Intelligent identification method and device for logging facies, electronic equipment and storage medium
Train the logging curve data through deep learning network models, solving the difficulties of traditional methods in fine logging phase recognition, and achieving high-precision and efficient logging phase recognition and classification.
Patent Information
- Application Number
- CN202311593162.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-11-27
- Publication Date
- 2025-05-27
AI Technical Summary
Traditional logging lithophase recognition methods have difficulties in fine logging phase recognition, making it difficult to achieve high-precision and efficient identification.
The deep learning network model is used to train the logging curve data, and the automatic identification and classification of target logging data is achieved by establishing the training sample set and the optimal network model.
It improves the accuracy of lithophase recognition, reduces the error of graph identification, realizes automatic interpretation of well logging phases, and is suitable for intelligent interpretation of large-scale logging data.
Smart Images
Figure CN120047706A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of geophysical logging data interpretation, and more specifically, relates to a method, device, electronic device and storage medium for intelligent identification of logging facies. Background Art
[0002] How to use geological, logging and other data to achieve rapid and efficient identification of logging lithofacies has always been a concern of reservoir geologists. Currently, most traditional methods are based on geological data such as logging and core, and combined with theoretical charts to manually divide the logging lithofacies of key wells. Then, through multi-parameter cross-plot analysis of the logging responses of different lithofacies, a logging lithofacies identification chart and quantitative interpretation criteria are established, and then the logging lithofacies of non-cored wells are discriminated and divided. Traditional methods can better distinguish large logging lithofacies, but there are difficulties in more refined logging facies identification. Summary of the Invention
[0003] The object of the present invention is to propose a method, device, electronic device and storage medium for intelligent identification of logging facies, so as to improve the accuracy of lithofacies identification and the efficiency of interpretation work.
[0004] To achieve the above object, in the first aspect, the present invention proposes an intelligent identification method for logging facies, including:
[0005] Preparing a training sample set, each training sample including a feature vector and a label value, the feature vector including at least one logging curve data, and the label value being the logging facies type corresponding to the logging curve data;
[0006] Training a deep learning network model using the training sample set to obtain an optimal network model;
[0007] Identifying the logging facies type of target logging data using the optimal network model.
[0008] Optionally, the logging facies type as the label value adopts a one-hot encoding method.
[0009] Optionally, the training of the deep learning network model using the training sample set includes:
[0010] Under the Bayesian framework, by iteratively calculating the maximum likelihood function of the training sample set, the weights of each class output of the logging lithofacies classification function and the hyperparameters for training are optimized to obtain the optimal network model.
[0011] Optionally, the logging facies classification function is defined as:
[0012]
[0013] Wherein, K(x n , x i ) is the kernel function, w k,i is the weight to be determined, N is the number of samples, x n is the nth training sample, x i is the training data of the ith sample, y k is the classification function, w k,0 is the weight bias, k is the category to which the logging facies type belongs, w k is the weight vector corresponding to the output of the kth category.
[0014] Optionally, the maximum likelihood function is:
[0015]
[0016] The likelihood function of the entire training sample set is:
[0017]
[0018] The Gaussian prior probability distribution constraint parameters defined by each weight w are:
[0019]
[0020] Wherein, t n,k is the one-hot encoded value of the nth sample belonging to the kth logging facies, y k,n is the classification function prediction of the nth sample belonging to the kth category, w k,n is the weight to be determined, α k,n is the hyperparameter corresponding to w k,n , C is the total number of logging facies types, σ is a connection function applied to the output y function, and α is the hyperparameter for training.
[0021] Optionally, the identifying the logging facies type of the target logging data by using the optimal network model includes:
[0022] Calculating the logging facies classification information of the target logging data by using the logging lithofacies classification function with determined weights;
[0023] Calculating the posterior probability information of the logging classification of the target logging data by using the posterior probability calculation formula, and taking the category with the largest posterior probability as the logging facies category to which the target logging data belongs.
[0024] Optionally, the posterior probability calculation formula is:
[0025]
[0026] Wherein, σ(y) is a connection function applied to the output y function, y(x i; x is the target logging data under the classification weight w i The corresponding logging facies classification.
[0027] In a second aspect, the present invention provides an intelligent recognition device for logging facies, comprising:
[0028] A training sample making module, configured to prepare a training sample set, each training sample including a feature vector and a label value, the feature vector including at least one logging curve data, and the label value being the logging facies type corresponding to the logging curve data;
[0029] A model training module, configured to train a deep learning-based network model using the training sample set to obtain an optimal network model;
[0030] A logging facies recognition module, configured to recognize the logging facies type of the target logging data using the optimal network model.
[0031] In a third aspect, the present invention provides an electronic device, the electronic device comprising:
[0032] At least one processor; and,
[0033] A memory communicatively connected to the at least one processor; wherein,
[0034] The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the intelligent recognition method for logging facies according to any one of the first aspects.
[0035] In a fourth aspect, the present invention provides a non-transitory computer-readable storage medium, which stores computer instructions for causing a computer to execute the intelligent recognition method for logging facies according to any one of the first aspects.
[0036] The beneficial effects of the present invention are as follows:
[0037] The intelligent identification method of well logging facies of the present invention realizes a well logging facies analysis and identification method based on deep learning. By establishing a sample label data set with known well logging facies as samples, a learning model is established using a deep learning network to achieve data classification, thereby completing the automatic division and identification of well logging facies, realizing the automatic interpretation of well logging facies using conventional well logging data, reducing the error of lithofacies identification by the chart method. The method of the present invention can realize the intelligent identification and division of well logging facies based on conventional well logging curves, provide accurate geological information for sedimentary facies analysis and reservoir prediction, and can complete the well logging facies interpretation of multiple wells in a short time, be applicable to the intelligent interpretation of large-scale well logging data, and can quickly, efficiently and accurately realize the identification and classification of well logging facies, thereby helping geological personnel quickly obtain the longitudinal development of formation lithofacies, etc., and laying a foundation for further research on sedimentary facies and reservoir characteristics.
[0038] The system of the present invention has other characteristics and advantages, which will be obvious from the accompanying drawings incorporated herein and the subsequent detailed description, or will be described in detail in the accompanying drawings incorporated herein and the subsequent detailed description. These accompanying drawings and detailed description are used together to explain the specific principles of the present invention. Brief Description of the Drawings
[0039] By describing the exemplary embodiments of the present invention in more detail in conjunction with the accompanying drawings, the above and other objects, features and advantages of the present invention will become more obvious. In the exemplary embodiments of the present invention, the same reference numerals generally represent the same components.
[0040] Figure 1 The flowchart showing the steps of an intelligent identification method of well logging facies according to the present invention is shown.
[0041] Figure 2 The schematic diagram showing the known conventional well logging curve samples and well logging facies labels in an intelligent identification method of well logging facies in Embodiment 2 of the present invention is shown.
[0042] Figure 3 The comparison diagram showing the predicted well logging facies and the original well logging facies in an intelligent identification method of well logging facies in Embodiment 2 of the present invention is shown. Detailed Description of the Embodiments
[0043] The present invention will be described in more detail below with reference to the accompanying drawings. Although the preferred embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to make the present invention more thorough and complete, and to convey the scope of the present invention fully to those skilled in the art.
[0044] Embodiment 1
[0045] As Figure 1As shown, this embodiment provides an intelligent identification method for logging facies, including:
[0046] S1: Prepare a training sample set. Each training sample includes a feature vector and a label value. The feature vector includes at least one logging curve data, and the label value is the logging facies type corresponding to the logging curve data.
[0047] Preferably, the logging facies type as the label value adopts the one-hot encoding method.
[0048] S2: Use the training sample set to train a deep learning network model to obtain an optimal network model.
[0049] In this step, training the deep learning network model with the training sample set includes:
[0050] Under the Bayesian framework, by performing iterative calculations on the maximum likelihood function of the training sample set, the weights of each class output of the logging lithofacies classification function and the hyperparameters for training are optimized to obtain an optimal network model.
[0051] Among them, the logging facies classification function is defined as:
[0052]
[0053] In the formula, K(x n ,x i ) is the kernel function, w k,i is the undetermined weight, N is the number of samples, x n is the nth training sample, x i is the training data of the ith sample, y k is the classification function, w k,0 is the weight bias, k is the category to which the logging facies type belongs, and w k is the weight vector corresponding to the output of the kth class.
[0054] The maximum likelihood function is:
[0055]
[0056] The likelihood function of the entire training sample set is:
[0057]
[0058] The constraint parameters of the Gaussian prior probability distribution defined by each weight w are:
[0059]
[0060] Among them, t n,k is the one-hot encoding value of the nth sample belonging to the kth logging facies, and y k,nThe classification function prediction for the nth sample belonging to the kth class, w k,n is the weight to be determined, α k,n is for w k,n corresponding hyperparameter, C is the total number of log facies types, σ is a connection function applied to the output y function, and α is a hyperparameter for training.
[0061] S3: Use the optimal network model to identify the log facies type of the target log data.
[0062] This step includes:
[0063] Calculate the log facies classification information of the target log data using the log lithofacies classification function with determined weights;
[0064] Calculate the posterior probability information of the log classification of the target log data using the posterior probability calculation formula, and take the class with the maximum posterior probability as the log facies class to which the target log data belongs.
[0065] Among them, the posterior probability calculation formula is:
[0066]
[0067] Among them, σ(y) is a connection function applied to the output y function, and y(x i ; w) is the log facies classification corresponding to the target log data x i under the classification weight w.
[0068] Example 2
[0069] This embodiment provides an intelligent identification method for log facies. Using the well data of known log facies in the target work area as training data, training is carried out by establishing a deep learning network model, and log facies intelligent identification is realized by reasoning on the target well based on the optimal network model obtained from training. The specific steps are as follows:
[0070] 1) Network training of known data
[0071] Given a set of data with m features (i.e., data corresponding to m conventional log curves) as training data, where N is the number of samples, and the log facies type corresponding to each sample point is label data. Figure 2 Shows the samples of known conventional log curves (left 3 columns) and log facies (right column) labels for model training.
[0072] Among them, the log facies type adopts the one-hot encoding method, that is, it is represented by a C-bit output vector to indicate which class it belongs to. When it belongs to the kth class, the kth bit is 1, and the other bits are 0 accordingly. The output t of each class k has its corresponding weight vector wk and the hyperparameter α for training k . Then the classification function of the logging lithofacies data can be defined as:
[0073]
[0074] where K(x n , x i ) is the kernel function, and w k,i is the undetermined weight.
[0075] Since the training samples have independent distribution characteristics, the likelihood function of the entire sample set is expressed as:
[0076]
[0077] To avoid overfitting problems, Gaussian prior probability distribution constraints are defined for each weight w:
[0078]
[0079] Thus, the maximized likelihood function is obtained:
[0080]
[0081] In the Bayesian framework, the estimation of the parameter vector w can be achieved by maximizing the likelihood function. That is, through iterative calculation of the above formula, the optimization of the parameters w and the hyperparameter α is realized. Finally, most of the w i tend to 0, and a small number of w i tend to stable finite values, realizing model sparsification and obtaining the optimal network model.
[0082] 2) Identification of logging facies of unknown target data
[0083] After obtaining the weight w (i.e., the optimal network model), use the posterior probability calculation formula to calculate the posterior probability information for the target data x of the unknown logging facies type i , and use the formula to take the category with the largest posterior probability as the logging facies category to which the data x i belongs, thereby realizing intelligent identification of logging facies. Figure 3 Shows the comparison between the logging facies (red) obtained by intelligent prediction and the original logging facies (black), and it can be seen that there is a high correlation between the two.
[0084] This method can realize intelligent identification and classification of logging facies based on conventional logging curves, providing accurate geological information for sedimentary facies analysis and reservoir prediction.
[0085] Example 3
[0086] This example provides an intelligent identification device for logging facies, including:
[0087] A training sample preparation module for preparing a training sample set, where each training sample includes a feature vector and a label value. The feature vector includes at least one logging curve data, and the label value is a logging facies type corresponding to the logging curve data.
[0088] A model training module for training a deep learning-based network model using the training sample set to obtain an optimal network model.
[0089] A logging facies identification module for identifying the logging facies type of target logging data using the optimal network model.
[0090] For the specific functions of each module in this embodiment, reference may be made to Embodiment 1.
[0091] Embodiment 4
[0092] This embodiment provides an electronic device, which includes:
[0093] At least one processor; and,
[0094] A memory communicatively connected to the at least one processor; wherein,
[0095] The memory stores instructions executable by the at least one processor. When the instructions are executed by the at least one processor, the at least one processor is enabled to execute the intelligent identification method of the logging facies described in any of the above embodiments.
[0096] The electronic device according to an embodiment of the present disclosure includes a memory and a processor. The memory is used to store non-temporary computer-readable instructions. Specifically, the memory may include one or more computer program products, and the computer program products may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory, etc. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.
[0097] The processor may be a central processing unit (CPU) or other forms of processing units with data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device to perform desired functions. In an embodiment of the present disclosure, the processor is used to run the computer-readable instructions stored in the memory.
[0098] Those skilled in the art should be able to understand that, in order to solve the technical problem of how to obtain good user experience effects, the present embodiment may also include well-known structures such as communication buses and interfaces, and these well-known structures should also be included in the protection scope of the present disclosure.
[0099] For the detailed description of this embodiment, reference may be made to the corresponding descriptions in the foregoing embodiments, which will not be repeated here.
[0100] Embodiment 5
[0101] The present invention provides a non-transitory computer-readable storage medium that stores computer instructions for causing a computer to execute the intelligent identification method of logging facies described in any of the foregoing embodiments.
[0102] According to an embodiment of the present disclosure, a non-transitory computer-readable instruction is stored on a computer-readable storage medium. When the non-transitory computer-readable instruction is run by a processor, all or part of the steps of the methods of the foregoing embodiments of the present disclosure are executed.
[0103] The above computer-readable storage media include, but are not limited to: optical storage media (such as CD-ROMs and DVDs), magneto-optical storage media (such as MOs), magnetic storage media (such as magnetic tapes or external hard drives), media with built-in rewritable non-volatile memories (such as memory cards), and media with built-in ROMs (such as ROM cartridges).
[0104] The various embodiments of the present invention have been described above. The above description is exemplary and not exhaustive, and is not limited to the disclosed embodiments. Many modifications and variations are obvious to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments.
Claims
1. An intelligent identification method for logging facies, characterized in that, it includes: Preparing a training sample set, each training sample includes a feature vector and a label value, the feature vector includes at least one logging curve data, and the label value is the logging facies type corresponding to the logging curve data; Using the training sample set to train a deep learning network model to obtain an optimal network model; Using the optimal network model to identify the logging facies type of target logging data.
2. The intelligent identification method for logging facies according to claim 1, characterized in that, The logging facies type as the label value adopts a one-hot encoding method.
3. The intelligent identification method for logging facies according to claim 1, characterized in that, The training of the deep learning network model using the training sample set includes: Under the Bayesian framework, by iteratively calculating the maximum likelihood function of the training sample set, the weights of each class output of the logging lithofacies classification function and the training hyperparameters are optimized to obtain the optimal network model.
4. The intelligent identification method for logging facies according to claim 3, characterized in that, The logging facies classification function is defined as: where K(x n , x i ) is the kernel function, w k,i is the weight to be determined, N is the number of samples, x n is the nth training sample, x i is the training data of the ith sample, y k is the classification function, w k,0 is the weight bias, k is the category to which the logging facies type belongs, w k is the weight vector corresponding to the output of the kth class.
5. The intelligent identification method for logging facies according to claim 4, characterized in that, The maximum likelihood function is: The likelihood function of the entire training sample set is: The constraint parameters of the Gaussian prior probability distribution defined by each weight w are: where t n,k is the one - hot encoding value that the n - th sample belongs to the k - th logging facies, y k,n is the classification function prediction that the n - th sample belongs to the k - th class, w k,n is the weight to be determined, α k,n is the corresponding hyperparameter of w k,n C is the total number of logging facies types, σ is a connection function applied to the output y function, and α is a hyperparameter for training.
6. The intelligent identification method for logging facies according to claim 5, characterized in that, The identification of the logging facies type of target logging data using the optimal network model includes: Calculating the logging facies classification information of the target logging data using the logging lithofacies classification function with determined weights; Calculating the posterior probability information of the logging classification of the target logging data using the posterior probability calculation formula, and taking the class with the maximum posterior probability as the logging facies class to which the target logging data belongs.
7. The intelligent identification method for logging facies according to claim 6, characterized in that, The posterior probability calculation formula is: where σ(y) is a link function applied to the output y function, and y(x i ; w) is the log facies classification corresponding to the target log data x i under the classification weight w.
8. An intelligent identification device for logging facies, characterized in that, it includes: A training sample making module for preparing a training sample set, each training sample includes a feature vector and a label value, the feature vector includes at least one logging curve data, and the label value is the logging facies type corresponding to the logging curve data; A model training module for training a network model based on deep learning using the training sample set to obtain an optimal network model; A logging facies identification module for identifying the logging facies type of target logging data using the optimal network model.
9. An electronic device, characterized in that, The electronic device includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the intelligent identification method for logging facies according to any one of claims 1-7.
10. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium stores computer instructions for causing a computer to execute the intelligent identification method of the logging facies according to any one of claims 1-7.