CN114782744A - Model processing method, device and equipment for reservoir prediction based on closed-loop network - Google Patents

Model processing method, device and equipment for reservoir prediction based on closed-loop network Download PDF

Info

Publication number
CN114782744A
CN114782744A CN202210406342.8A CN202210406342A CN114782744A CN 114782744 A CN114782744 A CN 114782744A CN 202210406342 A CN202210406342 A CN 202210406342A CN 114782744 A CN114782744 A CN 114782744A
Authority
CN
China
Prior art keywords
image
model
reservoir
label
sample
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.)
Pending
Application number
CN202210406342.8A
Other languages
Chinese (zh)
Inventor
陆文凯
宋操
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tsinghua University
Original Assignee
Tsinghua University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Tsinghua University filed Critical Tsinghua University
Priority to CN202210406342.8A priority Critical patent/CN114782744A/en
Publication of CN114782744A publication Critical patent/CN114782744A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • 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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Molecular Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Evolutionary Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Image Analysis (AREA)

Abstract

The embodiment of the application provides a model processing method, a device and equipment for reservoir prediction based on a closed-loop network, wherein the method comprises the following steps: obtaining sample data, wherein the sample data comprises a sample image of a first region, and a first label image and a second label image which correspond to the sample image, the first label image is used for indicating the probability of a reservoir existing in a first part of the first region, and the second label image is used for indicating the probability of a reservoir existing in a second part of the first region; processing the sample image through a first model to obtain a reservoir probability prediction image of the first region; and updating the model parameters of the first model according to the first label image, the second label image and the reservoir probability prediction image. The accuracy of the model for reservoir identification is improved.

Description

Model processing method, device and equipment for reservoir prediction based on closed-loop network
Technical Field
The application relates to the technical field of geophysical, in particular to a model processing method, a device and equipment for reservoir prediction based on a closed-loop network.
Background
A reservoir is a rock formation that may store and permeate fluids. Reservoir prediction can help to identify the distribution condition of the underground oil reservoir, and further the difficulty of oil and gas exploration and development is reduced.
At present, a convolutional neural network can be trained through a one-dimensional label obtained from seismic data, and then a reservoir stratum is identified through the trained convolutional neural network. For example, a plurality of attributes (e.g., ant body, texture attribute, etc.) of the seismic data are spliced along the horizontal direction to obtain a one-dimensional tag, and then the convolutional neural network is trained through the one-dimensional tag. However, reservoirs at the bottom of the ground have spatial continuity, the probability that the reservoirs exist in the space is influenced mutually, the convolutional neural network is trained through the one-dimensional label, only the longitudinal one-dimensional depth characteristic is used, and therefore the accuracy of the model for reservoir identification is low.
Disclosure of Invention
The embodiment of the application provides a method, a device and equipment for processing a model of reservoir prediction based on a closed-loop network, which are used for solving the technical problem that the accuracy of the model for reservoir identification is low in the prior art.
In a first aspect, an embodiment of the present application provides a model processing method for reservoir prediction based on a closed-loop network, where the method includes:
obtaining sample data, wherein the sample data comprises a sample image of a first region, a first label image and a second label image, the first label image corresponds to the sample image, the first label image is used for indicating the probability that a reservoir exists in a first partial region of the first region, and the second label image is used for indicating the probability that a reservoir exists in a second partial region of the first region;
processing the sample image through a first model to obtain a reservoir probability prediction image of the first region;
and updating the model parameters of the first model according to the first label image, the second label image and the reservoir probability prediction image.
In one possible embodiment, acquiring a first label image corresponding to the sample image includes:
processing the sample images through a second model to obtain first label subimages, wherein the second model is obtained by learning multiple groups of first samples, and each group of first samples comprises a sample image and a sample label subimage of a first area;
and adding a first mask in the first label subimage to obtain the first label image, wherein the first mask is used for shielding the second partial area corresponding to the sample image.
In a possible embodiment, obtaining a second label image corresponding to the sample image includes:
performing polynomial linear regression processing on the sample image to obtain a second label subimage;
and adding a second mask in the second label subimage to obtain the second label image, wherein the second mask is used for shielding the first partial area corresponding to the sample image.
In one possible implementation, the first model includes an encoder, a converter, and a decoder; processing the sample image through a first model to obtain a predicted reservoir probability image of the first region, wherein the predicted reservoir probability image comprises the following steps:
obtaining, by the encoder, features of the sample image and sending the features of the sample image to the converter;
converting, by the converter, the features of the sample image into feature vectors, and sending the feature vectors to the decoder;
and converting the characteristic vector into a partial characteristic corresponding to the sample image through the decoder, and generating the reservoir probability predicted image according to the partial characteristic.
In one possible embodiment, updating the model parameters of the first model according to the first tag image, the tag label image and the reservoir probability prediction image includes:
obtaining a first weight of a first loss function between the first label image and the reservoir probability predicted image;
obtaining a second weight of a second loss function between the second label image and the reservoir probability predicted image;
updating the model parameters of the first model according to the first loss function, the second loss function, the first weight and the second weight.
In one possible implementation, after updating the model parameters of the first model according to the first label image, the label labeling image and the reservoir probability prediction image, the method further includes:
processing the reservoir probability predicted image through a third model to obtain a target sample image corresponding to the reservoir probability predicted image, wherein the third model is obtained by learning multiple groups of second samples, and each group of second samples comprises the sample reservoir probability predicted image and a sample image of the first area;
and updating the model parameters of the first model according to the sample image and the target sample image.
In a second aspect, the present application provides a method for model processing of reservoir prediction based on a closed-loop network, the method comprising:
acquiring at least one reservoir image set of a first region;
and processing the at least one reservoir image set according to a first model to obtain a reservoir probability predicted image of the first region, wherein the first model is the first model in any one of the first aspect.
In a third aspect, the present application provides a model processing apparatus for reservoir prediction based on a closed-loop network, including an obtaining module, a processing module, and an updating module, where:
the obtaining module is used for obtaining sample data, wherein the sample data comprises a sample image of a first area, a first label image and a second label image, the first label image corresponds to the sample image, the first label image is used for indicating the probability that a reservoir exists in a first partial area of the first area, and the second label image is used for indicating the probability that a reservoir exists in a second partial area of the first area;
the processing module is used for processing the sample image through a first model to obtain a reservoir probability predicted image of the first region;
the updating module is used for updating the model parameters of the first model according to the first label image, the second label image and the reservoir probability prediction image.
In a possible implementation manner, the obtaining module is specifically configured to:
processing the sample images through a second model to obtain first label subimages, wherein the second model is obtained by learning multiple groups of first samples, and each group of first samples comprises a sample image and a sample label subimage of a first area;
and adding a first mask in the first label subimage to obtain the first label image, wherein the first mask is used for shielding the second partial area corresponding to the sample image.
In a possible implementation manner, the obtaining module is specifically configured to:
performing polynomial linear regression processing on the sample image to obtain a second label subimage;
and adding a second mask in the second label sub-image to obtain the second label image, wherein the second mask is used for blocking the first partial area corresponding to the sample image.
In a possible implementation, the processing module is specifically configured to:
obtaining, by the encoder, features of the sample image and sending the features of the sample image to the converter;
converting, by the converter, the features of the sample image into feature vectors, and sending the feature vectors to the decoder;
and converting the characteristic vector into a partial characteristic corresponding to the sample image through the decoder, and generating the reservoir probability predicted image according to the partial characteristic.
In a possible implementation manner, the update module is specifically configured to:
obtaining a first weight of a first loss function between the first label image and the reservoir probability predicted image;
obtaining a second weight of a second loss function between the second tag image and the reservoir probability prediction image;
updating the model parameters of the first model according to the first loss function, the second loss function, the first weight and the second weight.
In a possible implementation, the processing module is further configured to:
processing the reservoir probability predicted image through a third model to obtain a target sample image corresponding to the reservoir probability predicted image, wherein the third model is obtained by learning multiple groups of second samples, and each group of second samples comprises the sample reservoir probability predicted image and a sample image of the first area;
and updating the model parameters of the first model according to the sample image and the target sample image.
In a fourth aspect, the present application provides a model processing apparatus for reservoir prediction based on a closed-loop network, including an obtaining module and a processing module, wherein:
the acquisition module is used for acquiring at least one reservoir image set of the first area;
the processing module is configured to process the at least one reservoir image set according to a first model to obtain a reservoir probability predicted image of the first region, where the first model is the first model according to any one of claims 1 to 6.
In a fifth aspect, an embodiment of the present application provides a model processing apparatus for reservoir prediction based on a closed-loop network, including: a memory for storing program instructions, a processor for invoking the program instructions in the memory to perform the method of model processing for closed-loop network-based reservoir prediction according to any of the first aspects, or the method of model processing for closed-loop network-based reservoir prediction according to any of the second aspects, and a communication interface.
In a sixth aspect, an embodiment of the present application provides a readable storage medium, on which a computer program is stored; the computer program is for implementing a method of model processing for closed-loop network-based reservoir prediction according to any of the first aspects.
The embodiment of the application provides a method, a device and equipment for processing a model of reservoir prediction based on a closed-loop network, sample data is obtained, wherein the sample data comprises a sample image of a first area, and a first label image and a second label image which correspond to the sample image, the first label image is used for indicating the probability of a reservoir existing in a first partial area in the first area, the second label image is used for indicating the probability of a reservoir existing in a second partial area in the first area, the sample image is processed through the first model to obtain a reservoir probability predicted image of the first area, and model parameters of the first model are updated according to the first label image, the second label image and the reservoir probability predicted image. In the method, the sample data of the first model comprises the two-dimensional label image, and the first label image and the second label image can accurately reflect the probability that the reservoirs exist in different partial areas under the first area, so that the label accuracy of the first model is high, and the label is a two-dimensional image label, and can accurately reflect the spatial continuity between the reservoirs, thereby improving the accuracy of the first model in identifying the reservoirs.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the embodiments or the description of the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and those skilled in the art can obtain other drawings based on the drawings without inventive labor.
Fig. 1 is a schematic view of an application scenario provided in an embodiment of the present application;
fig. 2 is a schematic flowchart of a model processing method for reservoir prediction based on a closed-loop network according to an embodiment of the present disclosure;
FIG. 3 is a schematic diagram of a sample image provided by an embodiment of the present application;
fig. 4 is a schematic structural diagram of a first model provided in an embodiment of the present application;
FIG. 5 is a schematic diagram of model parameters for updating a first model according to an embodiment of the present disclosure;
fig. 6 is a schematic flowchart of another model processing method for reservoir prediction based on a closed-loop network according to an embodiment of the present application;
FIG. 7 is a schematic diagram of a process for obtaining a reservoir probability identification model according to an embodiment of the present disclosure;
FIG. 8 is a schematic diagram illustrating a process for obtaining a first label sub-image of a exemplar according to an embodiment of the present application;
FIG. 9 is a schematic diagram of another process for updating model parameters of a first model according to an embodiment of the present application;
fig. 10 is a schematic flowchart of a model processing method for reservoir prediction based on a closed-loop network according to an embodiment of the present application;
fig. 11 is a schematic diagram of a reservoir probability prediction image provided in an embodiment of the present application;
fig. 12 is a schematic structural diagram of a model processing apparatus for reservoir prediction based on a closed-loop network according to an embodiment of the present application;
fig. 13 is a schematic structural diagram of a terminal device provided in the present application.
Detailed Description
Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, like numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the exemplary embodiments below are not intended to represent all implementations consistent with the present disclosure. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present disclosure, as detailed in the appended claims.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
In the related technology, reservoir prediction can help to identify the distribution condition of the underground oil reservoir, so that oil and gas exploration and development are assisted, and the cost and difficulty of the oil and gas exploration and development are reduced. Currently, the reservoir may be identified by a trained convolutional neural network, where the training samples of the convolutional neural network may include one-dimensional labels obtained through seismic data. For example, a plurality of attributes of seismic data are spliced along the horizontal direction to obtain a one-dimensional label, and the one-dimensional label is used as a label in a training sample of a convolutional neural network to train the convolutional neural network. However, the convolutional neural network obtained through one-dimensional label training can only identify longitudinal one-dimensional depth features in seismic data, reservoirs at the ground have spatial continuity, and the probability of the reservoirs in the space can affect each other, so that the accuracy of the model for reservoir identification is low.
In order to solve the technical problem that the accuracy of reservoir identification by a model in the related art is low, the embodiment of the present application provides a method for model processing for reservoir prediction based on a closed-loop network, obtaining sample data, where the sample data includes a sample image of a first region, a first label image and a second label image corresponding to the sample image, the first label image is used to indicate a probability that a first part of the first region has a reservoir, the second label image is used to indicate a probability that a second part of the first region has a reservoir, obtaining features of the sample image by an encoder of a first model, sending the features of the sample image to a converter of the first model, receiving the features of the sample image by the converter, converting the features of the sample image into feature vectors, sending the feature vectors to a decoder of the first model, receiving the feature vectors by the decoder, and converting the characteristic vector into a partial characteristic corresponding to the sample image, generating a reservoir probability predicted image according to the partial characteristic, and updating the model parameter of the first model according to the first label image, the second label image and the reservoir probability predicted image. Therefore, the training sample data of the first model comprise the two-dimensional label images, and the first label image and the second label image can accurately reflect the probability that reservoirs exist in different partial regions under the first region, so that the label accuracy of the first model is high, the label is a two-dimensional image label, the spatial continuity between the reservoirs can be accurately reflected, and the accuracy of the first model in reservoir identification is improved.
Next, an application scenario of the present application will be described with reference to fig. 1.
Fig. 1 is a schematic view of an application scenario provided in an embodiment of the present application. Please refer to fig. 1, which includes: a first model. The method comprises the steps of inputting a sample image, a first label image and a second label image into a first model, processing the sample image by the first model to obtain a reservoir probability prediction image, and updating model parameters of the first model according to the reservoir probability prediction image, the first label image and the second label image. The training sample data of the first model comprises the two-dimensional label image, and the first label image and the second label image can accurately reflect the probability of the reservoir existing in different partial areas under the first area, so that the label accuracy of the first model is high, the label is a two-dimensional image label, the spatial continuity between the reservoirs can be accurately reflected, and the accuracy of the first model in identifying the reservoir is improved.
The following describes the technical solutions of the present application and how to solve the above technical problems with specific embodiments. These several specific embodiments may be combined with each other below, and details of the same or similar concepts or processes may not be repeated in some embodiments. Embodiments of the present application will be described below with reference to the accompanying drawings.
Fig. 2 is a schematic flowchart of a model processing method for reservoir prediction based on a closed-loop network according to an embodiment of the present disclosure. Referring to fig. 2, the method may include:
s201, obtaining sample data.
The execution subject of the embodiment of the application can be a server, and can also be a model processing device which is arranged in the server and is based on reservoir prediction of a closed-loop network. The model processing device for reservoir prediction based on the closed-loop network can be realized by software, and the model processing device for reservoir prediction based on the closed-loop network can also be realized by the combination of software and hardware.
Optionally, the sample data includes a sample image of the first area, and a first label image and a second label image corresponding to the sample image. The first region may be a region in which reservoir prediction is to be performed. For example, if reservoir prediction is required for region a, the first region may be region a, and if reservoir prediction is required for region B, the first region may be region B. Alternatively, the first zone may be the zone in which the well is logged. For example, the logging is a well for producing natural gas, the location of the logging may be predetermined, and the depth region of the geology where the logging is located may be determined as the first region. For example, in practical applications, a three-dimensional geological model including a plurality of logs may be constructed, and a cross-section of the log may be determined as the first region.
Alternatively, the sample image may be seismic data of the first region. The seismic data may include at least one of: seismic data, texture properties, dip derivative properties, attenuation properties, absorption properties, and ant bodies. Alternatively, the seismic data may be information such as the intensity of the seismic in the first region. For example, the seismic data may be a two-dimensional image and the seismic data for the first region may be displayed by a seismic intensity image. Optionally, the texture property, the dip derivative property, the attenuation property, the absorption property and the ant body are geological properties with a higher degree of correlation with the reservoir. For example, texture properties, dip derivative properties, attenuation properties, absorption properties, and ant bodies may all be represented by two-dimensional images.
Alternatively, seismic data for the first region may be acquired in a database. For example, when the server acquires a seismic intensity image of the first region, the image may be stored in a database, and when the reservoir probability of the first region needs to be predicted, the server may acquire the seismic intensity image in the database.
Optionally, texture properties, dip derivative properties, attenuation properties, absorption properties, and ant bodies of the first region may be obtained from the seismic data. For example, the seismic data (two-dimensional image) is processed by a preset algorithm (e.g., a geological analysis algorithm), so as to obtain a texture attribute image, an inclination derivative attribute image, an attenuation attribute image, an absorption attribute image, and an ant body image of the first region.
Next, the sample image will be described with reference to fig. 3.
Fig. 3 is a schematic diagram of a sample image according to an embodiment of the present disclosure. See fig. 3, which includes a seismic data image. The seismic data image is an image of a cross-section geological surface of a well log, and the seismic data image comprises the well log. And processing the seismic data image through a preset algorithm to obtain a texture attribute image, an inclination derivative attribute image, an attenuation attribute image, an absorption attribute image and an ant body image of the cross-section geological surface.
Optionally, the first label image is used to indicate a probability that the reservoir exists in a first partial region of the first region. For example, the first label image may be a distribution image of reservoir probabilities of the first partial region, and the first label image may be used to obtain probabilities that reservoirs exist at various positions in the first partial region. Optionally, the first partial region may be a region within a preset logging range. For example, the first region may be a cross-sectional geological surface including a log, and the first sub-region in the first region may be a region within 5 meters of the log.
Optionally, the first label image corresponding to the sample image may be obtained through the following feasible implementation manners: and processing the sample image through the second model to obtain a first label subimage. The second model is obtained by learning a plurality of groups of first samples, and each group of first samples comprises a sample image of the first area and a first label subimage of the sample. The sets of first samples may be pre-labeled samples. For example, for a sample image 1, a sample first label sub-image 1 corresponding to the sample image 1 is obtained, and a group of samples is obtained, where the group of samples includes the sample image 1 and the sample first label sub-image 1. In this way, multiple sets of first samples may be obtained. For example, multiple sets of first samples may be as shown in table 1:
TABLE 1
Multiple sets of first samples Sample image Sample first label subimage
First group of first samples Sample image 1 Sample first label subimage 1
Second group of first samples Sample image 2 Sample first label subimage 2
Third group of first samples Sample image 3 Sample first label subimage 3
…… …… ……
It should be noted that table 1 illustrates the plurality of sets of first samples by way of example only, and does not limit the plurality of sets of first samples.
For example, if the seismic data input into the second model is the sample image 1, the second model outputs the first tag sub-image corresponding to the seismic data as the sample first tag sub-image 1; if the seismic data input into the second model are sample images 2, the second model outputs first label subimages corresponding to the seismic data as sample first label subimages 2; if the seismic data input into the second model is the sample image 3, the second model outputs the first label subimage corresponding to the seismic data as the sample first label subimage 3.
Optionally, a first mask is added to the first label sub-image to obtain a first label image. The first mask is used for blocking a second partial area corresponding to the sample image. For example, the first label sub-image may accurately reflect the probability of the reservoir existing in the first partial region of the first region, and therefore, when the first model is trained, the second partial region may be occluded by the first mask, so as to improve the accuracy of the reservoir identification by the first model.
Optionally, the second label image is used to indicate a probability that the reservoir exists in the second partial region of the first region. For example, the second label image may be a distribution image of the reservoir probability of the second partial region, and the probability that the reservoir exists at each position in the second partial region may be obtained through the second label image. Optionally, the second partial region may be a region outside a preset logging range. For example, the first region may be a cross-sectional geological surface including a log, and the second sub-region in the first region may be a region outside of 5 meters around the log. Optionally, the first region includes a first partial region and a second partial region. For example, the first sub-region may be a region within a predetermined range around the logging in the first region, and the second sub-region may be a region outside the predetermined range around the logging in the first region.
Optionally, the second label image corresponding to the sample image may be obtained according to the following feasible implementation manners: and performing polynomial linear regression processing on the sample image to obtain a second label subimage. For example, when obtaining the seismic data of the first region, the relationship between the seismic data and the reservoir probability distribution may be fitted by a regression analysis method (e.g., a least square method, etc.), so as to obtain the reservoir probability distribution map of the first region, i.e., the second tag sub-image.
And adding a second mask in the second label subimage to obtain a second label image. The second mask is used for shielding a first partial area corresponding to the sample image. For example, the accuracy of the second label sub-image obtained by the polynomial linear regression method is unknown, that is, the second label sub-image is a pseudo label, but since the first label image can accurately reflect the probability that the first partial region has a reservoir, a mask needs to be added to the first partial region of the second label sub-image to obtain the second label image, so that the first model can learn the reservoir probability of the second partial region in the second label image.
S202, processing the sample image through the first model to obtain a reservoir probability prediction image of the first area.
Optionally, the reservoir probability predicted image is used to indicate the probability of the reservoir existing in each portion of the first region. For example, by predicting the image according to the reservoir probability, the part of the first area where the reservoir exists can be accurately determined, and therefore the exploration of oil and gas is facilitated.
Optionally, the first model includes an encoder, a converter and a decoder, and the reservoir probability prediction image of the first region may be obtained through the following feasible implementation manners: features of the sample image are obtained by the encoder and sent to the converter. For example, the encoder may include 3 convolutional layers, and the input sample image may be downsampled by the 3 convolutional layers, thereby obtaining the features of the sample image.
The features of the sample image are converted into feature vectors by a converter and the feature vectors are sent to a decoder. For example, the converter part may include 9 residual neural network ResNets structures, each ResNets structure including 2 convolution layers, by which the features of the sample image may be converted into feature vectors and the original features of the sample image may be retained.
And converting the characteristic vector into a partial characteristic corresponding to the sample image through a decoder, and generating a reservoir probability predicted image according to the partial characteristic. Wherein the partial features may be low-level features in the sample image. For example, the decoder includes 3 deconvolution layers, the feature vector can be up-sampled by the 3 deconvolution layers, so as to restore the low-level features of the sample image in the feature vector, and obtain a reservoir probability prediction image according to the low-level features.
Next, the structure of the first model will be described with reference to fig. 4.
Fig. 4 is a schematic structural diagram of a first model according to an embodiment of the present disclosure. Please refer to fig. 4, which includes: a first model. Wherein the first model comprises an encoder, a converter and a decoder. The encoder comprises a convolution layer A, a convolution layer B and a convolution layer C, the converter comprises 9 residual error neural networks, each residual error neural network comprises 2 convolution layers, and the converter comprises 18 convolution layers. The decoder includes a deconvolution layer A, a deconvolution layer B, and a deconvolution layer C.
Referring to fig. 4, when the first model receives the sample image, the encoder in the first model may down-sample the sample image to extract features in the sample image, the converter may convert the features in the sample image into feature vectors, and the decoder may up-sample the feature vectors to restore low-level features in the sample image and obtain a predicted reservoir probability image according to the low-level features.
And S203, updating model parameters of the first model according to the first label image, the second label image and the reservoir probability prediction image.
Optionally, the model parameters of the first model may be updated according to the following feasible implementation manners: a first weight of a first loss function between the first tagged image and the reservoir probability predicted image is obtained. Wherein the first loss function is used to adjust model parameters of the first model. Optionally, the first weight is a weight of the first loss function when adjusting the model parameter. Alternatively, the first loss function may be obtained by the following formula:
Figure BDA0003602299070000111
wherein,
Figure BDA0003602299070000112
is a first loss function, YaugIs the first tag sub-image (two-dimensional augmented well log tag), maskaugAs the first mask, in practical application, the first mask is used to determine whether a certain point belongs to the first partial region (logging and well bypass region), fWp(X) is a first model, and Wp is a model parameter of the first model.
Optionally, maskaugThe formula of (1) is as follows:
Figure BDA0003602299070000113
wherein j is the logging depth, the logging depth range is 0 to α, k is the column number corresponding to the logging position, the range is r- β to r + β, the total number of the well side channels is 2 β, and optionally, β may be 20.
A second weight of a second loss function between the second tag image and the reservoir probability prediction image is obtained. Wherein the second loss function is used to adjust model parameters of the first model. Optionally, the second weight is a proportion of the second loss function in adjusting the model parameter. Alternatively, the second loss function may be obtained by the following equation:
Figure BDA0003602299070000114
wherein,
Figure BDA0003602299070000115
is a second loss function, YpvugFor the second label sub-image, maskpvugAs the second mask, in practical application, the second mask is used to determine whether a certain point belongs to the second partial region (the region outside the well logging and well bypass), fWp(X) is a first model, and Wp is a model parameter of the first model.
Optionally, maskpvugThe formula of (1) is as follows:
Figure BDA0003602299070000116
wherein j is the logging depth, the logging depth range is 0 to alpha, k is the column number corresponding to the logging position, the range is r-beta to r + beta, the total number of the well side channels is 2 beta, and optionally, beta can be 20.
And updating the model parameters of the first model according to the first loss function, the second loss function, the first weight and the second weight. For example, the entire loss function of the first model is obtained by the first weight, the second weight, the first loss function, and the second loss function, and the model parameters of the first model are updated by the entire loss function. For example, a first product of a first weight and a first loss function is obtained, a second product of a second weight and a second loss function is obtained, and the sum of the first product and the second product is determined as the loss function of the first model as a whole.
Next, a process of updating the model parameters of the first model will be described with reference to fig. 5.
Fig. 5 is a schematic diagram of model parameters for updating a first model according to an embodiment of the present disclosure. Referring to fig. 5, a first model is included. And inputting the sample image into the first model, and obtaining a reservoir probability prediction image corresponding to the sample image by the first model. The first loss function is obtained by the first model through the first label image and the reservoir probability prediction image, and the second loss function is obtained by the first model through the second label image and the reservoir probability prediction image. The first model updates the model parameters of the first model through the first loss function and the second loss function, and when the model parameters are updated, the complete loss function of the first model can be obtained through the first weight corresponding to the first loss function and the second weight corresponding to the second loss function, so that the model parameters are updated.
The embodiment of the application provides a model processing method for reservoir prediction based on a closed-loop network, which comprises the steps of obtaining sample data, wherein the sample data comprises a sample image of a first area, a first label image and a second label image, the first label image corresponds to the sample image, the first label image is used for indicating the probability that a first part area in the first area has a reservoir, the second label image is used for indicating the probability that a second part area in the first area has a reservoir, obtaining the characteristics of the sample image through an encoder of the first model, sending the characteristics of the sample image to a converter of the first model, receiving the characteristics of the sample image through the converter, converting the characteristics of the sample image into characteristic vectors, sending the characteristic vectors to a decoder of the first model, receiving the characteristic vectors through the decoder, converting the characteristic vectors into part characteristics corresponding to the sample image, and generating a reservoir probability predicted image according to the partial features, and updating the model parameters of the first model according to the first tag image, the second tag image and the reservoir probability predicted image. Therefore, the training sample data of the first model comprises the two-dimensional label image, and the first label image and the second label image can accurately reflect the probability that reservoirs exist in different partial regions under the first region, so that the label accuracy of the first model is high, the label is a two-dimensional image label, the spatial continuity between the reservoirs can be accurately reflected, and the reservoir identification accuracy of the first model is further improved.
Based on the embodiment shown in fig. 2, the model processing method for reservoir prediction based on the closed-loop network is further described below with reference to fig. 6.
Fig. 6 is a schematic flowchart of another model processing method for reservoir prediction based on a closed-loop network according to an embodiment of the present disclosure. Referring to fig. 6, the method flow includes:
and S601, acquiring sample data.
Optionally, the sample data includes a sample image of the first region, and a first label image and a second label image corresponding to the sample image, where the first label image is used to indicate a probability that a reservoir exists in a first partial region in the first region, and the second label image is used to indicate a probability that a reservoir exists in a second partial region in the first region.
Optionally, the second model may be obtained by training the second model through the sample image and the sample first label subimage, and the second model may also be obtained by iterative training of the model. Alternatively, the second model may be obtained by the following feasible implementation: and further training the reservoir probability recognition model to obtain a second model. The reservoir probability identification model is obtained by training a sample image and a sample reservoir probability prediction image. For example, the training samples of the reservoir probability identification model are the sample image and the artificially labeled reservoir probability prediction image, and the accuracy of the artificially labeled reservoir probability prediction image is low, so that the reservoir probability identification model can be further trained to improve the accuracy of reservoir identification. Optionally, when the reservoir probability recognition model is further trained to obtain a second model, a one-dimensional logging label of a cross-section geological surface may be obtained, where the one-dimensional logging label may be determined by existing seismic data, and therefore, the one-dimensional logging label may accurately reflect the probability that a reservoir exists in a preset region around a log, and since the reservoir probability recognition model is a two-dimensional training model, a layer of mask may be added on the surface of the one-dimensional logging label to obtain a two-dimensional logging label, where only the reservoir probability in the preset region around the log is accurate (the one-dimensional label is accurate), and further, the reservoir probability recognition model is further trained through a sample image and the two-dimensional logging label (the one-dimensional logging label with the mask added), and after the training is completed, the second model is obtained. Next, a process of obtaining the reservoir probability recognition model will be described with reference to fig. 7.
Fig. 7 is a schematic diagram of a process for obtaining a reservoir probability identification model according to an embodiment of the present disclosure. See fig. 7, which includes a reservoir probabilistic identification model. The method comprises the steps of inputting a sample image into a reservoir probability identification model, outputting a reservoir probability predicted image by the reservoir probability identification model, determining a loss function of the reservoir probability identification model through a pseudo tag and the reservoir probability predicted image, updating model parameters of the reservoir probability identification model through the loss function, and processing a two-dimensional image by the reservoir probability identification model through the training method although the pseudo tag is low in accuracy. Optionally, the loss function of the reservoir probability identification model may be determined by the following formula:
Figure BDA0003602299070000131
wherein L isopenIdentifying a loss function for the model for the reservoir probability; y ispvugIs a false label; f. ofWpAnd (X) is a reservoir probability recognition model, and Wp is a model parameter of the reservoir probability recognition model.
Next, a process of the second model training will be described with reference to fig. 8.
Fig. 8 is a schematic process diagram of a second model training process according to an embodiment of the present disclosure. Please refer to fig. 8, which includes: and a second model. And the second model is a trained reservoir probability recognition model. The sample image is input into the second model, and the second model outputs a first label sub-image (reservoir probability prediction image output by the reservoir identification model). And adding a layer of mask on the surface of the one-dimensional logging label to obtain a two-dimensional logging label, determining a loss function through the first label subimage and the two-dimensional logging label, and updating model parameters in the second model through the loss function so that the second model can output an accurate first label subimage.
Alternatively, the loss function of the second model may be determined by the following equation:
Figure BDA0003602299070000132
wherein L isopenA loss function that is a second model; y iswellIs a one-dimensional well logging label; f. ofWp(X) is a second model, and Wp is a model parameter of the second model; an h _ is a dot product; maskwellAnd the mask is a mask corresponding to the one-dimensional logging label.
The formula for the mask may be:
Figure BDA0003602299070000141
wherein, maskwellThe mask corresponding to the one-dimensional logging label is used, j is the logging depth, the logging depth range is 0-alpha, k is the column number corresponding to the logging position, and r is the column number corresponding to the logging.
Optionally, when the sample first label subimages are obtained, each group of seismic data may train a reservoir probability recognition model, and then a plurality of sample first label subimages corresponding to a plurality of seismic data may be obtained in a multi-model processing manner. For example, if 79 one-dimensional well-logging labels are determined from the seismic data, 79 second models are trained, respectively.
S602, processing the sample image through the first model to obtain a reservoir probability prediction image of the first region.
It should be noted that, the execution process of step S602 may refer to step S202, which is not described again in this embodiment of the application.
And S603, updating the model parameters of the first model according to the first label image, the second label image and the reservoir probability predicted image.
It should be noted that, the step S203 may be referred to in the execution process of the step S603, and details of this embodiment are not described herein again.
And S604, processing the reservoir probability predicted image through the third model to obtain a target sample image corresponding to the reservoir probability predicted image.
Optionally, the third model is obtained by learning multiple sets of the second samples. And each group of second samples comprises a sample reservoir probability prediction image and a sample image of the first region. The sets of second samples may be pre-labeled samples. For example, for the sample reservoir probability predicted image 1, the sample image 1 corresponding to the sample reservoir probability predicted image 1 is obtained, and a group of samples is obtained, wherein the group of samples comprises the sample reservoir probability predicted image 1 and the sample image. In this way, a plurality of sets of second samples can be obtained. For example, the sets of second samples may be as shown in table 2:
TABLE 2
Multiple sets of second samples Sample reservoir probability predictive image Sample image
First set of second samples Sample reservoir probability prediction image 1 Sample image 1
Second group of second samples Sample reservoir probability prediction image 2 Sample image 2
Third set of second samples Sample reservoir probability prediction image 3 Sample image 3
…… …… ……
It should be noted that table 2 illustrates the plurality of second samples by way of example only, and does not limit the plurality of first samples.
For example, if the image input into the third model is the sample reservoir probability predicted image 1, the third model outputs the sample image 1 corresponding to the sample reservoir probability predicted image 1; if the image input into the third model is a sample reservoir probability predicted image 2, the third model outputs a sample image 2 corresponding to the sample reservoir probability predicted image 1; and if the image input into the third model is the sample reservoir probability predicted image 3, the third model outputs the sample image 3 corresponding to the sample reservoir probability predicted image 1.
And S605, updating the model parameters of the first model according to the sample image and the target sample image.
Optionally, the reservoir probability image is input into a third model, and the third model may generate a target sample image corresponding to the reservoir probability image. Optionally, the target sample image may be seismic data corresponding to the reservoir probability image. For example, the target sample image may include a seismic data image, a texture attribute image, an inclination derivative attribute image, an attenuation attribute image, an absorption attribute image, and an ant body image corresponding to the reservoir probability image.
Through the sample image and the target sample image, a third loss function can be determined, and then the model parameters of the first model are updated through the third loss function. For example, a third loss function corresponding to the first model can be obtained through the sample image and the target sample image, and then the complete loss function of the first model is updated according to a third weight occupied by the third loss function in the adjustment model parameters, and then the model parameters of the first model are updated through the updated complete loss function. Thus, the better the loop consistency, the higher the accuracy of the reservoir probability prediction image output by the first model.
Optionally, when the model parameters of the first model are updated, the model parameters of the first model may also be updated through the first label image and the second label image. For example, a first label image and a second label image are input into a third model, the third model outputs a corresponding first sample image, a fifth loss function and a sixth loss function are determined according to the sample image and the first sample image, a fifth weight of the fifth loss function and a sixth weight of the sixth loss function are determined, a complete loss function of the first model is updated according to the fifth weight and the fifth loss function, and model parameters of the first model are updated.
Optionally, after the first sample image is obtained through the third model, the first sample image may be processed through the first model to obtain a target reservoir probability prediction image, and the first model parameter is updated according to the target reservoir probability image, the first tag image, and the second tag image. For example, after the first sample image is processed by the first model, a target reservoir probability predicted image may be obtained, a seventh loss function and a seventh weight corresponding to the seventh loss function may be obtained by the target reservoir probability predicted image and the first tag image, an eighth loss function and an eighth weight may be obtained by the target reservoir probability predicted image and the second tag image, and the complete loss function of the first model may be updated by the seventh loss function, the seventh weight, the eighth loss function, and the eighth weight, and the model parameters of the first model may be updated.
Next, a process of updating the model parameters of the first model will be described with reference to fig. 9.
Fig. 9 is a schematic diagram of another process for updating model parameters of a first model according to an embodiment of the present application. Please refer to fig. 9, which includes a first model and a third model. Wherein X is a sample image, XcycleIn order to be the target sample image,
Figure BDA0003602299070000161
predicting an image for reservoir probability, YaugAs a first label subgraphImage, maskaugAs a first mask, YpvugFor the second label sub-image, maskpvugIs the second mask. And inputting the sample image to the first model, and outputting a reservoir probability prediction image by the first model.
The first model obtains a first label image through the synthesis of a first label sub-image and a first mask, obtains a second label image through the synthesis of a second label sub-image and a second mask, determines a first loss function through the first label image and a reservoir probability prediction image, and determines a second loss function through the second label image and the reservoir probability prediction image.
Referring to fig. 9, a composite image of the reservoir probability prediction image, the first label image and the second label image is input to a third model, the third model may output a target sample image, and a composite image of the reservoir probability prediction image and the second label image is input to the third model, which may output another target sample image. The third loss function and the fourth loss function of the first model can be determined from the two target sample images and the sample image.
Referring to fig. 9, the first label image with the mask and the second label image are input into the third model at the same time to obtain two corresponding target sample images
Figure BDA0003602299070000162
From the sample image, the two target sample images, the first mask and the second mask, a fifth loss function and a sixth loss function can be determined. Inputting the two target sample images into the first model again to obtain two reservoir probability predicted images Y after one cyclecycleAnd further through one of Ycycle、YaugAnd the first mask, to obtain a seventh loss function, passing another Ycycle、YpvugAnd a second mask to obtain an eighth loss function. Further, by the first loss function, the second loss function, the third loss function, the fourth loss function, the fifth loss function, the sixth loss function, the seventh loss function, the eighth loss function, and the weight corresponding to each loss function,an overall loss function of the first model is determined, and model parameters of the first model are updated through the overall loss function.
Alternatively, the formula of the third loss function is as follows:
Figure BDA0003602299070000163
wherein,
Figure BDA0003602299070000164
as a third loss function, YaugFor the first label sub-image, maskaugAs a first mask, YpvugFor the second label subimage, maskpvuhFor the second mask, fWp(X) is a first model, Wp is a model parameter of the first model, fWGThe WG is a model parameter of the third model.
The formula for the fourth loss function is as follows:
Figure BDA0003602299070000171
wherein,
Figure BDA0003602299070000172
is a fourth loss function, YpvugFor the second label sub-image, maskpvugFor the second mask, fWp(X*) Is a first model, X*As target sample image, Wp is the model parameter of the first model, fWGThe WG is a model parameter of the third model.
The formula for the fifth loss function is as follows:
Figure BDA0003602299070000173
wherein,
Figure BDA0003602299070000174
as a fifth loss function, YaugIs the first label subimage, X is the sample image, maskaugIs a first mask, fWGIs a third model.
The formula for the sixth loss function is as follows:
Figure BDA0003602299070000175
wherein,
Figure BDA0003602299070000176
as a sixth loss function, YpvugIs a second label subimage, X is a sample image, maskpvugAs a second mask, fWGIs the third model.
The formula for the seventh loss function is as follows:
Figure BDA0003602299070000177
wherein,
Figure BDA0003602299070000178
is a seventh loss function, YaugFor the first label sub-image, maskaugIs a first mask, fWFFirst model, f, being a cyclic inputWGIs a third model.
The formula of the eighth loss function is as follows:
Figure BDA0003602299070000179
wherein,
Figure BDA00036022990700001710
as an eighth loss function, YpvugFor the second label sub-image, maskpvugFor the second mask, fWFFirst model, f, input for cyclesWGIs the third model.
Alternatively, the overall loss function of the first model may be expressed as follows:
Figure BDA00036022990700001711
wherein L iscloseIs a loss function of the first model population, λ1、λ2、λ3Is the weight used to adjust the 8 loss functions.
The embodiment of the application provides a reservoir prediction model processing method based on a closed-loop network, sample data is obtained, a sample image is processed through a first model to obtain a reservoir probability prediction image of a first region, model parameters of the first model are updated according to the first tag image, a second tag image and the reservoir probability prediction image, the reservoir probability prediction image is processed through a third model to obtain a target sample image corresponding to the reservoir probability prediction image, and the model parameters of the first model are updated according to the sample image and the target sample image. Therefore, the training sample data of the first model comprise the two-dimensional label image, and the first label image and the second label image can accurately reflect the probability that reservoirs exist in different partial regions under the first region, so that the label accuracy of the first model is high, the label is a two-dimensional image label, the spatial continuity between the reservoirs can be accurately reflected, the model parameters of the first model can be updated through a closed-loop network, and the accuracy of the first model in reservoir identification is improved.
On the basis of any of the above embodiments, a process of performing model processing on the first model will be described below with reference to fig. 10.
Fig. 10 is a schematic flowchart of a model processing method for reservoir prediction based on a closed-loop network according to an embodiment of the present application. Referring to fig. 10, the method flow includes:
s1001, acquiring at least one reservoir image set of the first area.
Alternatively, the set of reservoir images are sample images in the embodiment shown in FIG. 2. For example, a seismic data image, a texture attribute image, an inclination derivative attribute image, an attenuation attribute image, an absorption attribute image, and an ant body image of the first region may be included in the reservoir image collection.
S1002, processing at least one reservoir image set according to the first model to obtain a reservoir probability prediction image of the first region.
Optionally, the first model is the first model in any one of the above embodiments. And inputting the reservoir image set of the first region into the first model, wherein the first model can output a reservoir probability prediction image corresponding to the first region.
Next, a reservoir probability prediction image will be described with reference to fig. 11.
Fig. 11 is a schematic diagram of a reservoir probability prediction image provided in an embodiment of the present application. Referring to fig. 11, the first image and the second image are included. The first image is a reservoir probability predicted image obtained according to the prior art, and the second image is a reservoir probability predicted image obtained according to the first model in the application. The position indicated by the arrow is the position of the log. In the first image, the one-dimensional label is used as a training sample of the model, so that the spatial continuity between the obtained reservoirs is poor, and the accuracy of reservoir prediction is low. The spatial continuity between reservoirs in the second image obtained through the first model of the application is good, and the accuracy of reservoir prediction is high.
The embodiment of the application provides a reservoir prediction model processing method based on a closed-loop network, which comprises the steps of obtaining at least one reservoir image set of a first area, and processing the at least one reservoir image set according to a first model to obtain a reservoir probability prediction image of the first area. Wherein the first model is the first model in any of the above embodiments. Because the first model can accurately determine the probability of the reservoir existing in the first area, the accuracy of reservoir prediction can be improved through the first model.
Fig. 12 is a schematic structural diagram of a model processing apparatus for reservoir prediction based on a closed-loop network according to an embodiment of the present disclosure. Referring to fig. 12, the model processing apparatus 10 for reservoir prediction based on closed-loop network includes an obtaining module 11, a processing module 12 and an updating module 13, wherein:
the obtaining module 11 is configured to obtain sample data, where the sample data includes a sample image of a first region, and a first tag image and a second tag image corresponding to the sample image, where the first tag image is used to indicate a probability that a reservoir exists in a first partial region of the first region, and the second tag image is used to indicate a probability that a reservoir exists in a second partial region of the first region;
the processing module 12 is configured to process the sample image through a first model to obtain a reservoir probability prediction image of the first region;
the updating module 13 is configured to update the model parameter of the first model according to the first tag image, the second tag image, and the reservoir probability prediction image.
In a possible implementation manner, the obtaining module 11 is specifically configured to:
processing the sample images through a second model to obtain first label subimages, wherein the second model is obtained by learning multiple groups of first samples, and each group of first samples comprises the sample images and the sample label subimages of a first area;
and adding a first mask in the first label sub-image to obtain the first label image, wherein the first mask is used for blocking the second partial area corresponding to the sample image.
In a possible implementation, the obtaining module 11 is specifically configured to:
performing polynomial linear regression processing on the sample image to obtain a second label subimage;
and adding a second mask in the second label subimage to obtain the second label image, wherein the second mask is used for shielding the first partial area corresponding to the sample image.
In a possible implementation, the processing module 12 is specifically configured to:
acquiring, by the encoder, features of the sample image and sending the features of the sample image to the converter;
converting, by the converter, the features of the sample image into feature vectors, and sending the feature vectors to the decoder;
and converting the characteristic vector into a partial characteristic corresponding to the sample image through the decoder, and generating the reservoir probability prediction image according to the partial characteristic.
In a possible implementation, the updating module 13 is specifically configured to:
obtaining a first weight of a first loss function between the first label image and the reservoir probability predicted image;
obtaining a second weight of a second loss function between the second label image and the reservoir probability predicted image;
updating the model parameters of the first model according to the first loss function, the second loss function, the first weight and the second weight.
In a possible implementation, the processing module 12 is further configured to:
processing the reservoir probability predicted image through a third model to obtain a target sample image corresponding to the reservoir probability predicted image, wherein the third model is obtained by learning multiple groups of second samples, and each group of second samples comprises the sample reservoir probability predicted image and a sample image of the first area;
and updating the model parameters of the first model according to the sample image and the target sample image.
The model processing device for reservoir prediction based on the closed-loop network provided by the embodiment of the application can execute the technical scheme shown in the embodiment of the method, the implementation principle and the beneficial effect are similar, and details are not repeated here.
The application also comprises another structure of a model processing device for reservoir prediction based on a closed-loop network, which comprises an acquisition module and a processing module, wherein:
the acquisition module is used for acquiring at least one reservoir image set of the first area;
the processing module is configured to process the at least one reservoir image set according to a first model to obtain a reservoir probability prediction image of the first region, where the first model is the first model according to any one of claims 1 to 6.
Fig. 13 is a schematic structural diagram of a terminal device provided in the present application. Referring to fig. 13, the terminal device 20 may include: a processor 21 and a memory 22, wherein the processor 21 and the memory 22 may be in communication; illustratively, the processor 21 and the memory 22 are in communication via a communication bus 23, the memory 22 is configured to store program instructions, and the processor 21 is configured to call the program instructions in the memory to perform a model processing method for closed-loop network-based reservoir prediction as illustrated in any of the method embodiments described above.
Optionally, the terminal device 20 may further comprise a communication interface, which may comprise a transmitter and/or a receiver.
Optionally, the Processor may be a Central Processing Unit (CPU), or may be another general-purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), or the like. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of a method disclosed in connection with the present application may be embodied directly in a hardware processor, or in a combination of the hardware and software modules in the processor.
A readable storage medium having a computer program stored thereon; the computer program is for implementing a model processing method for closed-loop network-based reservoir prediction as described in any of the embodiments above.
Embodiments of the present application provide a computer program product comprising instructions that, when executed, cause a computer to perform the above-described method for model processing for closed-loop network-based reservoir prediction.
All or a portion of the steps of implementing the above-described method embodiments may be performed by hardware associated with program instructions. The aforementioned program may be stored in a readable memory. When executed, the program performs steps comprising the method embodiments described above; and the aforementioned memory (storage medium) includes: read-only memory (ROM), RAM, flash memory, hard disk, solid state disk, magnetic tape (magnetic tape), floppy disk (flexible disk), optical disk (optical disk), and any combination thereof.
Embodiments of the present application are described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processing unit of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processing unit of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
It will be apparent to those skilled in the art that various changes and modifications can be made in the embodiments of the present application without departing from the spirit and scope of the application. Thus, if such modifications and variations of the embodiments of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to encompass such modifications and variations.
In this application, the terms "include," "includes," and variations thereof may refer to non-limiting inclusions; the term "or" and variations thereof may mean "and/or". The terms "first," "second," and the like in this application are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. In the present application, "a plurality" means two or more. "and/or" describes the association relationship of the associated objects, meaning that there may be three relationships, e.g., a and/or B, which may mean: a exists alone, A and B exist simultaneously, and B exists alone. The character "/" generally indicates that the former and latter associated objects are in an "or" relationship.

Claims (10)

1. A model processing method for reservoir prediction based on a closed-loop network is characterized by comprising the following steps:
obtaining sample data, wherein the sample data comprises a sample image of a first region, a first label image and a second label image, the first label image corresponds to the sample image, the first label image is used for indicating the probability that a reservoir exists in a first partial region of the first region, and the second label image is used for indicating the probability that a reservoir exists in a second partial region of the first region;
processing the sample image through a first model to obtain a reservoir probability prediction image of the first region;
and updating the model parameters of the first model according to the first label image, the second label image and the reservoir probability prediction image.
2. The method of claim 1, wherein obtaining the first label image corresponding to the sample image comprises:
processing the sample images through a second model to obtain first label subimages, wherein the second model is obtained by learning multiple groups of first samples, and each group of first samples comprises a sample image and a sample label subimage of a first area;
and adding a first mask in the first label sub-image to obtain the first label image, wherein the first mask is used for blocking the second partial area corresponding to the sample image.
3. The method of claim 1 or 2, wherein obtaining a second label image corresponding to the specimen image comprises:
performing polynomial linear regression processing on the sample image to obtain a second label subimage;
and adding a second mask in the second label sub-image to obtain the second label image, wherein the second mask is used for blocking the first partial area corresponding to the sample image.
4. The method of any of claims 1-3, wherein the first model comprises an encoder, a converter, and a decoder; processing the sample image through a first model to obtain a predicted reservoir probability image of the first region, wherein the predicted reservoir probability image comprises the following steps:
obtaining, by the encoder, features of the sample image and sending the features of the sample image to the converter;
converting, by the converter, the features of the sample image into feature vectors, and sending the feature vectors to the decoder;
and converting the characteristic vector into a partial characteristic corresponding to the sample image through the decoder, and generating the reservoir probability predicted image according to the partial characteristic.
5. The method according to any of claims 1-4, wherein updating the model parameters of the first model based on the first tag image, the tag label image, and the reservoir probability prediction image comprises:
obtaining a first weight of a first loss function between the first tag image and the reservoir probability prediction image;
obtaining a second weight of a second loss function between the second tag image and the reservoir probability prediction image;
updating the model parameters of the first model according to the first loss function, the second loss function, the first weight and the second weight.
6. The method according to any of claims 1-5, wherein after updating the model parameters of the first model based on the first label image, the label labeling image, and the reservoir probability prediction image, the method further comprises:
processing the reservoir probability predicted image through a third model to obtain a target sample image corresponding to the reservoir probability predicted image, wherein the third model is obtained by learning multiple groups of second samples, and each group of second samples comprises the sample reservoir probability predicted image and a sample image of the first area;
and updating the model parameters of the first model according to the sample image and the target sample image.
7. A model processing method for reservoir prediction based on a closed-loop network is characterized by comprising the following steps:
acquiring at least one reservoir image set of a first region;
processing the at least one reservoir image set according to a first model to obtain a reservoir probability prediction image of the first region, wherein the first model is the first model according to any one of claims 1 to 6.
8. A model processing device for reservoir prediction based on a closed-loop network is characterized by comprising an acquisition module, a processing module and an updating module, wherein:
the obtaining module is used for obtaining sample data, wherein the sample data comprises a sample image of a first area, and a first label image and a second label image which correspond to the sample image, the first label image is used for indicating the probability that a reservoir exists in a first part of the first area, and the second label image is used for indicating the probability that a reservoir exists in a second part of the first area;
the processing module is used for processing the sample image through a first model to obtain a reservoir probability prediction image of the first region;
the updating module is used for updating the model parameters of the first model according to the first label image, the second label image and the reservoir probability predicted image.
9. A terminal device, comprising: a processor coupled with a memory;
the memory is used for storing a computer program;
the processor is configured to execute the computer program stored in the memory to cause the terminal device to perform the method for model processing of closed-loop network-based reservoir prediction according to any one of the preceding claims 1-6, or the method for model processing of closed-loop network-based reservoir prediction according to claim 7.
10. A readable storage medium comprising a program or instructions for the method of model processing for closed loop network based reservoir prediction as claimed in any one of claims 1 to 6 or the method of model processing for closed loop network based reservoir prediction as claimed in claim 7 when the program or instructions are run on a computer.
CN202210406342.8A 2022-04-18 2022-04-18 Model processing method, device and equipment for reservoir prediction based on closed-loop network Pending CN114782744A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210406342.8A CN114782744A (en) 2022-04-18 2022-04-18 Model processing method, device and equipment for reservoir prediction based on closed-loop network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210406342.8A CN114782744A (en) 2022-04-18 2022-04-18 Model processing method, device and equipment for reservoir prediction based on closed-loop network

Publications (1)

Publication Number Publication Date
CN114782744A true CN114782744A (en) 2022-07-22

Family

ID=82431047

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210406342.8A Pending CN114782744A (en) 2022-04-18 2022-04-18 Model processing method, device and equipment for reservoir prediction based on closed-loop network

Country Status (1)

Country Link
CN (1) CN114782744A (en)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105593703A (en) * 2013-08-13 2016-05-18 界标制图有限公司 A simulation-to-seismic workflow construed from core based rock typing and enhanced by rock replacement modeling
US20200292722A1 (en) * 2019-03-12 2020-09-17 Saudi Arabian Oil Company Method for Dynamic Calibration and Simultaneous Closed-Loop Inversion of Simulation Models of Fractured Reservoirs
CN113109874A (en) * 2021-03-30 2021-07-13 中国石油大学(华东) Wave impedance inversion method using neural network and neural network system

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105593703A (en) * 2013-08-13 2016-05-18 界标制图有限公司 A simulation-to-seismic workflow construed from core based rock typing and enhanced by rock replacement modeling
US20200292722A1 (en) * 2019-03-12 2020-09-17 Saudi Arabian Oil Company Method for Dynamic Calibration and Simultaneous Closed-Loop Inversion of Simulation Models of Fractured Reservoirs
CN113109874A (en) * 2021-03-30 2021-07-13 中国石油大学(华东) Wave impedance inversion method using neural network and neural network system

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
NANZHE WANG等: "Deep learning based closed-loop optimization of geothermal reservoir production", HTTPS://DOI.ORG/10.48550/ARXIV.2204.08987, 15 April 2022 (2022-04-15) *
安鹏等: "基于ＬＳＴＭ 循环神经网络的储层物性参数预测方法研究", 《地球物理学进展》, vol. 34, no. 5, 31 May 2019 (2019-05-31), pages 1849 - 1858 *
董志明等: "基于闭环CNN-GRU 和迁移学习的地震波阻抗反演", 第四届油气地球物理学术年会论文集, 30 June 2021 (2021-06-30) *

Similar Documents

Publication Publication Date Title
CN111160276A (en) U-shaped cavity full-volume integral cutting network identification model based on remote sensing image
CN112731522B (en) Intelligent recognition method, device and equipment for seismic stratum and storage medium
CN110991284B (en) Optical remote sensing image statement description generation method based on scene pre-classification
CN110555841A (en) SAR image change detection method based on self-attention image fusion and DEC
CN113109874B (en) Wave impedance inversion method using neural network and neural network system
Wang et al. Seismic stratum segmentation using an encoder–decoder convolutional neural network
CN116400426B (en) Electromagnetic method-based data survey system
CN115830471A (en) Multi-scale feature fusion and alignment domain self-adaptive cloud detection method
CN115081719A (en) Earthquake phase automatic identification method based on combination of self-attention mechanism and U-shaped structure
CN115393656A (en) Automatic classification method for stratum classification of logging-while-drilling image
CN115308803A (en) Coal seam thickness prediction method, device, equipment and medium
CN114913434B (en) High-resolution remote sensing image change detection method based on global relation reasoning
CN116071346A (en) Heterogeneous double-time image change detection method and device
Asif et al. Automated transient electromagnetic data processing for ground-based and airborne systems by a deep learning expert system
CN117689960B (en) Lithology scene classification model construction method and classification method
CN118348326A (en) Electromagnetic spectrum map mapping method for three-dimensional complex urban environment
CN118091789A (en) Magnetotelluric data inversion method, magnetotelluric data inversion device, electronic equipment and readable storage medium
CN114782744A (en) Model processing method, device and equipment for reservoir prediction based on closed-loop network
CN116740521A (en) Landslide detection method and system based on multi-scale feature fusion
Yuan et al. Horizon auto-picking with quantitative uncertainty evaluation by using a modified VQ-VAE framework
CN115937565A (en) Hyperspectral image classification method based on self-adaptive L-BFGS algorithm
CN114879263A (en) Three-dimensional magnetic tensor gradient inversion method and device based on convolutional neural network
CN113920311A (en) Remote sensing image segmentation method and system based on edge auxiliary information
Jo et al. Conditioning stratigraphic, rule-Based models with generative adversarial networks: A deepwater lobe, deep learning example
Dou et al. Automatic identification of GPR targets on roads based on CNN and Grad-CAM

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