Method and system for predicting physical well data using deep learning-based iterative estimation

Abstract

The present invention provides a method for predicting physical logging data, comprising receiving physical logging data in which a predetermined number of missing data has occurred, estimating a first missing data corresponding to the physical logging data using a pre-trained conditional variation model, predicting rock formation sequence information corresponding to at least one of the physical logging data and the first missing data using a pre-trained long-short-term memory model, and, conditionally according to the rock formation sequence information, estimating a second missing data corresponding to the physical logging data using the pre-trained conditional variation model.

Description

Technology Field

[0001] The present invention relates to a method and system for predicting physical logging data that performs deep learning-based iterative estimation. Background Technology

[0002] Geophysical logging data may include data collected through various sensors to identify the physical characteristics of the subsurface. Such data is primarily utilized in the exploration and development of oil, gas, and mineral resources, and is also used to model subsurface structures and estimate resource locations. However, cases have been observed where some values ​​are missing due to sensor errors or changes in the external environment during the real-time collection of geophysical logging data.

[0003] Accordingly, simple imputation techniques (e.g., mean imputation or adjacent value imputation) or regression-based techniques are applied to process or restore missing data in physical logging data.

[0004] Meanwhile, deep learning is a field of artificial intelligence (AI) implemented to learn and predict patterns and characteristics of complex data through neural network models with a multi-layered structure. Deep learning demonstrates excellent performance, particularly in the process of solving large-scale data and complex problems, and is being used in various fields such as image recognition, speech recognition, natural language processing, and time series analysis. The problem to be solved

[0005] The present invention relates to a method and system for predicting physical logging data that performs deep learning-based iterative estimation, which can improve the restoration performance of physical logging data.

[0006] In addition, the present invention relates to a method and system for predicting physical logging data that performs deep learning-based iterative estimation, which can progressively improve the accuracy of two models through an iterative prediction process and perform more accurate restoration of physical logging data. means of solving the problem

[0007] To solve the problem described above, the method for predicting physical logging data according to the present invention may include: receiving physical logging data in which a predetermined number of missing data has occurred; estimating a first missing data corresponding to the physical logging data using a pre-learned conditional variation model; predicting rock formation sequence information corresponding to at least one of the physical logging data and the first missing data using a pre-learned long-term short-term memory model; and, conditionally according to the rock formation sequence information, estimating a second missing data corresponding to the physical logging data using the pre-learned conditional variation model.

[0008] In addition, the physical logging data prediction system according to the present invention includes: a storage unit in which physical logging data in which a predetermined number of missing data has occurred is stored; and a control unit that estimates a first missing data corresponding to the physical logging data using a pre-learned conditional variation model, wherein the control unit predicts rock formation sequence information corresponding to at least one of the physical logging data and the first missing data using a pre-learned long-term memory model, and, as a condition according to the rock formation sequence information, can estimate a second missing data corresponding to the physical logging data using the pre-learned conditional variation model.

[0009] Additionally, a program stored on a computer-readable recording medium according to the present invention is executed by one or more processes in an electronic device and is a program stored on a computer-readable recording medium, wherein the program may include instructions for performing the steps of: receiving physical logging data in which a predetermined missing data has occurred; estimating a first missing data corresponding to the physical logging data using a pre-learned conditional variation model; predicting a lithographic sequence information corresponding to at least one of the physical logging data and the first missing data using a pre-learned long-term short-term memory model; and, conditionally according to the lithographic sequence information, estimating a second missing data corresponding to the physical logging data using the pre-learned conditional variation model. Effects of the invention

[0010] According to various embodiments of the present invention, a method and system for predicting physical logging data that performs deep learning-based iterative estimation can improve the restoration performance of physical logging data by predicting lithological sequence information using physical logging data in which missing data has occurred, and estimating missing data for physical logging data conditionally according to the predicted lithological sequence information.

[0011] In addition, according to various embodiments of the present invention, a method and system for predicting physical logging data that performs deep learning-based iterative estimation can iteratively predict lithological sequence information and missing data through mutual feedback by a conditional variation model and a long short-term memory model, thereby gradually improving the accuracy of the two models through an iterative prediction process and performing more accurate restoration of physical logging data. Brief explanation of the drawing

[0012] FIG. 1 illustrates an embodiment of predicting missing data according to the present invention. FIG. 2 illustrates a physical logging data prediction system according to the present invention. FIG. 3 is a flowchart illustrating a method for predicting physical logging data according to the present invention. FIG. 4 illustrates an example of estimating first missing data. FIG. 5 illustrates an example of predicting cancer sequence information. FIG. 6 illustrates an example of repeatedly generating missing data and dark sequence information. Specific details for implementing the invention

[0013] Hereinafter, embodiments disclosed in this specification will be described in detail with reference to the attached drawings. Identical or similar components are assigned the same reference number regardless of the drawing symbols, and redundant descriptions thereof will be omitted. The suffixes "module" and "part" used for components in the following description are assigned or used interchangeably solely for the ease of drafting the specification and do not have distinct meanings or roles in themselves. Furthermore, in describing the embodiments disclosed in this specification, if it is determined that a detailed description of related prior art could obscure the essence of the embodiments disclosed in this specification, such detailed description will be omitted. Additionally, the attached drawings are intended only to facilitate understanding of the embodiments disclosed in this specification; the technical concept disclosed in this specification is not limited by the attached drawings, and it should be understood that they include all modifications, equivalents, and substitutions that fall within the spirit and technical scope of the present invention.

[0014] Terms including ordinal numbers, such as first, second, etc., may be used to describe various components, but said components are not limited by said terms. These terms are used solely for the purpose of distinguishing one component from another.

[0015] When it is stated that one component is "connected" or "connected" to another component, it should be understood that while it may be directly connected or connected to that other component, there may also be other components in between. On the other hand, when it is stated that one component is "directly connected" or "directly connected" to another component, it should be understood that there are no other components in between.

[0016] Singular expressions include plural expressions unless the context clearly indicates otherwise.

[0017] In this application, terms such as “comprising” or “having” are intended to specify the existence of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0018] FIG. 1 illustrates an embodiment for predicting missing data according to the present invention. FIG. 2 illustrates a physical logging data prediction system according to the present invention.

[0019] Referring to FIG. 1, the physical logging data prediction system (100) according to the present invention can estimate a first missing data corresponding to physical logging data in which a predetermined missing data has occurred using a pre-learned conditional variation model, predict lithological sequence information corresponding to at least one of the physical logging data and the first missing data using a pre-learned long-term memory model, and estimate a second missing data corresponding to the physical logging data again using a pre-learned conditional variation model.

[0020] Accordingly, the physical logging data prediction system (100) can predict rock formation sequence information from the physical logging data and repeat the process of estimating missing data for the physical logging data conditionally based on the predicted rock formation sequence information, thereby restoring missing data for the physical logging data.

[0021] Here, geophysical logging data may include various physical data collected through underground boreholes, etc., to evaluate underground rock and fluid characteristics. For example, geophysical logging data may include data corresponding to multiple geophysical logging items such as gamma rays, porosity, density, compressional wave velocity, shear wave velocity, electrical conductivity, thermal conductivity, and radioactivity.

[0022] In this regard, missing data may include data corresponding to at least one physical logging item in which a missing value occurred in the physical logging data. Such missing data may have occurred during the measurement process of the physical logging data or during the data processing and communication process.

[0023] Lithographic sequence information may include data representing a structure in which the physical and chemical properties of rocks are arranged in layers and continuously. For example, lithographic sequence information may include information on the types of rocks formed in a specific area and how the layered structures of each rock are distributed. Accordingly, lithographic sequence information may be arranged so that information about different rocks (e.g., types of rocks and thickness, etc.) is listed sequentially.

[0024] Meanwhile, the conditional variation model may be trained to estimate the physical logging data or missing data in the physical logging data based on conditions according to the lithological sequence information when physical logging data and lithological sequence information are input.

[0025] To this end, the conditional variation model may be a Conditional Variational Autoencoder (CVAE) implemented with an encoder and a decoder; thus, when physical logging data without missing values ​​and rock formation sequence information corresponding to said physical logging data are input, the conditional variation model can be trained by combining the physical logging data and the rock formation sequence information, extracting feature vectors in the latent space from the combined data, and generating a probability distribution based on the extracted feature vectors and the rock formation sequence information.

[0026] Through this, when physical logging data and rock formation sequence information containing missing data are input, the conditional variation model combines the physical logging data and rock formation sequence information, extracts feature vectors in the latent space from the combined data, and can generate missing data based on the probability distribution of the extracted feature vectors and rock formation sequence information.

[0027] The Long Short-Term Memory model may be trained to predict lithological sequence information corresponding to physical logging data when such data is input. In this case, the physical logging data input to the Long Short-Term Memory model may include missing data estimated through a conditional variation model.

[0028] To this end, the long-short-term memory model may be a Long-Short Term Memory (LSTM), and thus, when training physical logging data containing missing data is input, the long-short-term memory model sequentially stores sequences of multiple samples included in the training physical logging data in a pre-implemented memory, removes or preserves at least some data from the memory according to a predetermined rule, predicts training image sequence information based on the data in the finally updated memory, and can be trained based on a loss calculated by comparing the predicted training image sequence information with the ground truth image sequence information.

[0029] Accordingly, when physical logging data is input, the long-term and short-term memory model sequentially updates the memory according to the sequence of multiple samples included in the physical logging data, removes or preserves at least some data from the previously updated memory according to a predetermined rule, and can predict dark sequence information based on the data of the memory finally updated for the physical logging data.

[0030] Meanwhile, referring to FIG. 2, the physical logging data prediction system (100) according to the present invention may include an input unit (110), a storage unit (120), a control unit (130), and an output unit (140). These input unit (110), storage unit (120), control unit (130), and output unit (140) may be connected to each other through a system bus, a network, or other connection mechanism.

[0031] The input unit (110) may receive information necessary for the operation of the physical logging data prediction system (100) according to the present invention. To this end, the input unit (110) may be connected to a separate input device, server, or external storage device via a wireless or wired network. For example, the input unit (110) may receive specific information based on a touch screen, computer mouse, keyboard, keypad, touchpad, trackball, joystick, voice recognition module, or other similar device.

[0032] Accordingly, the input unit (110) can receive physical inspection data (10) from a separate input device, server, or external storage device, etc.

[0033] Additionally, the storage unit (120) may store commands and information necessary for the operation of the physical logging data prediction system (100). As a data storage unit (120), this storage unit (120) may include one or more non-transient computer-readable storage media that can be read and / or accessed by a control unit (130) (e.g., at least one of one or more processors).

[0034] One or more non-transient computer-readable storage media may include volatile and / or non-volatile storage components, such as optical, magnetic, organic, or other memory or disk storage devices. In some examples, the storage unit (120) may be implemented using a single physical device (e.g., one optical, magnetic, organic, or other memory or disk storage device), whereas in other examples, the storage unit (120) may be implemented using two or more physical devices.

[0035] Additionally, the storage unit (120) may include computer-readable instructions and additional data. The storage unit (120) may include a storage necessary to perform at least some of the methods, scenarios, and techniques described herein and / or at least some of the functions of the device and network.

[0036] Accordingly, the storage unit (120) can store physical logging data (10). Additionally, the storage unit (120) can store a conditional variation model (20) and a long-term short-term memory model (30), and can store missing data (40) estimated from the conditional variation model (20) and dark sequence information predicted from the long-term short-term memory model (30).

[0037] A control unit (130) can control the overall operation of the physical logging data prediction system (100). This control unit (130) may be one or more processors and may include one or more general-purpose processors and / or one or more special-purpose processors (e.g., digital signal processor, tensor processing unit (TPU), graphics processing unit (GPU), neural network processing unit (NPU), application integrated circuit, application semiconductor (ASIC), etc.). The control unit (130) may be configured to execute computer-readable instructions contained in the storage unit (120) and / or other instructions described herein.

[0038] Accordingly, the control unit (130) can estimate a first missing data corresponding to physical logging data (10) in which a predetermined number of missing data has occurred using a pre-learned conditional variation model (20), predict a rock pattern sequence information corresponding to at least one of the physical logging data (10) and the first missing data using a pre-learned long-term short-term memory model (30), and estimate a second missing data corresponding to the physical logging data again using a pre-learned conditional variation model (20).

[0039] Specifically, the control unit (130) can receive physical logging data (10) in which a predetermined number of missing values ​​have occurred. That is, the control unit (130) can receive physical logging data (10) in which data corresponding to at least one physical logging item is missing from the measurement results for a plurality of physical logging items.

[0040] Accordingly, the control unit (130) scales the physical logging data (10) to normalize it to a predetermined value range, checks the number of samples for each of the multiple physical logging items belonging to the physical logging data (10), compares the confirmed number of samples with a predetermined reference value, and can augment the physical logging data (10) of physical logging items with a number of samples less than the reference value based on the comparison result.

[0041] Accordingly, the control unit (130) can estimate first missing data corresponding to physical logging data (10) using a pre-learned conditional variation model (20). To this end, the control unit (130) can generate first missing data corresponding to physical logging data (10) by inputting the physical logging data (10) and pre-prepared initial rock phase sequence information into the pre-learned conditional variation model (20).

[0042] Additionally, the control unit (130) can predict rock formation sequence information corresponding to at least one of the physical logging data (10) and the first missing data by using a pre-learned long-term memory model (30). To this end, the control unit (130) inputs the physical logging data (10) and the first missing data into the pre-learned long-term memory model to predict rock formation sequence information corresponding to the physical logging data (10) and the first missing data.

[0043] Furthermore, the control unit (130) can estimate second missing data corresponding to physical logging data (10) using a pre-learned conditional variation model (20) as a conditional part according to the rock phase sequence information. That is, the control unit (130) can generate second missing data corresponding to rock phase sequence information and physical logging data (10) by inputting physical logging data (10) including first missing data estimated prior to the pre-learned conditional variation model (20) and rock phase sequence information predicted based on the first missing data and physical logging data (10).

[0044] At this time, the control unit (130) can repeatedly perform the process of predicting the rock formation sequence information corresponding to the physical logging data (10) and the second missing data using a pre-learned short-term memory model (30), and then estimating the missing data (40) corresponding to the physical logging data (10) using a conditional variation model (20) based on the predicted rock formation sequence information.

[0045] The output unit (140) may output information generated by the operation of the physical logging data prediction system (100) according to the present invention. To this end, the output unit (140) may be connected to a separate visual output device, server, or external storage device via a wireless or wired network.

[0046] For example, the output unit (140) may be configured to provide output to one or more user display devices, such as cathode ray tubes (CRTs), liquid crystal displays, light-emitting diodes (LEDs), displays using digital light processing (DLP) technology, and printers. Additionally, the output unit (140) may be configured to produce audible output using devices such as speakers, speaker jacks, audio output ports, audio output devices, earphones and / or other similar devices, and one or more tactile devices may be additionally configured to produce tactile outputs, such as vibrations and / or other outputs detectable by touch and / or physical contact.

[0047] Accordingly, the output unit (140) can output physical logging data (10), missing data (40), and rock sequence information, etc., so that a user can visually check them through a separate output device, server, or external storage device, and according to an embodiment, the physical logging data (10), missing data (40), and rock sequence information, etc., may also be transmitted to another device.

[0048] Meanwhile, the physical logging data prediction system (100) may be a computing device comprising one or more sensors and / or power systems. Here, one or more sensors may be configured to measure conditions within the computing device and / or conditions of the computing device environment and to provide data regarding these conditions. The power system may include one or more batteries and / or one or more external power interfaces for providing power to the computing device.

[0049] Based on the configuration of the physical logging data prediction system (100) examined above, the method for predicting physical logging data will be explained in more detail below.

[0050] FIG. 3 is a flowchart illustrating a method for predicting physical logging data according to the present invention. FIG. 4 illustrates an embodiment for estimating first missing data. FIG. 5 illustrates an embodiment for predicting lithographic sequence information. FIG. 6 illustrates an embodiment for repeatedly generating missing data and lithographic sequence information.

[0051] Referring to FIG. 3, the physical logging data prediction system (100) according to the present invention can receive physical logging data in which a predetermined number of missing data has occurred (S100).

[0052] Specifically, the physical logging data prediction system (100) can receive physical logging data in which data corresponding to at least one physical logging item is missing from the measurement results for a plurality of physical logging items.

[0053] For example, the physical logging data prediction system (100) can obtain physical logging data for multiple physical logging items excluding the missing physical logging item, such as gamma rays, porosity, density, compressional wave velocity, shear wave velocity, electrical conductivity, thermal conductivity and radioactivity, where at least one physical logging item is missing.

[0054] That is, the physical logging data prediction system (100) can obtain physical logging data including gamma rays, porosity, density, compressional wave velocity, shear wave velocity, electrical conductivity, thermal conductivity, and radioactivity when physical logging data for compressional wave velocity is missing among a plurality of physical logging items including gamma rays, porosity, density, compressional wave velocity, shear wave velocity, electrical conductivity, thermal conductivity, and radioactivity.

[0055] Alternatively, the physical logging data prediction system (100) can obtain physical logging data including gamma rays, porosity, density, compressive wave velocity, shear wave velocity, and radioactivity when physical logging data for electrical conductivity and thermal conductivity is missing among a plurality of physical logging items including gamma rays, porosity, density, compressive wave velocity, shear wave velocity, electrical conductivity, thermal conductivity, and radioactivity.

[0056] Alternatively, the physical logging data prediction system (100) can obtain physical logging data including gamma rays, density, compressional wave velocity, shear wave velocity, electrical conductivity, thermal conductivity, and radioactivity when physical logging data for porosity is missing among a plurality of physical logging items including gamma rays, porosity, density, compressional wave velocity, shear wave velocity, electrical conductivity, thermal conductivity, and radioactivity.

[0057] In this regard, the physical logging data prediction system (100) may, according to an embodiment, generate data corresponding to missing physical logging items in the physical logging data as a predetermined initial value (e.g., 0, 1, etc.).

[0058] Furthermore, the physical logging data prediction system (100) can scale the physical logging data to normalize it into a predetermined range of values.

[0059] For example, the physical logging data prediction system (100) can normalize the range of values ​​for each of the multiple physical logging items belonging to the physical logging data to a predetermined range of values ​​through a Min-Max scaling technique. In one embodiment, the physical logging data prediction system (100) can normalize the physical logging data so that it appears as a value between 0 and 1.

[0060] Furthermore, the physical logging data prediction system (100) can verify the number of samples for each of the multiple physical logging items belonging to the physical logging data, compare the number of samples between the physical logging items based on the verification result, and, based on the comparison result, if the difference in the number of samples between the physical logging items is greater than a predetermined threshold, the physical logging data of the physical logging item can be augmented.

[0061] For example, a physical logging data prediction system (100) can compare the number of samples for each of a plurality of physical logging items belonging to physical logging data, and if the difference in the number of samples according to the comparison result is greater than a predetermined threshold, the physical logging item with the smaller number of samples among the compared physical logging items can be specified as the target item.

[0062] At this time, the physical logging data prediction system (100) may specify one or more target items by performing a pairwise comparison for each of the multiple physical logging items belonging to the physical logging data, or may calculate the average value of the number of samples for each of the multiple physical logging items belonging to the physical logging data and specify the target item based on a comparison between the calculated average value and each of the multiple physical logging items. Additionally, the physical logging data prediction system (100) may specify the target item based on a comparison between one of the multiple physical logging items with the largest number of samples and other multiple physical logging items.

[0063] Accordingly, the physical logging data prediction system (100) can increase the number of samples for a target item based on SMOTE (Synthetic Minority Over-sampling Technique). To this end, the physical logging data prediction system (100) can specify each of the multiple samples belonging to the target item as a target sample, specify one or more neighbor samples adjacent to each target sample, and generate an augmentation sample between the specified one or more neighbor samples and each target sample.

[0064] At this time, the physical logging data prediction system (100) can generate augmented samples corresponding to the average number of samples for each of the multiple physical logging items, or generate augmented samples corresponding to the number of samples of any one physical logging item that has the largest number of samples.

[0065] Referring to FIG. 3, the physical logging data prediction system (100) according to the present invention can estimate first missing data corresponding to physical logging data using a pre-learned conditional variation model (S200).

[0066] Specifically, the physical logging data prediction system (100) can generate first missing data corresponding to the physical logging data by inputting physical logging data and pre-prepared initial rock phase sequence information into a pre-trained conditional variation model.

[0067] For example, with reference to FIG. 4, the physical logging data prediction system (100) can generate initial rock formation sequence information (51) with a predetermined initial value. According to an embodiment, the physical logging data prediction system (100) may generate initial rock formation sequence information (51) with different initial values ​​depending on the physical logging items missing in the physical logging data (10), and to this end, the physical logging data prediction system (100) can select the physical logging items missing in the physical logging data (10) based on user input.

[0068] Accordingly, the physical logging data prediction system (100) can generate the first missing data (41) by inputting the physical logging data (10) and the initial rock phase sequence information (51) into a pre-trained conditional variation model (20).

[0069] Here, the conditional variation model (20) may be learned using physical logging data for multiple physical logging items such as gamma rays, porosity, density, compressional wave velocity, shear wave velocity, electrical conductivity, thermal conductivity, and radioactivity, along with rock phase sequence information in which the physical logging data was measured. At this time, the conditional variation model (20) may be learned through the physical logging data prediction system (100) according to the present invention or through a separate learning system.

[0070] To this end, in one embodiment, the conditional variation model (20) may be a CVAE implemented with an encoder and a decoder. Accordingly, the conditional variation model (20) can be learned by combining physical logging data and rock formation sequence information corresponding to the physical logging data when physical logging data without missing data and rock formation sequence information corresponding to the physical logging data are input, extracting feature vectors in a latent space from the combined data, and generating a probability distribution according to the extracted feature vectors and rock formation sequence information.

[0071] Accordingly, the physical logging data prediction system (100) can combine physical logging data (10) containing missing data and initial rock formation sequence information (51) through a conditional variation model (20), extract a feature vector in a latent space from the combined data, and generate a first missing data (41) based on the probability distribution of the extracted feature vector and the initial rock formation sequence information (51).

[0072] Referring again to FIG. 3, the physical logging data prediction system (100) according to the present invention can predict rock pattern sequence information corresponding to at least one of physical logging data and first missing data using a pre-learned long-term and short-term memory model (S300).

[0073] Specifically, the physical logging data prediction system (100) can predict rock formation sequence information corresponding to the physical logging data and the first missing data by inputting the physical logging data and the first missing data into a pre-trained short-term memory model.

[0074] For example, with reference to FIG. 5, the long-term short-term memory model (30) may be trained to analyze the sequence of multiple samples appearing from the physical logging data (11) using the training physical logging data and ground truth sequence information, and to predict the rock sequence information (53) corresponding to the physical logging data (11) based on the analysis results. At this time, the long-term short-term memory model (30) may be trained through the physical logging data prediction system (100) according to the present invention or through a separate learning system.

[0075] To this end, in one embodiment, the long short-term memory model (30) may be an LSTM. Accordingly, when training physical logging data containing missing data is input, the long short-term memory model (30) sequentially stores sequences of multiple samples included in the training physical logging data in a pre-implemented memory, removes at least some data from the memory according to a predetermined rule, predicts training image sequence information based on the data in the finally updated memory, and can be trained based on a loss calculated by comparing the predicted training image sequence information with the correct image sequence information.

[0076] Accordingly, the physical logging data prediction system (100) sequentially updates the memory according to the sequence of multiple samples included in the physical logging data (11) through the long-term and short-term memory model (30), removes at least some of the data from the previously updated memory according to a predetermined rule, and can predict the dark sequence information (53) based on the data of the memory that was finally updated for the physical logging data (11).

[0077] At this time, the physical logging data (11) input to the short-term and long-term memory model (30) may include (or be combined with) the missing data (e.g., first missing data (41)) generated from the conditional variation model in addition to the previously received physical logging data (11).

[0078] Referring again to FIG. 3, the physical logging data prediction system (100) according to the present invention can estimate second missing data corresponding to physical logging data using a pre-learned conditional variation model based on the rock phase sequence information (S400).

[0079] Specifically, the physical logging data prediction system (100) can input physical logging data including first missing data estimated prior to a pre-learned conditional variation model, and rock formation sequence information predicted based on the first missing data and physical logging data, to generate second missing data corresponding to rock formation sequence information and physical logging data.

[0080] For example, the physical logging data prediction system (100) can input physical logging data and initial rock phase sequence information into a conditional variation model to generate first missing data and combine the physical logging data, and input the physical logging data and the first missing data into a long-term short-term memory model to input the predicted rock phase sequence information and the previously combined physical logging data into the conditional variation model. Through this, the physical logging data prediction system (100) can estimate second missing data corresponding to the previously combined physical logging data and the previously predicted rock phase sequence information from the conditional variation model.

[0081] Furthermore, the physical logging data prediction system (100) can repeatedly perform the process of predicting the rock formation sequence information corresponding to the physical logging data and the second missing data using a pre-learned short-term memory model, and then estimating the missing data corresponding to the physical logging data using a conditional variation model based on the predicted rock formation sequence information.

[0082] For example, with reference to FIG. 6, a physical logging data prediction system (100) can generate missing data (43) corresponding to physical logging data (13) and rock formation sequence information (55) using a conditional variation model (20), update physical logging data (13) by combining the previously generated missing data (43) with the physical logging data, generate rock formation sequence information (55) corresponding to the previously updated physical logging data (13) (or missing data (43)) using a long-term short-term memory model (30), and repeat the process of generating missing data (43) corresponding to the previously updated physical logging data (13) and rock formation sequence information (55) again using a conditional variation model (20).

[0083] At this time, the physical logging data prediction system (100) can compare the difference between previously predicted missing data and currently predicted missing data through a conditional variation model (20), and repeat the process of predicting the rock sequence information (55) and missing data (43) until the difference resulting from the comparison becomes smaller than a predetermined threshold.

[0084] Alternatively, the physical logging data prediction system (100) may repeat the process of predicting rock sequence information (55) and missing data (43) a predetermined number of times.

[0085] Through the above configurations, the physical logging data prediction system (100) according to the present invention can improve the restoration performance of physical logging data by predicting rock formation sequence information using physical logging data in which missing values ​​have occurred, and estimating missing data for physical logging data conditionally according to the predicted rock formation sequence information.

[0086] In addition, the physical logging data prediction system (100) according to the present invention can perform more accurate physical logging data restoration by repeatedly predicting lithological sequence information and missing data through mutual feedback by a conditional variation model and a long-term short-term memory model, thereby gradually improving the accuracy of the two models through a repeated prediction process.

[0087] Furthermore, the present invention described above can be implemented as a program that is executed by one or more processes in an electronic device and stored in a computer-readable recording medium.

[0088] Accordingly, the present invention can be implemented as computer-readable code or instructions on a medium on which a program is recorded. That is, various control methods according to the present invention can be provided in the form of integrated or individual programs.

[0089] Meanwhile, computer-readable media include all types of recording devices in which data that can be read by a computer system is stored. Examples of computer-readable media include HDD (Hard Disk Drive), SSD (Solid State Disk), SSD (Silicon Disk Drive), ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage device, etc.

[0090] Furthermore, the computer-readable medium may be a server or cloud storage that includes a storage and is accessible to an electronic device via communication. In this case, the computer may download the program according to the present invention from the server or cloud storage via wired or wireless communication.

[0091] Furthermore, in the present invention, the computer described above is an electronic device equipped with a processor, namely a CPU (Central Processing Unit), and no special limitations are placed on its type.

[0092] Meanwhile, the above detailed description should not be interpreted restrictively in all respects but should be considered exemplary. The scope of the invention shall be determined by a reasonable interpretation of the appended claims, and all modifications within the equivalent scope of the invention are included within the scope of the invention.

Claims

Claim 1 A method for predicting physical logging data, comprising: receiving physical logging data in which a predetermined number of missing data has occurred; estimating first missing data corresponding to said physical logging data using a pre-learned conditional variation model; predicting rock formation sequence information corresponding to at least one of said physical logging data and said first missing data using a pre-learned long-term short-term memory model; and, conditionally according to said rock formation sequence information, estimating second missing data corresponding to said physical logging data using the pre-learned conditional variation model. Claim 2 A method for predicting physical logging data according to claim 1, further comprising the step of re-predicting lithographic sequence information corresponding to the physical logging data and the second missing data using the long-term / short-term memory model, and re-estimating the missing data corresponding to the physical logging data using the conditional variation model conditionally according to the re-predicted lithographic sequence information. Claim 3 A method for predicting physical logging data according to claim 2, wherein the step of re-estimating the missing data comprises repeating the process of generating the missing data using the conditional variation model, updating the physical logging data by combining the generated missing data with the physical logging data, generating the rock formation sequence information corresponding to the updated physical logging data using the long-term short-term memory model, and re-generating the missing data corresponding to the updated physical logging data and the rock formation sequence information using the conditional variation model. Claim 4 A method for predicting physical logging data according to claim 1, wherein the conditional variation model is a Conditional Variational Autoencoder (CVAE) implemented with an encoder and a decoder. Claim 5 In claim 1, the long-short term memory model is a Long-Short Term Memory (LSTM), a method for predicting physical logging data. Claim 6 A method for predicting physical logging data according to claim 1, wherein the step of receiving physical logging data includes the step of scaling the physical logging data to normalize it to a predetermined value range. Claim 7 A method for predicting physical logging data according to claim 6, wherein the step of receiving physical logging data further comprises the step of verifying the number of samples for each of a plurality of physical logging items belonging to the physical logging data, comparing the number of samples between the plurality of physical logging items based on the verification result, and, based on the comparison result, augmenting the physical logging data of a specific physical logging item when the difference in the number of samples between the plurality of physical logging items is greater than a predetermined threshold. Claim 8 A method for predicting physical logging data according to claim 1, wherein the physical logging data is a result of measurement for a plurality of physical logging items in which data corresponding to at least one physical logging item is missing. Claim 9 A physical logging data prediction system comprising: a storage unit for storing physical logging data in which a predetermined number of missing data has occurred; and a control unit for estimating first missing data corresponding to said physical logging data using a pre-learned conditional variation model, wherein the control unit predicts rock formation sequence information corresponding to at least one of said physical logging data and said first missing data using a pre-learned long-term memory model, and, as a condition according to said rock formation sequence information, estimates second missing data corresponding to said physical logging data using the pre-learned conditional variation model. Claim 10 A program stored on a computer-readable recording medium that is executed by one or more processes in an electronic device, the program comprising: a step of receiving physical logging data in which a predetermined number of missing data has occurred; a step of estimating a first missing data corresponding to the physical logging data using a pre-learned conditional variation model; a step of predicting a lithographic sequence information corresponding to at least one of the physical logging data and the first missing data using a pre-learned short-term memory model; and a step of estimating a second missing data corresponding to the physical logging data using the pre-learned conditional variation model, conditionally according to the lithographic sequence information.

Images