Deposition type identification model establishment, deposition type identification method and related apparatus

Abstract

The application discloses a deposition type identification model establishment method, a deposition type identification method and related devices. The deposition type identification model establishment method comprises the following steps: obtaining a sample set according to a plurality of selected well logging curves and deposition type data of at least one well, wherein a sample in the sample set comprises a curve segment of each selected well logging curve and a deposition type in a certain depth range of a corresponding well; setting a network structure and network parameters of a deep neural network according to a set rule; and training the deep neural network after the setting by using a periodic attenuation type learning rate of the sample set, so as to obtain a deposition type identification model used for predicting deposition type data of a target well. The model established by the method can quickly and reasonably complete the prediction of the deposition type data of the target well based on well data. The loss function is optimized by using the periodic attenuation type learning rate, so that the convergence speed of the loss function is improved, and the loss function has strong ability to escape from a saddle point and a local minimum value, thereby improving the prediction efficiency and accuracy.

Description

Technical Field

[0001] This invention relates to the field of geophysical technology, and in particular to the establishment of sediment type identification models, sediment type identification methods, and related devices. Background Technology

[0002] Sedimentary facies and lithology classification is an important research area in geophysical data interpretation, providing technical support for reservoir prediction and bioherm discovery, and serving as a prerequisite for subsequent refined data interpretation. For example, it can provide facies constraints for subsequent inversion work, thereby improving inversion accuracy, or be used for well-seismic calibration and sequence framework establishment, thereby providing support for low-frequency seismic prediction. In the field of geophysics, traditional sedimentary facies and lithology classification methods mostly utilize waveform analysis, seismic attribute calculation, and manual interpretation of seismic data. The limitations of these methods are reflected in the following aspects: (1) There are many processing and interpretation steps, and the process is complex. Errors generated in each step may be passed on and accumulated at each level, affecting the final accuracy; (2) They rely too much on data quality and the professional knowledge and subjective judgment of the interpreters regarding the research area; (3) They have low timeliness when dealing with large amounts of data; (4) There are difficulties in knowledge accumulation, improvement, and transfer. Summary of the Invention

[0003] The inventors discovered that in recent years, with the improvement of computing power, the rise of related algorithm research, and the accumulation of large amounts of data, intelligent methods have demonstrated efficiency and accuracy advantages surpassing human capabilities in some application areas such as classification, regression, dimensionality reduction, and clustering. Taking deep learning as an example, which contains more nonlinear layers and has powerful feature extraction and representation capabilities, the learning rate, as an important hyperparameter, is usually set manually as a small constant or a decaying array to optimize the loss function. However, when using stochastic gradient descent optimization methods in deep learning, the above two learning rate settings may result in a large gap between the found local minimum and the true global minimum, and may even lead to getting stuck at a saddle point in the loss function, affecting the model's prediction accuracy. In addition, to ensure successful model training, a small learning rate is usually given, which will reduce training efficiency, especially in the later stages of training. While setting a large learning rate manually may improve learning efficiency in the early stages, it may cause the loss function to fail to converge, leading to training failure.

[0004] Meanwhile, well data features high vertical resolution and encompasses a variety of subsurface physical properties. Therefore, a deep learning-based intelligent method is proposed to classify sedimentary facies and lithology based on well data, in order to meet the production demands of increasingly concealed exploration targets and ever-increasing exploration accuracy requirements.

[0005] In view of the above problems, this invention is proposed to provide a sedimentation type identification model establishment, sedimentation type identification method and related apparatus to overcome or at least partially solve the above problems. The model established by this method can quickly and reasonably predict the sedimentation type data of the target well based on well data; the loss function is optimized by using a periodically decaying learning rate, which improves the convergence speed of the loss function and has a strong ability to escape saddle points and local minima, thereby improving the prediction accuracy.

[0006] In a first aspect, embodiments of the present invention provide a method for establishing a deposition type identification model, comprising:

[0007] A sample set is obtained based on multiple selected logging curves and sediment type data from at least one well. The samples in the sample set include curve segments and sediment types of each selected logging curve within a certain depth range of the corresponding well.

[0008] The deep neural network structure and parameters are set according to the set rules, and the set deep neural network is trained with a periodically decaying learning rate using the sample set to obtain a sedimentation type identification model for predicting sedimentation type data of the target well.

[0009] Secondly, embodiments of the present invention provide a deposition type identification method, including:

[0010] Input the selected logging curves of the target well into the sedimentation type identification model established by the above method, and predict the sedimentation type data of the target well based on the output results of the model.

[0011] Thirdly, embodiments of the present invention provide a deposition type identification model establishment apparatus, comprising:

[0012] The sample set establishment module is used to obtain a sample set based on multiple selected logging curves and sedimentation type data from at least one well. The samples in the sample set include curve segments and sedimentation types of each selected logging curve within a certain depth range of the corresponding well.

[0013] The sedimentation type identification model building module is used to set the network structure and network parameters of the deep neural network according to the set rules, and to train the set deep neural network with a periodically decaying learning rate through the sample set to obtain a sedimentation type identification model for predicting sedimentation type data of the target well.

[0014] Fourthly, embodiments of the present invention provide a computer program product, including a computer program / instruction, wherein the computer program / instruction, when executed by a processor, implements the above-mentioned deposition type identification model establishment method, or implements the above-mentioned deposition type identification method.

[0015] Fifthly, embodiments of this disclosure provide a server, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the above-described deposition type identification model establishment method or the above-described deposition type identification method.

[0016] The beneficial effects of the above-described technical solutions provided in the embodiments of the present invention include at least the following:

[0017] The sedimentation type identification model establishment method provided in this invention can quickly and reasonably predict the sedimentation type data of the target well based on well data; the loss function is optimized by using a periodically decaying learning rate, which improves the convergence speed of the loss function and has a strong ability to escape saddle points and local minima, thereby improving prediction efficiency and accuracy.

[0018] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings.

[0019] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0020] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0021] Figure 1 This is a flowchart of the deposition type identification model establishment method in Embodiment 1 of the present invention;

[0022] Figure 2 This is a schematic diagram of the deep neural network structure in Embodiment 1 of the present invention;

[0023] Figure 3 This is a schematic diagram of the periodically decaying learning rate in Embodiment 1 of the present invention;

[0024] Figure 4 for Figure 1 The flowchart of the specific implementation of training the deep neural network in step S12;

[0025] Figure 5 This is a flowchart illustrating the specific implementation of the deposition type identification model establishment method in Embodiment 2 of the present invention.

[0026] Figure 6 This is a schematic diagram of well logging curve data visualization in Embodiment 2 of the present invention;

[0027] Figure 7 This is a cross-plot of various types of well logging curve values ​​and lithology types in Embodiment 2 of the present invention;

[0028] Figure 8 This is a bar chart showing the distribution of the number of each label sample in Embodiment 2 of the present invention;

[0029] Figure 9 This is a graph showing the change of loss function / prediction accuracy of the training set and validation set with the number of training rounds in Embodiment 2 of the present invention.

[0030] Figure 10 A comparison chart showing the actual lithological results, the predicted results obtained through various machine learning methods, and the predicted results obtained in Example 2 of this invention;

[0031] Figure 11 A comparison chart showing the prediction results and the true categories using unsupervised machine learning algorithms, supervised neural network algorithms (Example 3), and semi-supervised algorithms;

[0032] Figure 12 This is a schematic diagram of the structure of the deposition type identification model establishment device in an embodiment of the present invention. Detailed Implementation

[0033] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0034] It should be understood that the terminology used herein is merely for describing particular embodiments and is not intended to limit the invention. Unless otherwise stated, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. While only preferred methods and materials have been described herein, any methods and materials similar or equivalent to those described herein may be used in the implementation or testing of this invention. All references to this specification are incorporated by way of citation to disclose and describe methods and / or materials associated with those references. In the event of any conflict with any incorporated reference, the content of this specification shall prevail.

[0035] To address the problem of difficulty in quickly and reasonably identifying sediment types in existing technologies, embodiments of the present invention provide sediment type identification model establishment, sediment type identification method and related apparatus, which can quickly and reasonably predict the sediment type data of the target well based on well data.

[0036] Example 1

[0037] Embodiment 1 of the present invention provides a method for establishing a deposition type identification model, the process of which is as follows: Figure 1 As shown, it includes the following steps:

[0038] Step S11: Obtain a sample set based on multiple selected logging curves and sedimentation type data from at least one well.

[0039] The sample set includes curve segments and sedimentary types for each selected well logging curve within a certain depth range of the corresponding well. The sedimentary type can be either sedimentary facies or lithology. Each sedimentary type is unique within a certain depth range.

[0040] Furthermore, multiple selected logging curves and sedimentary type data for the well can include logging curves such as natural gamma, resistivity, density porosity, and neutron porosity, as well as some artificially interpreted sedimentary facies or lithological classification results; logging curves can also include logging curves such as P-wave impedance, Poisson's ratio, density, and P-wave and S-wave velocities, which are obtained by further calculation from conventional logging curves.

[0041] First, the obtained logging curves are preprocessed, including outlier removal, normalization, and mean zeroing.

[0042] The selection of the logging curve type can be achieved by first obtaining multiple logging curves and sediment type data from at least one well, and then filtering out the selected logging curve type through correlation analysis between the logging curve type and the sediment type.

[0043] In some embodiments, after obtaining the sample set, the method may further include determining whether the sample set meets the training quantity requirement; if not, supplementing the sample set by selecting a machine learning method.

[0044] For a sample set to meet the training quantity requirements, firstly, the total number of samples must meet the quantity requirement; secondly, it may also include drawing a distribution histogram of the sediment types included in the sample set to determine that the number of samples corresponding to each sediment type is roughly equal. If the number of samples corresponding to one or more sediment types is significantly less than the number of samples corresponding to other sediment types, then it is necessary to supplement the sediment types with fewer samples.

[0045] Step S12: Set the network structure and network parameters of the deep neural network according to the set rules, and train the set deep neural network with a periodically decaying learning rate through the sample set to obtain a sedimentation type identification model for predicting sedimentation type data of the target well.

[0046] Setting the network structure of a deep neural network according to predefined rules may include setting the number of input neurons of the deep neural network based on the number of types of selected logging curves, and setting the number of output neurons of the deep neural network based on the number of types of sediments included in the sample set.

[0047] Even better, the number of input neurons in the deep neural network is set to be the same as the number of types of selected logging curves; the number of output neurons in the deep neural network is set to be the same as the number of types of sedimentation types included in the sample set.

[0048] Configure the network parameters of the deep neural network according to the set rules, which may include:

[0049] Set the activation function of the deep neural network to the softmax function:

[0050]

[0051] Where, x i S(x) represents the probability that the output belongs to the i-th class. i ) represents the activation function value corresponding to the probability of the output being of class i, and n is the total number of classes.

[0052] The softmax function normalizes the output components corresponding to each category, making the sum of the components equal to 1.

[0053] Set the loss function of the deep neural network to the cross-entropy loss function:

[0054]

[0055] Where J is the cross-entropy loss function value, y i Let y be the label for the i-th class. If the class is i, then y i =1, otherwise equals 0.

[0056] See Figure 2 The diagram shown is a schematic of the network structure of a deep neural network; see also... Figure 3 The diagram shows a periodically decaying learning rate, where the horizontal axis represents the training rounds and the vertical axis represents the learning rate value for the corresponding training round.

[0057] Furthermore, the learning rate exhibits a periodic decay pattern, meaning that the learning rate changes periodically as the number of iterations increases; at the same time, each large cycle contains several small cycles, and within a large cycle, the peak value of each small cycle decays from large to small.

[0058] For periodically decaying learning rates, it is necessary to pre-set the number of small cycles contained in a large cycle, the width of the small cycle (i.e., the number of training rounds covered by the small cycle), the maximum peak value of the small cycle, and the peak decay rate.

[0059] In some embodiments, after setting the network structure and parameters of the deep neural network, the specific model training process is described in [reference needed]. Figure 4 As shown, the following steps may be included:

[0060] Step S121: Divide the sample set into a training set and a test set.

[0061] For example, the training and test sets can be randomly divided in an 8:2 ratio; alternatively, other ratios can be used.

[0062] Step S122: Using a periodically decaying learning rate on the training set, iteratively train the set deep neural network according to the set initial training rounds, and determine the optimal training rounds based on the output training logs.

[0063] The initial training epochs are often set to be relatively large so that the curve of the loss function changing with each epoch can be seen to show the process of the validation set loss function changing from large to small and then back to large.

[0064] Based on the output training logs, the training epoch corresponding to the point where the loss function of the validation set just reaches its minimum with increasing training epochs can be taken as the optimal training epoch.

[0065] Step S123: According to the preferred training rounds, the set deep neural network is iteratively trained using a periodically decaying learning rate through the training set to obtain a sedimentation type identification model for predicting sedimentation type data of the target well.

[0066] The sedimentation type identification model establishment method provided in Embodiment 1 of this invention can quickly and reasonably predict the sedimentation type data of the target well based on well data; the loss function is optimized by using a periodically decaying learning rate, which improves the convergence speed of the loss function and has a strong ability to escape saddle points and local minima, thereby improving prediction efficiency and accuracy.

[0067] Example 2

[0068] Embodiment 2 of the present invention provides a specific implementation flow of a method for establishing a deposition type identification model, as follows: Figure 5 As shown, it includes the following steps:

[0069] Step S51: Input natural gamma, spontaneous potential, well diameter, shallow / medium / deep resistivity, acoustic wave, neutron porosity, density porosity logging curves and lithological interpretation results.

[0070] Step S52: Perform outlier removal, mean zeroing, and variance normalization on the input well logging curves.

[0071] Mean normalization means that the average value of the processed logging curve is zero, which can reduce the amount of data and calculation, facilitate data management, and facilitate training and learning; variance unitization is normalization.

[0072] The above processing of the logging curves facilitates subsequent data visualization, data optimization, and model training.

[0073] Step S53: Visualize logging curves, plot the intersection of logging curve values ​​for each type and sedimentation type, filter sensitive logging curve types, and establish a sample set for training the sedimentation type identification model.

[0074] Well logging curve data visualization reference Figure 6 The image shows a visualization of logging curves for natural gamma, resistivity, and two types of porosity.

[0075] Taking lithological type as an example, the sedimentary type is referenced. Figure 7 The figure shows the intersection of various types of logging curve values ​​and sediment types; the logging curves are selected for subsequent training based on the quality and quantity of data, the correlation between data, and the distinguishability between data and targets.

[0076] The logging curves used in Example 2 include: natural gamma ray, spontaneous potential, well diameter, resistivity, density porosity, neutron porosity, and acoustic wave. After data preprocessing and cross-plot analysis, natural gamma ray, resistivity, density porosity, and neutron porosity were selected for use as input to the neural network. The artificially interpreted lithological labels corresponding to the input data include nine types: barrier limestone, argillaceous limestone, dolomite, grainy argillaceous limestone, mudstone, marine siltstone shale, fluvial siltstone, coastal siltstone, and sandstone.

[0077] A bar chart showing the number of samples for each label (e.g.) Figure 8 The sample set is obtained by filtering labels with a small number of samples and supplementing the samples of the label with the selected machine learning method.

[0078] Step S54: Divide the sample set into a training set and a test set.

[0079] Step S55: Build a neural network model and train the neural network model using a periodically decaying learning rate on the training set.

[0080] Example 2 was implemented using Python 3.6. The neural network structure was built using the PyTorch framework, and the operating system used for training and testing was Windows 10. The machine configuration was as follows: CPU: Intel(R) Xeon(R) Gold 6234 CPU @ 3.30GHz; Memory: 256GB; GPU: Nvidia Quadro RTX 5000.

[0081] The trained neural network model is used to predict test set data, and the training rounds, neural network depth and width are fine-tuned by combining the convergence curve of the loss function with the changes in rounds.

[0082] The loss functions for the training and validation sets change with each training epoch, as follows: Figure 9 The more training epochs, the smaller the loss function; however, too many training epochs can lead to overfitting, which in turn reduces the accuracy of model predictions. Therefore, it is possible to... Figure 9 The optimal training epoch is the one in which the loss function of the validation set just reaches its minimum.

[0083] Figure 9 It also verifies that the periodically decaying learning rate improves training efficiency and enables convergence quickly. Compared with neural network training with constant or decaying learning rates, the training time of Example 2 is reduced by 60% while achieving the same 80% prediction accuracy.

[0084] Figure 10 This study compares the actual lithological results with predictions obtained through various machine learning methods and deep learning. Support vector machine (SVM) and deep learning methods showed higher prediction accuracy than other methods. Sandstone and marine siltstone showed relatively high prediction accuracy, while barrier limestone and dolomite showed slightly lower accuracy, which is related to the insufficient number of original labeled samples for these types of rocks.

[0085] Example 3

[0086] Embodiment 3 of the present invention provides an application example of a method for establishing a sedimentation type identification model.

[0087] Using only one well as a known well, to simulate a situation with insufficient labeled samples, five curves were used as input, with sample labels including sandstone and shale. After training the model on this well, classification prediction was performed on another well. The experimental results show (e.g.) Figure 11 Although the prediction effect of semi-supervised learning, which uses the results of machine learning classification as input, is somewhat inferior to that of supervised learning, represented by deep learning, it greatly alleviates the problem of insufficient labeled samples and verifies that the present method has the ability to solve the small sample problem.

[0088] Based on the inventive concept of this invention, embodiments of this invention also provide a deposition type identification method, including:

[0089] Input the selected logging curves of the target well into the sedimentation type identification model established by the above method, and predict the sedimentation type data of the target well based on the output results of the model.

[0090] Based on the inventive concept of this invention, embodiments of this invention also provide a deposition type identification model establishment device, the structure of which is as follows: Figure 12 As shown, it includes:

[0091] The sample set establishment module 121 is used to obtain a sample set based on multiple selected logging curves and sedimentation type data of at least one well. The samples in the sample set include curve segments and sedimentation types of each selected logging curve within a certain depth range of the corresponding well.

[0092] The sedimentation type identification model establishment module 122 is used to set the network structure and network parameters of the deep neural network according to the set rules, and to train the set deep neural network with a periodically decaying learning rate through the sample set to obtain a sedimentation type identification model for predicting sedimentation type data of the target well.

[0093] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0094] Based on the inventive concept of the present invention, embodiments of the present invention also provide a computer program product, including a computer program / instruction, wherein the computer program / instruction, when executed by a processor, implements the above-mentioned deposition type identification model establishment method, or implements the above-mentioned deposition type identification method.

[0095] Unless otherwise specifically stated, terms such as processing, calculation, operation, determination, display, etc., may refer to the actions and / or processes of one or more processing or computing systems or similar devices that represent the manipulation and conversion of data representing physical (e.g., electronic) quantities within the registers or memory of the processing system into other data similarly representing physical quantities within the memory, registers, or other such information storage, transmission, or display devices of the processing system. Information and signals can be represented using any of a variety of different techniques and methods. For example, data, instructions, commands, information, signals, bits, symbols, and chips mentioned throughout the above description can be represented by voltage, current, electromagnetic waves, magnetic fields or particles, light fields or particles, or any combination thereof.

[0096] It should be understood that the specific order or hierarchy of steps in the disclosed process is an example of an exemplary method. Based on design preferences, it should be understood that the specific order or hierarchy of steps in the process may be rearranged without departing from the scope of this disclosure. The appended method claims provide elements of various steps in an exemplary order and are not intended to limit the scope to the specific order or hierarchy described.

[0097] In the detailed description above, various features are combined together in a single embodiment to simplify this disclosure. This approach to disclosure should not be construed as reflecting an intention that embodiments of the claimed subject matter require more features than are explicitly stated in each claim. Rather, as reflected in the appended claims, the invention is presented with fewer features than all of the features in a single disclosed embodiment. Therefore, the appended claims are hereby explicitly incorporated into the detailed description, with each claim representing a separate preferred embodiment of the invention.

[0098] Those skilled in the art will also understand that the various illustrative logic blocks, modules, circuits, and algorithm steps described in conjunction with the embodiments herein can be implemented as electronic hardware, computer software, or a combination thereof. To clearly illustrate the interchangeability between hardware and software, the various illustrative components, blocks, modules, circuits, and steps described above are generally described in terms of their functionality. Whether such functionality is implemented as hardware or software depends on the specific application and the design constraints imposed on the overall system. Those skilled in the art can implement the described functionality in alternative ways for each specific application; however, such implementation decisions should not be construed as departing from the scope of this disclosure.

[0099] The steps of the methods or algorithms described in conjunction with the embodiments herein can be directly embodied in hardware, software modules executed by a processor, or a combination thereof. The software modules can reside in RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, hard disks, removable disks, CD-ROMs, or any other form of storage medium well known in the art. An exemplary storage medium is connected to the processor, enabling the processor to read information from and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and storage medium can reside in an ASIC. The ASIC can reside in a user terminal. Alternatively, the processor and storage medium can exist as discrete components in the user terminal.

[0100] For software implementation, the techniques described in this application can be implemented using modules (e.g., procedures, functions, etc.) that perform the functions described in this application. This software code can be stored in memory units and executed by a processor. The memory units can be implemented within the processor or outside the processor; in the latter case, they are communicatively coupled to the processor via various means, as is well known in the art.

[0101] The foregoing description includes examples of one or more embodiments. It is certainly impossible to describe all possible combinations of components or methods in order to describe the above embodiments, but those skilled in the art will recognize that further combinations and arrangements of the various embodiments are possible. Therefore, the embodiments described herein are intended to cover all such changes, modifications, and variations that fall within the scope of the appended claims. Furthermore, the term "comprising" as used in the specification or claims is interpreted in a manner similar to the term "including," as interpreted when used as a conjunction in the claims. Additionally, the use of any term "or" in the specification of the claims is intended to mean "non-exclusive or."

Claims

1. A method for establishing a sedimentation type identification model, characterized in that, include: A sample set is obtained based on multiple selected logging curves and sediment type data from at least one well. The samples in the sample set include curve segments and sediment types of each selected logging curve within a certain depth range of the corresponding well. The deep neural network structure and parameters are set according to the set rules, and the set deep neural network is trained with a periodically decaying learning rate using the sample set to obtain a sedimentation type identification model for predicting sedimentation type data of the target well. The process of setting the network structure of the deep neural network according to the set rules specifically includes: setting the number of input neurons of the deep neural network to be the same as the number of types of the selected logging curves; and setting the number of output neurons of the deep neural network to be the same as the number of types of sedimentation types included in the sample set. Configure the network parameters of the deep neural network according to the set rules, specifically including: Set the activation function of the deep neural network to the softmax function: ； Where, x i This represents the probability that the output belongs to the i-th class. Let n be the activation function value corresponding to the probability of outputting the i-th class, where n is the total number of classes; Set the loss function of the deep neural network to the cross-entropy loss function: ； in, The value of the cross-entropy loss function. The label is for the i-th class.

2. The method as described in claim 1, characterized in that, The step of training the configured deep neural network on the sample set using a periodically decaying learning rate to obtain a sedimentation type identification model for predicting sedimentation type data of the target well specifically includes: The sample set is divided into a training set and a test set; Using a periodically decaying learning rate on the training set, the deep neural network is iteratively trained according to the set initial training rounds, and the optimal training rounds are determined based on the output training logs. According to the preferred training rounds, the set deep neural network is iteratively trained using the periodically decaying learning rate on the training set to obtain a sedimentation type identification model for predicting sedimentation type data of the target well.

3. The method as described in claim 1, characterized in that, After obtaining the sample set based on multiple selected logging curves and sedimentary type data from at least one well, the process also includes: Determine whether the sample set meets the training quantity requirement; If not, supplement the sample set by selecting a machine learning method.

4. The method as described in claim 1, characterized in that, Before obtaining the sample set based on multiple selected logging curves and sedimentary type data from at least one well, the process also includes: Based on multiple logging curves and sediment type data from at least one well, the selected logging curve type is screened out through correlation analysis between logging curve type and sediment type.

5. The method according to any one of claims 1 to 4, characterized in that, The selected logging curve is the normalized logging curve.

6. The method as described in claim 5, characterized in that, The selected logging curve is the logging curve after the mean has been zeroed.

7. The method according to any one of claims 1 to 4, characterized in that, The sedimentary type is either a sedimentary facies type or a lithological type.

8. A method for identifying sediment types, characterized in that, include: The selected logging curves of the target well are input into the sedimentation type identification model established by any one of claims 1 to 7, and the sedimentation type data of the target well are predicted based on the output results of the model.

9. A device for establishing a sedimentation type identification model, characterized in that, include: The sample set establishment module is used to obtain a sample set based on multiple selected logging curves and sedimentation type data from at least one well. The samples in the sample set include curve segments and sedimentation types of each selected logging curve within a certain depth range of the corresponding well. The sedimentation type identification model building module is used to set the network structure and network parameters of the deep neural network according to the set rules, and to train the set deep neural network with a periodically decaying learning rate through the sample set to obtain a sedimentation type identification model for predicting the sedimentation type data of the target well. The method of setting the network structure of the deep neural network according to the set rules is specifically used to: set the number of input neurons of the deep neural network to be the same as the number of types of the selected logging curves; and set the number of output neurons of the deep neural network to be the same as the number of types of sedimentation types included in the sample set. Set the network parameters of the deep neural network according to the set rules, specifically for: Set the activation function of the deep neural network to the softmax function: ； Where, x i This represents the probability that the output belongs to the i-th class. Let n be the activation function value corresponding to the probability of outputting the i-th class, where n is the total number of classes; Set the loss function of the deep neural network to the cross-entropy loss function: ； in, The value of the cross-entropy loss function. The label is for the i-th class.

10. A computer program product, comprising a computer program / instructions, characterized in that, When the computer program / instruction is executed by the processor, it implements the deposition type identification model establishment method according to any one of claims 1 to 7, or the deposition type identification method according to claim 8.

Images