CN116226623B - Mark layer division method and device based on SegNet segmentation model and computer equipment - Google Patents
Mark layer division method and device based on SegNet segmentation model and computer equipment Download PDFInfo
- Publication number
- CN116226623B CN116226623B CN202310144716.8A CN202310144716A CN116226623B CN 116226623 B CN116226623 B CN 116226623B CN 202310144716 A CN202310144716 A CN 202310144716A CN 116226623 B CN116226623 B CN 116226623B
- Authority
- CN
- China
- Prior art keywords
- model
- data
- data set
- training
- layer
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 230000011218 segmentation Effects 0.000 title claims abstract description 61
- 238000000034 method Methods 0.000 title claims abstract description 43
- 238000012549 training Methods 0.000 claims abstract description 81
- 238000012545 processing Methods 0.000 claims abstract description 29
- 238000012360 testing method Methods 0.000 claims abstract description 25
- 238000012795 verification Methods 0.000 claims abstract description 14
- 238000004140 cleaning Methods 0.000 claims abstract description 12
- 238000005520 cutting process Methods 0.000 claims abstract description 8
- 239000003550 marker Substances 0.000 claims description 15
- 238000004458 analytical method Methods 0.000 claims description 13
- 238000004590 computer program Methods 0.000 claims description 10
- 238000000638 solvent extraction Methods 0.000 claims description 10
- 238000009826 distribution Methods 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 4
- 230000003044 adaptive effect Effects 0.000 claims description 3
- 238000010276 construction Methods 0.000 claims description 3
- 230000006870 function Effects 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 8
- 230000000694 effects Effects 0.000 description 7
- 230000015572 biosynthetic process Effects 0.000 description 6
- 238000005755 formation reaction Methods 0.000 description 6
- 238000003860 storage Methods 0.000 description 6
- 238000003709 image segmentation Methods 0.000 description 5
- 238000004891 communication Methods 0.000 description 4
- 238000013527 convolutional neural network Methods 0.000 description 4
- 238000013473 artificial intelligence Methods 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000010200 validation analysis Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000000919 ceramic Substances 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 230000008094 contradictory effect Effects 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000005553 drilling Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 239000003208 petroleum Substances 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/21—Design, administration or maintenance of databases
- G06F16/215—Improving data quality; Data cleansing, e.g. de-duplication, removing invalid entries or correcting typographical errors
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The application discloses a mark layer dividing method, device and computer equipment based on a SegNet segmentation model. The method comprises the following steps: collecting log curve data and standard layer data; cleaning, cutting off and complementing the selected logging curve data; performing standardized processing on the selected logging curve data and standard layer data, and constructing a training data set, a verification data set and a test data set; constructing a full-segment model data set and a segment model data set; constructing a SegNet model based on the VGG16 model, and training the SegNet model by utilizing the full-segment model data set and the segment model data set respectively; determining a segmentation model with highest precision; and inputting the test data set into an optimal segment model, predicting the marking layer data of the well to be layered, and calculating to obtain marking layer depth data. In the application, the prediction error depth of the segmented model is greatly reduced compared with the overall model error, and the application provides a novel high-precision intelligent prediction method for stratum division, contrast and mark layer identification.
Description
Technical Field
The application relates to the technical field of petroleum exploration, in particular to a mark layer dividing method, device and computer equipment based on a SegNet segmentation model.
Background
The oil and gas reservoir stratum division and comparison system layer is one of key steps of oil reservoir description and reservoir characterization, the result of the stratum system layer directly determines an oil reservoir grid, the spatial distribution of an oil reservoir body in the oil reservoir is further controlled, and the development of the oil and gas reservoir is finally affected. A marking layer refers to a layer or group of formations that have distinct characteristics that can be used as a contrast marking for the formation. The marking layer should have the characteristics of obvious fossil and lithology characteristics, stable horizon, wide distribution range and easy identification. The general steps of stratum system layer are to identify the mark layer, build stratum section on the basis of the mark layer, and then to identify and compare other stratum between layers.
The core work of marker layer identification is achieved by geology researchers through logging curves in single well shafts, footed area geology awareness and specific stratum contrast modes through manual division. The manual interpretation by geological researchers depends on the professional knowledge level, experience knowledge and the like of the researchers to a great extent, and the division results of different researchers can have large differences, so that unified comparison standards are difficult to establish.
In order to improve the working efficiency, students at home and abroad develop a large number of semi-automatic and automatic stratum intelligent dividing and comparing research works realized by means of computers by means of methods and technologies such as signal processing, mathematical statistics, artificial intelligence and the like. The intelligent stratum dividing and comparing reduces some problems existing in the manual interpretation process to a certain extent, such as non-uniform comparing modes and standards, difficulty in determining stratum interfaces and the like, and greatly improves the working efficiency of stratum dividing and comparing.
The surge of Artificial Intelligence (AI) and Deep Learning (DL) is being rolled around the world. Deep learning is good at mining abstract feature representations from the raw input data, with good generalization capability. The convolutional neural network (ConvolutionalNeural Networks, CNN for short) has achieved a series of breakthrough successes in the fields of image classification, target positioning, target detection, image semantic segmentation and the like due to the strong feature learning and classifying capability, and is communicated with feature extraction and analysis of a logging curve in artificial stratum comparison.
In recent years, with the development of artificial intelligence technology, a large number of machine learning algorithms are applied to intelligent division and comparison research of strata, and have made progress. At present, the scheme adopts the thought of CNN image segmentation to carry out stratum division, and specific models comprise SegNet, U-Net and the like. Xu Chaohui and 2019 are based on SegNet algorithm, based on well logging data of a dense well pattern area, automatic stratum division and comparison are achieved through 3 logging curves such as natural potential and microelectrode, and good effects are achieved. The SegNet method is applied to the ceramic group stratum in a winning area, and results show that compared with the traditional manual method, the SegNet method can realize accurate automatic stratum division and improve stratum interpretation efficiency. However, the inventors have realized that these stratigraphic partitioning methods often utilize an overall model, and that there are still large errors in the partitioning accuracy, although higher than manual.
Disclosure of Invention
Based on the above, aiming at the technical problems, a mark layer dividing method, a mark layer dividing device and computer equipment based on a SegNet segmentation model are provided, so that the technical problem that larger errors still exist in the existing stratum dividing method is solved.
In order to achieve the above object, the present application provides the following technical solutions:
in a first aspect, a method for partitioning a marker layer based on SegNet segmentation model includes:
s1, collecting logging curve data and standard layer data, and selecting a plurality of logging variable characteristics in the logging curve data;
s2, cleaning the selected logging curve data, and cutting off and completing the selected logging curve data according to the depth range of the standard layer data;
s3, carrying out standardized processing on the selected logging curve data and standard layer data, and constructing a training data set, a verification data set and a test data set;
s4, carrying out segmentation processing on the training data set to construct a full-segment model data set and a segmentation model data set;
s5, constructing a SegNet model based on the VGG16 model, and training the SegNet model by utilizing the full-segment model data set and the segment model data set respectively to obtain a corresponding full-segment model and segment model;
s6, performing precision analysis on the full-segment model and the segment model, and determining the segment model with highest precision;
s7, inputting the test data set into an optimal segmentation model, predicting the marking layer data of the well to be layered, and calculating to obtain marking layer depth data.
Optionally, the depth of the logging curve data is 300m, and the sampling interval of the logging curve data is 0.05m; the depth range of the standard layer data is 200m.
Optionally, the plurality of logging variable features includes micro-gradients, micro-potentials, natural gamma, and natural potentials.
Optionally, step S3 includes:
performing Z-score standardization and 0-1 normalization treatment on the selected logging curve data;
performing pixel 0-255 space processing on the selected logging curve data, storing the logging curve data as a logging curve training picture, and recording a depth range represented by a current picture;
standardizing the standard layer data, generating a layered label picture with the same dimension as the well logging curve picture, and establishing a well logging curve training picture and a layered label picture pair data set;
dividing the data set according to the logging curve training pictures and the layered label pictures according to a preset proportion, and constructing a training data set, a verification data set and a test data set.
Further alternatively, the preset ratio is 6:2:2.
Optionally, in step S5, an adaptive moment-estimation optimizer is used in training the SegNet model, the learning rate is 0.0001, the batch size is 32, the training batch is 50, and ACC is used as a loss function to evaluate the error between the predicted value and the true value.
Optionally, the standard layer data includes five flag layers, and the segment model data set includes three segment data sets and five segment data sets.
Optionally, the method further comprises:
introducing wavelet data, and decomposing logging curve data by selecting a Coif5 wavelet method to obtain 7 low-frequency signals a 1-a 7 and 7 high-frequency signals d 1-d 7;
d6 and d7 signals of the plurality of logging variable characteristics are selected and added to training.
In a second aspect, a sign layer dividing apparatus based on SegNet segmentation model includes:
the data acquisition module is used for collecting logging curve data and standard layer data and selecting a plurality of logging variable characteristics in the logging curve data;
the data cleaning module is used for cleaning the selected logging curve data and cutting off and completing the selected logging curve data according to the depth range of the standard layer data;
the data set construction module is used for carrying out standardized processing on the selected logging curve data and standard layer data and constructing a training data set, a verification data set and a test data set;
the data set segmentation processing module is used for carrying out segmentation processing on the training data set to construct a full-segment model data set and a segmentation model data set;
the training module is used for constructing a SegNet model based on the VGG16 model, and training the SegNet model by utilizing the full-segment model data set and the segment model data set respectively to obtain a corresponding full-segment model and segment model;
the model precision analysis module is used for carrying out precision analysis on the full-segment model and the segment model and determining the segment model with the highest precision;
and the marking layer prediction module is used for inputting the test data set into an optimal segmentation model, predicting marking layer data of the well to be layered, and calculating to obtain marking layer depth data.
In a third aspect, a computer device comprises a memory storing a computer program and a processor implementing the steps of the method of any of the first aspects when the computer program is executed.
The application has at least the following beneficial effects:
the embodiment of the application provides a mark layer dividing method based on a SegNet segmentation model, which comprises the steps of collecting logging curve data and standard layer data; cleaning, cutting off and complementing the selected logging curve data; performing standardized processing on the selected logging curve data and standard layer data, and constructing a training data set, a verification data set and a test data set; constructing a full-segment model data set and a segment model data set; constructing a SegNet model based on the VGG16 model, and training the SegNet model by utilizing the full-segment model data set and the segment model data set respectively; determining a segmentation model with highest precision; inputting the test data set into an optimal segment model, predicting the mark layer data of the well to be layered, and calculating to obtain the mark layer depth data; the method is characterized in that a segNet network model with excellent performance in the image segmentation field is used as a basis, a segmented model data set is constructed to carry out multi-model segmented training, and after a proper segmented multi-model is determined, logging curve data obtained by drilling a well to be layered is used for predicting the marking layer data of the well; in the method, the average prediction error depth of the segment model is 0.31m, and the error is greatly reduced compared with the overall model, so the application provides a novel high-precision intelligent prediction method for stratum division, contrast and mark layer identification.
Drawings
FIG. 1 is a schematic flow chart of a method for dividing a marker layer based on a SegNet segmentation model according to an embodiment of the present application;
FIG. 2 is a schematic diagram of 12 curve samples constructed in one embodiment of the present application;
FIG. 3 is a plot of formation depth versus position before and after data truncation in accordance with one embodiment of the present application;
FIG. 4 is a graph of formation depth versus position before and after data completion in accordance with one embodiment of the present application;
FIG. 5 is a plot of formation depth versus position after data cleaning in accordance with one embodiment of the present application;
FIG. 6 is a log data picture and a label classification picture in one embodiment of the application;
FIG. 7 is a schematic diagram of a model in accordance with an embodiment of the application;
FIG. 8 is a diagram of a full segment mode model training effect in one embodiment of the application;
FIG. 9 is a schematic diagram of three-segment model training data creation in accordance with one embodiment of the present application;
FIG. 10 is a training error comparison of a three-segment model and a full-segment model in accordance with one embodiment of the present application;
FIG. 11 is a diagram illustrating the creation of five-segment model training data in accordance with one embodiment of the present application;
FIG. 12 is a training error comparison of a five-segment model with a whole and three-segment model in accordance with one embodiment of the present application;
FIG. 13 is a block diagram of a module architecture of a logo layer device based on a SegNet segmentation model according to an embodiment of the present application;
fig. 14 is an internal structural diagram of a computer device according to an embodiment of the present application.
Detailed Description
The present application will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present application more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application.
In one embodiment, as shown in fig. 1, there is provided a method for dividing a marker layer based on SegNet segmentation model, including the following steps:
s1, collecting logging curve data and standard layer data, and selecting a plurality of logging variable characteristics in the logging curve data.
The collected log data and standard layer data include known wells and wells to be zoned. The data collected in this example are as follows: the standard layer data comprises five mark layers (C1, C2, C3, C4 and C5), the depth range of the standard layer data is about 200m, the depth of a logging curve is about 300m, the sampling interval of the logging curve is 0.05m, and four logging variable characteristics of micro gradient (RMN), micro potential (RMG), natural Gamma (GR) and natural potential (SP) are selected as network inputs.
Meanwhile, wavelet data are introduced, a Coif5 wavelet method is selected to decompose logging signals (logging curve data), and finally 7 low-frequency signals a 1-a 7 and 7 high-frequency signals d 1-d 7 are obtained. D6 and d7 of the four curves RMN, RMG, GR, SP are selected and added to the sample training. The total of 12 sample curves (fig. 2, red horizontal line is the stratum position).
In general, this step completes data acquisition, analysis, and screening: and collecting log curve data and stratum data, determining a target log curve range (RMN, RMG, SP, GR), analyzing the log curve data and the depth range of the stratum data, and selecting effective data.
S2, cleaning the selected logging curve data, and cutting off and completing the selected logging curve data according to the depth range of the standard layer data.
After data collection, the data is cleaned, and firstly, common problems in logging data such as invalid values, missing values and the like are cleaned. Checking whether the data is satisfactory, finding out data which is out of normal range, logically unreasonable or contradictory, and cleaning the data. In addition, according to the depth range of the marking layer data, the sample depth space distribution range of the logging curve is determined, and the overlong logging data are truncated (shown in fig. 3) and are subjected to completion processing (shown in fig. 4).
According to the embodiment, a plurality of attempts are made, and the cutting-off and the completion of the logging data enable stratum data to be more balanced in depth space of a logging curve, so that the precision of a model is greatly improved. The depth space range of the treated marker layer in the log is shown in fig. 5.
In general, this step completes the data cleansing, truncation, and completion: and processing the data invalid value and the missing value, determining a sample depth space of the logging curve according to the depth range of stratum data, and cutting off and complementing corresponding logging curve data.
And S3, carrying out standardization processing on the selected logging curve data and standard layer data, and constructing a training data set, a verification data set and a test data set.
The training data set is constructed from data acquired from known wells and the test data set is constructed from data acquired from wells to be partitioned. The step S3 comprises the following steps:
performing Z-score standardization and 0-1 normalization treatment on the selected logging curve data;
performing pixel 0-255 space processing on the selected logging curve data, storing the logging curve data as a logging curve training picture, and recording a depth range represented by a current picture;
standardizing standard layer data, generating a layered label picture with the same dimension as the well logging curve picture, and establishing a well logging curve training picture and a layered label picture pair data set;
dividing the data set according to the logging curve training pictures and the layered label pictures according to a preset proportion, and constructing a training data set, a verification data set and a test data set.
In other words, the log is normalized first, Z-score normalized first, and then 0-1 normalized. In order to facilitate parallelization model training, data is required to be processed into a picture mode, so that pixel 0-255 space processing is performed, the data is stored as a sample log curve picture, and meanwhile, a Depth range (Pic) represented by a current image is recorded.
And (3) carrying out standardization processing on the standard layer depth data, generating classification pictures with the same dimension as the logging curve pictures, and establishing a logging training picture and layering label picture pair data set (shown in fig. 6). Meanwhile, sample data are divided according to a ratio relation of 6:2:2, and a training data set, a verification data set and a test data set are respectively constructed.
In general, this step completes data normalization and creation of the data set: and (3) carrying out standardization treatment on the logging curve and carrying out standardization treatment on stratum data. A training data set, a validation data set, and a test data set are constructed.
S4, carrying out segmentation processing on the training data set, and constructing a full-segment model data set and a segmentation model data set.
And carrying out sectional processing on the data according to the distribution range of the stratum data in the logging data depth space, and respectively constructing a full-section model data set, a three-section model data set and a five-section model data set.
S5, constructing a SegNet model based on the VGG16 model, and training the SegNet model by utilizing the full-segment model data set and the segment model data set respectively to obtain a corresponding full-segment model and segment model.
S6, performing precision analysis on the full-segment model and the segment model, and determining the segment model with the highest precision;
specifically, the SegNet network structure in this embodiment mainly includes two parts: an encoding network and a decoding network. This network architecture is designed based on the VGG16 model. The data logging training picture is input, the output is also a layered picture, and the model is trained and optimized by comparing the layered picture with the label layered picture. The network structure is shown in fig. 7; through multiple reference adjustment, an adaptive moment estimation (Adam) optimizer is adopted in training, the learning rate is 0.0001, the Batch Size (batch_size) is 32, the training Batch (Epoch) is 50, and the error between the predicted value and the true value is estimated by adopting ACC as a loss function.
Training was performed using full-segment data with training accuracy and validation accuracy of 99.74% and 98.57%, model depth error of 0.79m, and training effect as shown in fig. 8.
And constructing a three-section model data set and performing model training. In view of the SegNet model of the present experiment constructed based on VGG16, the optimal dimension of the input image of VGG16 is 299×299, then the minimum error for longitudinal classification is 1/299×depth (Pic), which is the Depth range of the log, and if trying to reduce the Depth range Depth (Pic) of the log, there is a possibility of reducing the model error. Based on the thought, the experiment carries out sectional treatment on logging data, and adopts two sectional schemes, a three-section scheme and a five-section scheme.
The three-stage scheme builds three models, model 1 predicts the C1 and C2 formations, model 2 predicts C3, model three predicts C4 and C5. The segmentation range for the log is 20% -50%, 50% -70%, 60% -100%. The training data generated is shown in fig. 9.
Using the same SegNet model, the training effect is compared with the previous full segment overall model effect pair such as FIG. 10. The training precision and the verification precision of the three models reach more than 99%, the average depth error of five layers is 0.35m, and compared with the 0.79m of the whole model, the error is obviously reduced.
And constructing a five-segment model data set and performing model training. In view of the improvement of the three-section model precision, the embodiment further performs five-section processing on the logging data so as to obtain a smaller logging image depth image.
The five-segment scheme builds five models, one for each marker layer. The segmentation limit for the log is determined from the analysis in step S2, and is 22-43%, 28-50%, 50-72%, 60-83%, 72-100%, respectively. The training data generated is shown in fig. 11. The training effect is compared with the previous whole model and three-segment model effect pairs such as fig. 12 using the same SegNet model. The training precision and the verification precision of the five models reach more than 99%, the average depth error of the five layering layers is 0.31m, and the error is further reduced compared with the model error of the 3-section model. Thus, a five-segment training model was determined.
In general, steps S5 and S6 complete model building, training, and tuning: based on VGG16, a SegNet model is constructed. Respectively carrying out full-segment model training on the full-segment model data sets, dividing three-segment data sets according to data analysis, and carrying out three-segment model training; further optimizing, dividing the five-section data set, and carrying out five-section model training. And (3) carrying out precision analysis on the full-segment model, the three-segment model and the five-segment model, and determining a segmentation scheme with highest precision.
S7, inputting the test data set into an optimal segmentation model, predicting the marking layer data of the well to be layered, and calculating to obtain marking layer depth data.
And respectively inputting the 5 segments of data into a corresponding prediction model to obtain a predicted value (percentage value) of each marker layer, and calculating to obtain the depth value of each marker layer according to the predicted value of the stratum point and the depth range of the test picture.
In general, the marker layer data is predicted using an optimal segment model: and inputting the test data set into an optimal segmentation model, predicting the marking layer data of the well to be layered, and converting the output data into marking layer depth data.
In the method for dividing the marker layer based on the segNet segmentation model provided by the embodiment, the segNet model is used for dividing the stratum based on the image segmentation thought, and the model is output as a layered image with the same dimension as the input image. The SegNet of the experiment is built by VGG16, and the optimal dimension of the input image is 299×299, so that the minimum error of the longitudinal classification is 1/299×depth (Pic), and the Depth (Pic) is the Depth range of the input image (logging picture). The application tries to reduce the Depth range (Pic) of the input image (logging picture) by using the thought of the segmentation simulation so as to reduce the model error and further improve the model prediction precision.
The application takes a SegNet network model with excellent performance in the image segmentation field as a basis, analyzes the relative positions of a well logging curve and a mark layer of a known well in a depth space, adopts a mode of segmenting the well logging curve and grouping the mark layer, reduces the depth range of a sample image, constructs segmented sample image data and label classification images, takes the segmented sample image data and the label classification images as a training set to carry out multi-model segmentation training, respectively establishes mapping relations between the well logging curve segments and corresponding mark layer data in multiple models, and predicts the mark layer data of the well by using the well logging curve data drilled by a well to be stratified after determining a proper segmented multiple model. The method adopts the idea of constructing multiple models in a segmented way to improve the prediction precision of the mark layer, and secondly adopts the idea of analyzing sample data to determine the range of the depth section of the logging curve corresponding to the mark layer group, thereby scientifically and effectively improving the segmentation scale problem of the sample space. The prediction error depth average of the segment model is 0.31m through the experimental test, and the error is greatly reduced compared with the whole model. In a word, the scheme provides a novel high-precision intelligent prediction method based on the image segmentation thought for the stratum segmentation, contrast and mark layer identification fields.
It should be understood that, although the steps in the flowchart of fig. 1 are shown in sequence as indicated by the arrows, the steps are not necessarily performed in sequence as indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders. Moreover, at least a portion of the steps in fig. 1 may include a plurality of steps or stages, which are not necessarily performed at the same time, but may be performed at different times, and the order of the steps or stages is not necessarily sequential, but may be performed in rotation or alternatively with at least a portion of the steps or stages in other steps or other steps.
In one embodiment, as shown in fig. 13, there is provided a flag layer partitioning apparatus based on SegNet segmentation model, including the following program modules:
the data acquisition module 131 is configured to collect log curve data and standard layer data, and select a plurality of logging variable features in the log curve data;
the data cleaning module 132 is configured to clean the selected log data, and intercept and complement the selected log data according to the depth range of the standard layer data;
the data set construction module 133 is configured to perform standardization processing on the selected log data and standard layer data, and construct a training data set, a verification data set and a test data set;
a data set segmentation processing module 134, configured to perform segmentation processing on the training data set, and construct a full-segment model data set and a segment model data set;
the training module 135 is configured to construct a SegNet model based on the VGG16 model, and train the SegNet model by using the full-segment model dataset and the segment model dataset, respectively, to obtain a corresponding full-segment model and segment model;
the model precision analysis module 136 is configured to perform precision analysis on the full-segment model and the segment model, and determine the segment model with the highest precision;
the mark layer prediction module 137 inputs the test data set into an optimal segment model, predicts the mark layer data of the well to be stratified, and calculates the mark layer depth data.
For specific limitation of a SegNet segmentation model-based flag layer partitioning apparatus, reference may be made to the above limitation of a SegNet segmentation model-based flag layer partitioning method, which is not described herein. The above-mentioned various modules in a SegNet segmentation model-based marker layer division apparatus may be implemented in whole or in part by software, hardware, and combinations thereof. The above modules may be embedded in hardware or may be independent of a processor in the computer device, or may be stored in software in a memory in the computer device, so that the processor may call and execute operations corresponding to the above modules.
In one embodiment, a computer device is provided, which may be a terminal, and the internal structure thereof may be as shown in fig. 14. The computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The communication interface of the computer device is used for carrying out wired or wireless communication with an external terminal, and the wireless mode can be realized through WIFI, an operator network, NFC (near field communication) or other technologies. The computer program, when executed by a processor, implements a marker layer partitioning method based on a SegNet segmentation model. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, can also be keys, a track ball or a touch pad arranged on the shell of the computer equipment, and can also be an external keyboard, a touch pad or a mouse and the like.
It will be appreciated by those skilled in the art that the structure shown in fig. 14 is merely a block diagram of a portion of the structure associated with the present inventive arrangements and is not limiting of the computer device to which the present inventive arrangements are applied, and that a particular computer device may include more or fewer components than shown, or may combine certain components, or have a different arrangement of components.
In one embodiment, a computer device is provided, including a memory and a processor, the memory having stored therein a computer program, involving all or part of the flow of the methods of the embodiments described above.
In one embodiment, a computer readable storage medium having a computer program stored thereon is provided, involving all or part of the flow of the methods of the embodiments described above.
Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in embodiments provided herein may include at least one of non-volatile and volatile memory. The nonvolatile memory may include Read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, or the like. Volatile memory can include Random access memory (Random AccessMemory, RAM) or external cache memory. By way of illustration, and not limitation, RAM can take many forms, such as static random access memory (StaticRandomAccessMemory, SRAM) or dynamic random access memory (DynamicRandomAccessMemory, DRAM), among others.
The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.
The above examples illustrate only a few embodiments of the application, which are described in detail and are not to be construed as limiting the scope of the application. It should be noted that it will be apparent to those skilled in the art that several variations and modifications can be made without departing from the spirit of the application, which are all within the scope of the application. Accordingly, the scope of protection of the present application is to be determined by the appended claims.
Claims (9)
1. A mark layer dividing method based on a SegNet segmentation model comprises the following steps:
s1, collecting logging curve data and standard layer data, and selecting a plurality of logging variable characteristics in the logging curve data; the standard layer data comprises five mark layers, which are marked as C1, C2, C3, C4 and C5;
s2, cleaning the selected logging curve data, and cutting off and completing the selected logging curve data according to the depth range of the standard layer data;
s3, carrying out standardized processing on the selected logging curve data and standard layer data, and constructing a training data set, a verification data set and a test data set;
s4, carrying out segmentation processing on the training data set to construct a full-segment model data set and a segmentation model data set; according to the distribution range of stratum data in the logging data depth space, carrying out segmentation processing on the training data set, and respectively constructing a full-segment model data set, a three-segment model data set and a five-segment model data set;
s5, constructing a SegNet model based on the VGG16 model, and training the SegNet model by utilizing the full-segment model data set and the segment model data set respectively to obtain a corresponding full-segment model and segment model; training by using the three-section model data set to obtain three models, wherein a model I predicts C1 and C2, a model II predicts C3 and a model III predicts C4 and C5, and the segmentation ranges of the logging curves corresponding to the three models are 20% -50%, 50% -70% and 60% -100% respectively; training the five-section model data set to obtain five models, wherein each model corresponds to one mark layer, and the sectional ranges of the logging curves corresponding to the five models are 22-43%, 28-50%, 50-72%, 60-83% and 72-100% respectively;
s6, performing precision analysis on the full-segment model and the segment model, and determining the segment model with highest precision;
s7, inputting the test data set into an optimal segmentation model, predicting the marking layer data of the well to be layered, and calculating to obtain marking layer depth data.
2. The SegNet segmentation model-based marker layer segmentation method as set forth in claim 1, wherein the depth of the log data is 300m and the sampling interval of the log data is 0.05m; the depth range of the standard layer data is 200m.
3. The SegNet segmentation model based marker layer segmentation method as set forth in claim 1, wherein the plurality of logging variable features comprises micro-gradients, micro-potentials, natural gamma, and natural potentials.
4. The SegNet segmentation model-based logo layer segmentation method as claimed in claim 1, wherein step S3 comprises:
performing Z-score standardization and 0-1 normalization treatment on the selected logging curve data;
performing pixel 0-255 space processing on the selected logging curve data, storing the logging curve data as a logging curve training picture, and recording a depth range represented by a current picture;
standardizing the standard layer data, generating a layered label picture with the same dimension as the well logging curve picture, and establishing a well logging curve training picture and a layered label picture pair data set;
dividing the data set according to the logging curve training pictures and the layered label pictures according to a preset proportion, and constructing a training data set, a verification data set and a test data set.
5. The method for partitioning a marker layer based on a SegNet segmentation model as set forth in claim 4, wherein the predetermined ratio is 6:2:2.
6. The method according to claim 1, wherein in step S5, an adaptive moment-estimation optimizer is used for training the SegNet model, the learning rate is 0.0001, the batch size is 32, the training batch is 50, and the ACC is used as a loss function to evaluate the error between the predicted value and the actual value.
7. The SegNet segmentation model-based logo layer partitioning method as claimed in claim 1, further comprising:
introducing wavelet data, and decomposing logging curve data by selecting a Coif5 wavelet method to obtain 7 low-frequency signals a 1-a 7 and 7 high-frequency signals d 1-d 7;
d6 and d7 signals of the plurality of logging variable characteristics are selected and added to training.
8. A sign layer partitioning device based on SegNet segmentation model, comprising:
the data acquisition module is used for collecting logging curve data and standard layer data and selecting a plurality of logging variable characteristics in the logging curve data; the standard layer data comprises five mark layers, which are marked as C1, C2, C3, C4 and C5;
the data cleaning module is used for cleaning the selected logging curve data and cutting off and completing the selected logging curve data according to the depth range of the standard layer data;
the data set construction module is used for carrying out standardized processing on the selected logging curve data and standard layer data and constructing a training data set, a verification data set and a test data set;
the data set segmentation processing module is used for carrying out segmentation processing on the training data set to construct a full-segment model data set and a segmentation model data set; according to the distribution range of stratum data in the logging data depth space, carrying out segmentation processing on the training data set, and respectively constructing a full-segment model data set, a three-segment model data set and a five-segment model data set;
the training module is used for constructing a SegNet model based on the VGG16 model, and training the SegNet model by utilizing the full-segment model data set and the segment model data set respectively to obtain a corresponding full-segment model and segment model; training by using the three-section model data set to obtain three models, wherein a model I predicts C1 and C2, a model II predicts C3 and a model III predicts C4 and C5, and the segmentation ranges of the logging curves corresponding to the three models are 20% -50%, 50% -70% and 60% -100% respectively; training the five-section model data set to obtain five models, wherein each model corresponds to one mark layer, and the sectional ranges of the logging curves corresponding to the five models are 22-43%, 28-50%, 50-72%, 60-83% and 72-100% respectively;
the model precision analysis module is used for carrying out precision analysis on the full-segment model and the segment model and determining the segment model with the highest precision;
and the marking layer prediction module is used for inputting the test data set into an optimal segmentation model, predicting marking layer data of the well to be layered, and calculating to obtain marking layer depth data.
9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any of claims 1 to 7 when the computer program is executed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310144716.8A CN116226623B (en) | 2023-02-21 | 2023-02-21 | Mark layer division method and device based on SegNet segmentation model and computer equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310144716.8A CN116226623B (en) | 2023-02-21 | 2023-02-21 | Mark layer division method and device based on SegNet segmentation model and computer equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116226623A CN116226623A (en) | 2023-06-06 |
| CN116226623B true CN116226623B (en) | 2023-11-21 |
Family
ID=86570957
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310144716.8A Active CN116226623B (en) | 2023-02-21 | 2023-02-21 | Mark layer division method and device based on SegNet segmentation model and computer equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116226623B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110618158A (en) * | 2019-10-28 | 2019-12-27 | 中石化石油工程技术服务有限公司 | Method for constructing capillary pressure curve of rock core by utilizing nuclear magnetic resonance information |
| CN112859192A (en) * | 2019-11-28 | 2021-05-28 | 北京国双科技有限公司 | Rock core saturation prediction model construction method and rock core saturation prediction method |
| CN112862139A (en) * | 2019-11-27 | 2021-05-28 | 北京国双科技有限公司 | Fluid type prediction model construction method, fluid type prediction method and device |
| CN113553780A (en) * | 2021-09-22 | 2021-10-26 | 西南石油大学 | Stratum pore pressure prediction method based on machine learning |
| CN115017791A (en) * | 2021-12-18 | 2022-09-06 | 中国铁道科学研究院集团有限公司电子计算技术研究所 | Tunnel surrounding rock grade identification method and device |
-
2023
- 2023-02-21 CN CN202310144716.8A patent/CN116226623B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110618158A (en) * | 2019-10-28 | 2019-12-27 | 中石化石油工程技术服务有限公司 | Method for constructing capillary pressure curve of rock core by utilizing nuclear magnetic resonance information |
| CN112862139A (en) * | 2019-11-27 | 2021-05-28 | 北京国双科技有限公司 | Fluid type prediction model construction method, fluid type prediction method and device |
| CN112859192A (en) * | 2019-11-28 | 2021-05-28 | 北京国双科技有限公司 | Rock core saturation prediction model construction method and rock core saturation prediction method |
| CN113553780A (en) * | 2021-09-22 | 2021-10-26 | 西南石油大学 | Stratum pore pressure prediction method based on machine learning |
| CN115017791A (en) * | 2021-12-18 | 2022-09-06 | 中国铁道科学研究院集团有限公司电子计算技术研究所 | Tunnel surrounding rock grade identification method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116226623A (en) | 2023-06-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Shi et al. | Development of subsurface geological cross-section from limited site-specific boreholes and prior geological knowledge using iterative convolution XGBoost | |
| CN111783825A (en) | Well logging lithology identification method based on convolutional neural network learning | |
| CN112228054B (en) | Method, device and equipment for determining shale gas yield based on convolutional neural network | |
| CN113642698B (en) | Geophysical well logging intelligent interpretation method, system and storage medium | |
| CN113378736B (en) | Remote sensing image semi-supervised semantic segmentation method based on transformation consistency regularization | |
| CN111472765A (en) | Stratigraphic division method and device for target well | |
| CN112800590B (en) | Grid coarsening method for machine learning-assisted two-phase flow oil reservoir random modeling | |
| CN116226623B (en) | Mark layer division method and device based on SegNet segmentation model and computer equipment | |
| CN116927771A (en) | Method, device, equipment and medium for predicting total organic carbon data of shale reservoir | |
| US20230184087A1 (en) | Multi-modal and Multi-dimensional Geological Core Property Prediction using Unified Machine Learning Modeling | |
| CN114036829B (en) | Geological profile generation method, system, equipment and storage medium | |
| CN115358285A (en) | Method, device and equipment for selecting key geological parameters of block to be surveyed | |
| CN116070089A (en) | Stratum division method and device based on ResNet regression model and computer equipment | |
| Ahmadi et al. | A sensitivity study of FILTERSIM algorithm when applied to DFN modeling | |
| CN115527107A (en) | Carbonate reservoir plane prediction method, system, equipment and storage medium | |
| CN117633658B (en) | Rock reservoir lithology identification method and system | |
| CN116108368B (en) | Deposition microphase identification method and device based on deep learning mixed model | |
| CN117574269B (en) | Intelligent identification method and system for natural cracks of land shale reservoir | |
| CN117235628B (en) | Well logging curve prediction method and system based on hybrid Bayesian deep network | |
| CN111859709B (en) | Geologic statistics simulation method and device for aquifer structure variation transition probability | |
| Cornelio et al. | Neural Network-Assisted Clustering for Improved Production Predictions in Unconventional Reservoirs | |
| CN108108512B (en) | Reservoir lithofacies characterization method and device | |
| Ramcharitar et al. | Using Machine Learning Methods to Identify Reservoir Compartmentalization in Mature Oilfields from Legacy Production Data | |
| Zhao et al. | A Data-Driven Approach for Hydraulic Fracturing Simulation in Shale Based on Time-Series Images of Fracture Propagation | |
| CN115937599A (en) | Deep learning remote sensing image classification method for improving self-adaptive pooling |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |