there is a significant covariance between the two curves, making them unsuitable for simultaneous use as sensitive curves. The correlation coefficients between natural potential and deep and shallow resistivity are 0.336 and 0.311 respectively; the correlation coefficients between natural gamma and deep and shallow lateral resistivity are 0.405 and 0.35 respectively; the correlation coefficient between density logging and deep lateral resistivity reaches 0.309, and the correlation coefficient with acoustic time difference reaches 0.387.

[0139] Therefore, the deep and shallow resistivity curves are determined as the first correlation features with a correlation coefficient greater than a threshold.

[0140] S205. Compare the weight values ​​of at least two features in the first associated features. The first associated features with smaller weight values ​​in the first associated features do not participate in the weight value sorting process, and the first associated features with larger weight values ​​in the first associated features participate in the weight value sorting process.

[0141] Among them, when there is a first associated feature with a correlation coefficient greater than a threshold, the weight values ​​of at least two features in the first associated features are compared, and the first associated feature with a smaller weight value in the first associated features does not participate in the weight value sorting process, and the first associated feature with a larger weight value in the first associated features participates in the weight value sorting process.

[0142] Specifically, by Figure 3 The deep and shallow resistivity curve is determined as the first correlation feature with a correlation coefficient greater than a threshold. Figure 4 , Figure 5 , Figure 6It can be seen that the contribution of deep resistivity to porosity, permeability, and water saturation is lower than that of shallow resistivity, because in the subsequent weight value sorting process, deep resistivity does not participate in the sorting, while shallow resistivity participates in the sorting.

[0143] S206. Sort the porosity weight value, the permeability weight value, and the water saturation weight value from large to small.

[0144] The weight ranking of multiple features of logging data on porosity is shown in the attached Figure 4 .

[0145] The ranking of the weights of multiple features of logging data affecting permeability is shown in the attached Figure 5 .

[0146] The weight ranking of the impact of multiple features of logging data on water saturation is shown in the attached Figure 6 .

[0147] S207 , taking features corresponding to the first preset number of porosity weight values ​​after sorting as target features of porosity.

[0148] Among them, the features corresponding to the first preset number of porosity weight values ​​after sorting are used as the target features of porosity. Figure 4 , six features including “lithology density”, “acoustic time difference”, “natural gamma”, “natural potential”, “shallow lateral resistivity” and “depth” were selected as the target features of porosity.

[0149] S208. Use the features corresponding to the first preset number of permeability weight values ​​after sorting as the target features of the permeability.

[0150] Among them, the features corresponding to the first preset number of permeability weight values ​​after sorting are used as the target features of permeability. Figure 5 , six features including “porosity”, “natural potential”, “depth”, “acoustic time difference”, “density” and “natural gamma” were selected as the target features of permeability.

[0151] S209: Using the features corresponding to the first preset number of water saturation weight values ​​after sorting as target features of water saturation.

[0152] Among them, the features corresponding to the first preset number of water saturation weight values ​​after sorting are used as the target features of water saturation. Figure 6 , seven features, namely “permeability”, “natural potential”, “shallow resistivity”, “porosity”, “sonic time difference”, “natural gamma” and “density”, were selected as the target features of water saturation.

[0153] S210. Construct a long short-term memory network model input layer with a dimension corresponding to the number of features according to the number of features of the initialized core data.

[0154] The number of features in the initialized core data is counted, which will determine the dimension of the input layer of the LSTM network model. Use Python's deep learning library (such as Keras or PyTorch) to build the LSTM network model. In the input layer, set the dimension to the parameter of the number of features to adapt to core data with different numbers of features.

[0155] S211. Construct a long short-term memory network model output layer with a dimension corresponding to the number of features according to the number of features of the target features of the logging data.

[0156] The number of target features in the well logging data is determined, which will determine the dimension of the output layer of the LSTM network model. Use Python's deep learning library (such as Keras or PyTorch) to build the LSTM network model. In the output layer, set the dimension to the parameter of the number of target features to adapt to well logging data with different numbers of target features.

[0157] S212. Construct the number of units and layers of the long short-term memory network model according to the target features of the initialized core data and logging data.

[0158] Among them, the number of units and layers of the long short-term memory network model are constructed according to the target features of the initialized core data and logging data. The number of units refers to the number of LSTM units in the LSTM layer. More units can increase the representation ability of the model, but may also lead to overfitting. Usually, you can start with a smaller number of units and then gradually increase it to improve model performance. The number of layers refers to the number of LSTM layers in the LSTM network. Increasing the number of layers can increase the complexity and representation ability of the model, but it may also increase the training time and the risk of overfitting. For simpler tasks, you can usually start with a single layer. For more complex tasks, you can consider increasing the number of layers. Based on the above principles, you can select the appropriate number of units and layers according to the target features of the initialized core data and logging data to construct a long short-term memory network model with a specified number of units and layers.

[0159] S213. According to the target features of the initialized core data and logging data, the long short-term memory network model is trained to obtain a trained long short-term memory network model.

[0160] The model is trained using the target features of the initialized core data and logging data, and parameters such as the training batch size, number of iterations, and validation set are specified. The performance of the trained LSTM network model is evaluated using the test set data. After multiple iterations of training, the trained LSTM network model is obtained, such as Figure 7 shown.

[0161] S214. Use the initialized core data as input data of the trained long short-term memory network model, and use the target features of the logging data as output data of the trained long short-term memory network model, and control the long short-term memory network model to perform target feature correlation analysis and processing on the core data and the logging data to obtain the correlation relationship between the target features of the core data and the logging data.

[0162] Among them, the initialized core data is used as the input data of the trained long short-term memory network model, and the target features of the logging data are used as the output data of the trained long short-term memory network model. The long short-term memory network model is controlled to perform correlation analysis on the target features of the core data and the logging data, and the predicted results are compared and analyzed with the actual target features of the logging data to evaluate the correlation between the target features of the core data and the logging data.

[0163] S215. Construct a target prediction model based on the long short-term memory network model and rock physics knowledge, use the correlation between the target characteristics of the core data and the logging data and the rock physics knowledge as input data of the target prediction model, and control the target prediction model to perform iterative processing.

[0164] Among them, a target prediction model is constructed based on the long short-term memory network model and rock physics knowledge. The target prediction model combines the pure data-driven long short-term memory network neural network and the long short-term memory network neural network combined with rock physics knowledge constraints. The rock physics-based reservoir parameter prediction model is as follows Figure 7 shown.

[0165] Specifically, the two networks are consistent in optimizer, batch size and number of training rounds, using Adam optimizer, batch size 64 and training rounds 200 respectively. The pure data-driven LSTM network model uses the root mean square error (RMSE) as the loss function, while the physically constrained LSTM network model uses a comprehensive loss function that combines rock physics knowledge constraints and data constraints. In order to analyze the prediction effects of pure data-driven neural networks and physically constrained neural networks, the same network hyperparameters are used for the same prediction target. For different prediction targets, systematic hyperparameter adjustments are performed.

[0166] S216. Obtain the total loss function of the target prediction model.

[0167] Among them, the total loss function includes the first loss function of the long short-term memory network model and the second loss function of rock physics knowledge.

[0168] The total loss function includes: data-driven rock property parameter prediction error first loss function RMSE data And the second loss function RMSE of the prediction error of the neural network for domain knowledge p .

[0169] Specifically, the total loss function is defined as:

[0170] LOSS = α1RMSE data +α2RMSE p

[0171] in:

[0172]

[0173] LOSE p =|PK(x i ,y i )-MK(x i ,y i )|-ξ

[0174]

[0175] In the formula, RMSE data Represents the prediction error of rock parameters using data-driven reservoir parameter prediction, RMSE p represents the prediction error of the neural network on reservoir parameters based on petrophysical domain knowledge, represents the true value, PK(x i ,y i ) represents the reservoir physical property parameters calculated by the neural network based on the knowledge of rock physics, MK(x i ,y i ) represents the reservoir physical property parameters obtained by data-driven calculation of rock parameters, ξ is the allowable range of model error, and α1 and α2 are weight coefficients.

[0176] S217: When it is determined that the total loss function converges, the target prediction model is controlled to stop iterative processing, and it is determined that the training of the target prediction model is completed.

[0177] Before training the model, a minimum loss threshold can be set. When the loss function is lower than the threshold, the training of the target prediction model is stopped, and the training of the target prediction model is determined to be completed. Setting the loss function threshold can ensure that the model stops training after reaching a certain performance level to avoid overfitting.

[0178] This embodiment proposes a method for predicting oil reservoir parameters based on rock physics. The method inputs the initialized logging data into a random forest model, calculates the weight values ​​of the influence of multiple features in the logging data on porosity, permeability, and water saturation, and sorts them in descending order according to the weight values; uses the features corresponding to the first preset number of porosity weight values, permeability weight values, and water saturation weight values ​​after sorting as porosity target features, permeability target features, and water saturation target features; constructs a long short-term memory network model based on the initialized core data and logging data; and uses the initialized core data as the input of the trained long short-term memory network model. Data, logging data target features as the output data of the trained long short-term memory network model, control the long short-term memory network model to perform target feature correlation analysis and processing of core data and logging data, and obtain the correlation between core data and logging data target features; build a target prediction model based on the long short-term memory network model and rock physics knowledge, use the correlation between core data and logging data target features and rock physics knowledge as input data of the target prediction model, control the target prediction model to iterate; obtain the total loss function of the target prediction Moses, and when the total loss function converges, control the target prediction model to stop iterative processing and determine that the training of the target prediction model is completed. Avoid overfitting, improve prediction accuracy, and improve work efficiency.

[0179] Figure 8 The schematic diagram of the structure of the oil reservoir parameter prediction device based on rock physics provided in this application is as follows: Figure 8 As shown, the oil reservoir parameter prediction device 300 based on petrophysics provided in this embodiment includes:

[0180] An acquisition module 301 is used to acquire core data and logging data of the area to be tested;

[0181] The processing module 302 is used to perform initialization processing on the core data and the well logging data to obtain initialized core data and well logging data;

[0182] A calculation module 303 is used to input the initialized logging data into a random forest model, calculate the weights of multiple features of the logging data to oil reservoir parameters, and obtain target features of the logging data;

[0183] The processing module 302 is further used to construct a long short-term memory network model according to the initialized core data and the target features of the logging data, and the long short-term memory network model is used to determine the correlation relationship between the core data and the target features of the logging data;

[0184] The processing module 302 is further used to construct a target prediction model based on the long short-term memory network model and rock physics knowledge, and train the target prediction model, wherein the target prediction model predicts parameters of oil reservoirs based on rock physics.

[0185] Optionally, the petrophysics-based reservoir reservoir parameter prediction device further includes: a determination module 304;

[0186] The calculation module 303 is further used to input the initialized logging data into a random forest model, and respectively calculate the influence weight values ​​of the multiple features on the porosity, permeability, and water saturation to obtain a porosity weight value, a permeability weight value, and a water saturation weight value;

[0187] The processing module 302 is further used to sort the porosity weight value, the permeability weight value, and the water saturation weight value from large to small weight values;

[0188] The determination module 304 is used to use the features corresponding to the first preset number of porosity weight values ​​after sorting as the target features of porosity;

[0189] The determination module 304 is further configured to use the features corresponding to the first preset number of permeability weight values ​​after sorting as the target features of the permeability;

[0190] The determination module 304 is further configured to use the features corresponding to the first preset number of water saturation weight values ​​after sorting as target features of water saturation.

[0191] Optionally, the oil reservoir parameter prediction device based on petrophysics further includes: a judgment module 305;

[0192] The judging module 305 is used to judge whether there is a first associated feature with a correlation coefficient greater than a threshold value among the multiple features;

[0193] The processing module 302 is further configured to compare the weight values ​​of at least two features among the plurality of features when there is a first associated feature whose correlation coefficient is greater than a threshold value, so that the first associated feature with a smaller weight value among the first associated features does not participate in the weight value sorting process, and the first associated feature with a larger weight value among the first associated features participates in the weight value sorting process;

[0194] The processing module 302 is further configured to, when there is no first associated feature with a correlation coefficient greater than a threshold value among the multiple features, continue to sort the multiple weight values ​​in descending order of weight value.

[0195] Optionally, the calculation module 303 is further configured to calculate the correlation coefficient between the multiple features using the following formula:

[0196]

[0197] Among them, γ is the correlation coefficient, is the mean value of the logging curve, x i ,y i Represents the logging curve value corresponding to the i-th sample;

[0198] The calculation module 303 is further used to calculate the absolute value of the correlation coefficient γ;

[0199] The judging module 305 is further configured to judge whether |γ| is greater than a threshold.

[0200] Optionally, the processing module 302 is further used to construct a long short-term memory network model according to the target features of the initialized core data and the logging data;

[0201] The processing module 302 is further used to train the long short-term memory network model according to the target features of the initialized core data and the logging data to obtain a trained long short-term memory network model;

[0202] The processing module 302 is also used to use the initialized core data as input data of the trained long short-term memory network model, and the target features of the logging data as output data of the trained long short-term memory network model, and control the long short-term memory network model to perform correlation analysis and processing on the target features of the core data and the logging data, so as to obtain the correlation relationship between the core data and the target features of the logging data.

[0203] Optionally, the processing module 302 is further used to construct a long short-term memory network model input layer with a dimension corresponding to the number of features according to the number of features of the initialized core data;

[0204] The processing module 302 is further used to construct a long short-term memory network model output layer with a dimension corresponding to the number of features according to the number of features of the target features of the logging data;

[0205] The processing module 302 is further used to construct the number of units and layers of the long short-term memory network model according to the target features of the initialized core data and the logging data.

[0206] Optionally, the processing module 302 is further used to use the correlation between the core data and the target characteristics of the logging data and rock physics knowledge as input data of the target prediction model to control the target prediction model to perform iterative processing;

[0207] The acquisition module 301 is further used to acquire the total loss function of the target prediction model, wherein the total loss function includes a first loss function of a long short-term memory network model and a second loss function of petrophysical knowledge;

[0208] The processing module 302 is further used to control the target prediction model to stop iterative processing when it is determined that the total loss function converges, and to determine that the training of the target prediction model is completed.

[0209] Fig. 9 This is a schematic diagram of the structure of the oil reservoir parameter prediction device based on rock physics provided in this application. Fig. 9 As shown, the present application provides a petrophysics-based oil reservoir parameter prediction device, and the petrophysics-based oil reservoir parameter prediction device 400 includes: a receiver 401, a transmitter 402, a processor 403 and a memory 404.

[0210] Receiver 401, used for receiving instructions and data;

[0211] A transmitter 402, used for sending instructions and data;

[0212] Memory 404, for storing computer-executable instructions;

[0213] The processor 403 is used to execute the computer-executable instructions stored in the memory 404 to implement the various steps performed by the method for predicting oil reservoir parameters based on petrophysics in the above embodiment. For details, please refer to the relevant description in the above embodiment of the method for predicting oil reservoir parameters based on petrophysics.

[0214] Optionally, the memory 404 may be independent or integrated with the processor 403 .

[0215] When the memory 404 is independently provided, the electronic device further includes a bus for connecting the memory 404 and the processor 403 .

[0216] The present application also provides a computer storage medium, in which computer execution instructions are stored. When a processor executes the computer execution instructions, a petrophysics-based oil reservoir parameter prediction method as performed by the petrophysics-based oil reservoir parameter prediction device described above is implemented.

[0217] It will be appreciated by those skilled in the art that all or some of the steps, systems, and functional modules / units in the methods disclosed above may be implemented as software, firmware, hardware, and appropriate combinations thereof. In hardware implementations, the division between the functional modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed by several physical components in cooperation. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or implemented as hardware, or implemented as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or non-transitory medium) and a communication medium (or temporary medium). As known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, it is well known to those of ordinary skill in the art that communication media typically contain computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

[0218] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the invention disclosed herein. The present application is intended to cover any modification, use or adaptation of the present application, which follows the general principles of the present application and includes common knowledge or customary techniques in the art that are not disclosed in the present application. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present application are indicated by the following claims.

[0219] It should be understood that the present application is not limited to the precise structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. A method for predicting reservoir parameters based on rock physics, It is characterized in that The method comprises: Obtain core data and logging data of the area to be tested; Initializing the core data and the well logging data to obtain initialized core data and well logging data; Inputting the initialized logging data into a random forest model, calculating the weights of multiple features of the logging data to reservoir parameters, and obtaining target features of the logging data; According to the initialized core data and the target features of the logging data, a long short-term memory network model is constructed, wherein the long short-term memory network model is used to determine the correlation between the core data and the target features of the logging data; A target prediction model is constructed according to the long short-term memory network model and rock physics knowledge, and the target prediction model is trained. The target prediction model predicts parameters of oil reservoirs based on rock physics.

2. The method according to claim 1, It is characterized in that The reservoir parameters include porosity, permeability, and water saturation. The initialized logging data is input into a random forest model to calculate the weights of multiple features of the logging data on the reservoir parameters, and the target features of the logging data are obtained, including: Inputting the initialized logging data into a random forest model, respectively calculating the influence weight values ​​of the multiple features on the porosity, permeability, and water saturation, and obtaining a porosity weight value, a permeability weight value, and a water saturation weight value; The porosity weight value, the permeability weight value, and the water saturation weight value are sorted from large to small; The features corresponding to the first preset number of porosity weight values ​​after sorting are used as the target features of porosity; The features corresponding to the first preset number of permeability weight values ​​after sorting are used as the target features of the permeability; The features corresponding to the first preset number of water saturation weight values ​​after sorting are used as target features of water saturation.

3. The method according to claim 2, It is characterized in that Before sorting the porosity weight value, the permeability weight value, and the water saturation weight value from large to small, the method further includes: Determine whether there is a first associated feature with a correlation coefficient greater than a threshold value among the multiple features; If so, compare the weight values ​​of at least two features in the first associated features, the first associated features with smaller weight values ​​in the first associated features do not participate in the weight value sorting process, and the first associated features with larger weight values ​​in the first associated features participate in the weight value sorting process; If not, continue to sort the multiple weight values ​​in descending order of weight value.

4. The method according to claim 3, It is characterized in that The determining whether there is a first associated feature with a correlation coefficient greater than a threshold value among the multiple features includes: The correlation coefficient between the multiple features is calculated using the following formula: Among them, γ is the correlation coefficient, is the mean value of the logging curve, x i ,y i Represents the logging curve value corresponding to the i-th sample; The absolute value of the correlation coefficient γ is calculated, and it is determined whether |γ| is greater than a threshold.

5. The method according to claim 1, It is characterized in that The constructing of a long short-term memory network model according to the initialized core data and the target features of the logging data, wherein the long short-term memory network model is used to determine the correlation between the core data and the target features of the logging data, comprises: Constructing a long short-term memory network model according to the initialized core data and the target features of the logging data; According to the target features of the initialized core data and the logging data, a long short-term memory network model is trained to obtain a trained long short-term memory network model; The initialized core data is used as input data of the trained long short-term memory network model, and the target features of the well logging data are used as output data of the trained long short-term memory network model. The long short-term memory network model is controlled to perform correlation analysis on the target features of the core data and the well logging data to obtain the correlation relationship between the core data and the target features of the well logging data.

6. The method according to claim 5, It is characterized in that The initialized core data includes multiple features, and the long short-term memory network model is constructed according to the initialized core data and the target features of the logging data, including: According to the number of features of the initialized core data, a long short-term memory network model input layer with a dimension corresponding to the number of features is constructed; According to the number of features of the target features of the well logging data, a long short-term memory network model output layer with a dimension corresponding to the number of features is constructed; The number of units and the number of layers of the long short-term memory network model are constructed according to the target features of the initialized core data and the logging data.

7. The method according to claim 5, It is characterized in that The training of the target prediction model comprises: Using the correlation between the core data and the target features of the logging data and petrophysical knowledge as input data of the target prediction model, and controlling the target prediction model to perform iterative processing; Obtaining a total loss function of the target prediction model, wherein the total loss function includes a first loss function of a long short-term memory network model and a second loss function of petrophysical knowledge; When it is determined that the total loss function converges, the target prediction model is controlled to stop iterative processing, and it is determined that the training of the target prediction model is completed.

8. A device for predicting reservoir parameters based on rock physics. It is characterized in that The device comprises: An acquisition module is used to acquire core data and logging data of the area to be tested; A processing module, used for initializing the core data and the well logging data to obtain initialized core data and well logging data; A calculation module, used for inputting the initialized logging data into a random forest model, calculating the weights of multiple features of the logging data to reservoir parameters, and obtaining target features of the logging data; The processing module is further used to construct a long short-term memory network model according to the initialized core data and the target features of the logging data, and the long short-term memory network model is used to determine the correlation relationship between the core data and the target features of the logging data; The processing module is also used to construct a target prediction model based on the long short-term memory network model and rock physics knowledge, and train the target prediction model, wherein the target prediction model predicts parameters of oil reservoirs based on rock physics.

9. A device for predicting reservoir parameters based on rock physics. It is characterized in that include: Memory; processor; Wherein, the memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the petrophysics-based method for predicting oil reservoir parameters as described in any one of claims 1-7.

10. A computer storage medium, It is characterized in that The computer storage medium stores computer execution instructions, which, when executed by a processor, are used to implement the petrophysics-based reservoir parameter prediction method according to any one of claims 1 to 7.

Citation Information

Patent Citations

  • Reservoir logging characteristic parameter extraction method based on target preference coding

    CN111983722A

  • Complex lithology identification method based on long-term and short-term memory neural network

    CN112712025A

  • Carbonate reservoir sedimentary facies identification method and device, electronic equipment and medium

    CN113627607A

  • Method for realizing logging lithology identification based on convolution circulation deep learning network of self-attention mechanism

    CN116680638A

  • Oil reservoir development dynamic characterization method and device and related equipment

    CN117272038A

Cited By

  • Coal mine underground operation measurement mode and drilling parameter adjustment method and device, medium and product

    CN119507985A

  • A coal mine underground operation measurement mode and drilling parameter adjustment method, device, medium and product

    CN119507985B

  • Intelligent reservoir logging evaluation method driven by double models

    CN119807888A

  • Shale oiliness determination method and device, equipment and storage medium

    CN120296416A

  • Method and device for determining formation porosity and storage medium

    CN120468965A