CN111597753A - Data depth change characteristic self-adaptive two-dimensional resistivity inversion method and system - Google Patents

Data depth change characteristic self-adaptive two-dimensional resistivity inversion method and system Download PDF

Info

Publication number
CN111597753A
CN111597753A CN202010269141.9A CN202010269141A CN111597753A CN 111597753 A CN111597753 A CN 111597753A CN 202010269141 A CN202010269141 A CN 202010269141A CN 111597753 A CN111597753 A CN 111597753A
Authority
CN
China
Prior art keywords
resistivity
inversion
data
adaptive
depth
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.)
Granted
Application number
CN202010269141.9A
Other languages
Chinese (zh)
Other versions
CN111597753B (en
Inventor
刘斌
蒋鹏
郭谦
刘本超
聂利超
刘征宇
汤宇婷
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.)
Shandong University
Original Assignee
Shandong 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 Shandong University filed Critical Shandong University
Priority to CN202010269141.9A priority Critical patent/CN111597753B/en
Publication of CN111597753A publication Critical patent/CN111597753A/en
Application granted granted Critical
Publication of CN111597753B publication Critical patent/CN111597753B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • 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
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/30—Assessment of water resources

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biophysics (AREA)
  • Mathematical Physics (AREA)
  • Computing Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Biomedical Technology (AREA)
  • General Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Geometry (AREA)
  • Medical Informatics (AREA)
  • Computer Hardware Design (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Measurement Of Resistance Or Impedance (AREA)

Abstract

The invention provides a two-dimensional resistivity inversion method and a two-dimensional resistivity inversion system with self-adaptive data depth change characteristics, wherein a data set of apparent resistivity-resistivity model data pairs of different electrical models is constructed; constructing a self-adaptive convolution network with the amplitude and the offset of convolution kernels of different horizon depths and the self-adaptive variable characteristic according to the resistivity depth change; constructing an inversion loss function carrying vertical weight of a resistivity data item, training an adaptive convolution network controlled by the inversion loss function by using the data set, and establishing a mapping relation between apparent resistivity data and a resistivity model; the observation apparent resistivity data is input into the trained self-adaptive convolution network, the resistivity imaging result is output through the established mapping relation, the earth surface two-dimensional resistivity deep learning inversion is realized, the inversion quality can be effectively improved, and particularly the inversion precision of a deep region is improved.

Description

Data depth change characteristic self-adaptive two-dimensional resistivity inversion method and system
Technical Field
The disclosure belongs to the technical field of two-dimensional resistivity inversion, and relates to a data depth change characteristic self-adaptive two-dimensional resistivity inversion method and system.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
Surface two-dimensional resistivity survey is a common geophysical exploration method. Resistivity inversion imaging is a process of inversely calculating the resistivity distribution of an underground medium through observed apparent resistivity data, and is a core problem of resistivity detection. Resistivity inversion is a typical nonlinear problem, the existing general mature method converts a target function into a linear problem for solving through high-order item omission of the target function, and the problems of easy falling into local optimization, strong initial model dependency, insufficient inversion accuracy and the like exist. Aiming at the problems of the existing method, starting from the nonlinear nature of resistivity inversion, the imaging quality of resistivity inversion is improved by utilizing the strong nonlinear fitting capability of a complex function of a novel deep neural network, and the method becomes a brand new scheme for solving the problem of resistivity inversion.
According to the knowledge of the inventor, the current deep learning inversion scheme is not developed and popularized in the field of resistivity inversion, and the core problem is that a deep convolution network for natural image processing has weight sharing property and fixed convolution kernel amplitude. And depending on the resistivity data, the abnormal response characteristic of the resistivity data is different from that of a natural image, namely the abnormal response characteristic of the resistivity data changes along with the change of the depth position, namely the abnormal response characteristic of the resistivity data has a depth change characteristic. The depth-varying characteristics of the apparent resistivity data directly result in: firstly, abnormal features are difficult to effectively capture and distinguish, network output is fuzzy, and accurate inversion imaging is difficult to realize; secondly, the abnormal features of the deep region are not obvious, and the inversion effect of the deep abnormal body is not good.
Disclosure of Invention
In order to solve the problems, the invention provides a two-dimensional resistivity inversion method and a two-dimensional resistivity inversion system with self-adaptive data depth change characteristics.
According to some embodiments, the following technical scheme is adopted in the disclosure:
a data depth change characteristic self-adaptive two-dimensional resistivity inversion method comprises the following steps:
constructing a dataset of apparent resistivity-resistivity model data pairs for different electrical models;
constructing a self-adaptive convolution network with the amplitude and the offset of convolution kernels of different horizon depths and the self-adaptive variable characteristic according to the resistivity depth change;
constructing a resistivity deep learning inversion loss function carrying the vertical weight of the data item;
training the adaptive convolution network controlled by the inversion loss function by using the data set, and establishing a mapping relation between apparent resistivity data and a resistivity model;
and inputting observation apparent resistivity data into the trained adaptive convolution network, and outputting a resistivity imaging result through the established mapping relation to realize surface two-dimensional resistivity deep learning inversion.
In the technical scheme, the convolutional neural network is controlled through an inversion loss function carrying vertical weight of data items, the convolutional neural network is trained by utilizing a typical earth electric model data set, the depth change characteristics of the apparent resistivity can be adapted, the network learning capability of different depth positions is adjusted according to the data characteristics, and the end-to-end mapping relation between the apparent resistivity data and an inversion result is directly established through a depth characteristic adaptive deep learning inversion method.
As an alternative embodiment, the geoelectric model is a single or a plurality of different high/low resistance abnormal body combinations.
As an alternative embodiment, the depth-varying feature adaptive convolutional network includes:
the self-adaptive convolution aiming at the depth change characteristic of the apparent resistivity learns the amplitude and the offset of convolution kernels at different depth positions through network training, and then the self-adaptive convolution kernel of the depth change characteristic with the flexibility matched with the depth change characteristic is as follows:
Figure BDA0002442437240000031
wherein, α (h)k，β(h)k，A(h)kAdaptation parameters for the characteristics of the depth variation to be learned, α (h)kAnd β (h)kIs the horizontal and vertical position of network node k, A (h)kIs a convolution kernel wcFor a convolution kernel of size k and an input of vertical length h, the total amount of additional parameters 3 × k × h is assigned.
As an alternative embodiment, the resistivity deep learning inversion loss function carrying the vertical weight of the data item comprises:
and applying vertical weight of data items in the deep learning inversion loss function to carry out targeted training standard deployment on the depth change characteristics of the apparent resistivity data.
As a further embodiment, the data item vertical weight dw is in the form of:
Figure BDA0002442437240000032
wherein m isi，jIs a predictor of the resistivity model location (i, j), λ is a constant related to the electrode setup size and current electrode location, and parameter β depends on the data type and the dimensionality of the problem.
As a further embodiment, the resistivity deep learning inversion loss function carrying the vertical weight of the data item is:
Figure BDA0002442437240000041
wherein the inversion resistivity value is
Figure BDA0002442437240000042
Model value of resistivity is mi,j。
In the inversion process, the resistivity deep learning inversion loss function carrying the vertical weight of the data item aims at the attenuation characteristics of the earth surface excitation electric field at different vertical direction positions, and the purpose of controlling the self-adaptive convolution network is achieved by weighting and differentiating the inconsistency degree of the model predicted value and the actual value at the different vertical direction positions.
A two-dimensional resistivity inversion system with adaptive data depth variation characteristics, comprising:
a dataset construction module configured to construct a dataset of apparent resistivity-resistivity model data pairs for different electrical models;
the network model building module is configured to build convolution kernel amplitudes and offsets of different horizon depths into an adaptive convolution network which is adaptively variable according to the resistivity depth change characteristics;
the inversion loss function building module is configured to build a resistivity deep learning inversion loss function carrying the vertical weight of the data item;
the training module is configured to train the adaptive convolution network controlled by the inversion loss function by using the data set, and establish a mapping relation between apparent resistivity data and a resistivity model;
and the inversion module is configured to input the observation apparent resistivity data into the trained adaptive convolution network, output the resistivity imaging result through the established mapping relation and realize the surface two-dimensional resistivity deep learning inversion.
A computer readable storage medium having stored therein a plurality of instructions adapted to be loaded by a processor of a terminal device and to perform the steps of a method of data depth variation feature adaptive two-dimensional resistivity inversion.
A terminal device comprising a processor and a computer readable storage medium, the processor being configured to implement instructions; the computer readable storage medium is for storing instructions adapted to be loaded by a processor and for performing the steps of the method for data depth variation feature adaptive two-dimensional resistivity inversion.
Compared with the prior art, the beneficial effect of this disclosure is:
the method adopts the resistivity data depth change characteristic self-adaptive convolution kernel, namely the amplitude and the offset of the convolution kernel which is self-adaptively convolved into different horizon depths are self-adaptively variable according to the resistivity depth change characteristic, the coverage area of the convolution kernel of the deep abnormal characteristic is wider, the problems that the abnormal characteristic is difficult to effectively capture and distinguish and the output of a depth inversion network is fuzzy are solved, and the inversion quality is effectively improved.
According to the method, the resistivity deep learning inversion loss function carrying the vertical weight of the data item is adopted to redistribute the network capacity, the inversion effect of the deep area is effectively improved, and the vertical weight of the data item redistributes the network learning capacity according to the resistivity depth change characteristics, so that the network puts more abnormal characteristic learning capacities into the deep area.
The present disclosure avoids the high order omission of linear methods by a fully nonlinear deep neural network inversion method.
Aiming at the characteristic range increase generated by the response characteristics of the apparent resistivity data along with the depth increase, the characteristic boundary is not obvious, and the deep change characteristics that the abnormal response is not easy to identify are adopted, the amplitude and the offset of convolution kernels at different deep positions are changed through the self-adaptive convolution kernel of the depth change characteristics of the resistivity data, so that the coverage area of the convolution kernels at different deep positions is increased and is self-adaptive to the change of the deep characteristics, the abnormal response characteristics are captured more comprehensively, the variable scale learning of the abnormal response characteristics of the resistivity data with the deep change characteristics is realized, and the inversion precision is effectively improved.
According to the method, the network learning capability is redistributed through the resistivity deep learning inversion loss function carrying the vertical weight of the data item, and the network capability is concentrated to the deep region of the electric field attenuation, so that the inversion accuracy of the deep region is improved.
Drawings
The accompanying drawings, which are included to provide a further understanding of the disclosure, illustrate embodiments of the disclosure and together with the description serve to explain the disclosure and are not to limit the disclosure.
FIG. 1 is a flow chart of a two-dimensional resistivity deep learning inversion method with adaptive data depth variation characteristics;
FIG. 2 is a schematic diagram of the geoelectrical model in the database established in the present embodiment;
fig. 3 shows the deep learning inversion result in this embodiment.
The specific implementation mode is as follows:
the present disclosure is further described with reference to the following drawings and examples.
It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the disclosure. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments according to the present disclosure. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
A two-dimensional resistivity deep learning inversion method with self-adaptive data depth change characteristics comprises the following steps:
constructing a massive, typically electrical model dataset of single or multiple different morphological high/low resistance anomaly combinations, the dataset comprising apparent resistivity-resistivity model data pairs of different electrical models;
and constructing a depth change characteristic self-adaptive convolution network, wherein the amplitude and the offset of convolution kernels which are self-adaptively convolved into different horizon depths are self-adaptively variable according to the resistivity depth change characteristic, and the coverage area of the convolution kernels of the deep abnormal characteristics is wider.
And constructing a resistivity deep learning inversion loss function carrying the vertical weight of the data item, and redistributing the network learning ability by the vertical weight of the data item according to the resistivity depth change characteristics, so that the network puts more abnormal characteristic learning ability into a deep region.
Training a mass typical electrical model data set through a depth change characteristic self-adaptive convolution network controlled by a resistivity deep learning inversion loss function carrying data item vertical weight, and establishing a complex mapping relation between apparent resistivity data and a resistivity model through the network.
And inputting observation apparent resistivity data, and outputting a resistivity imaging result through established network mapping to realize surface two-dimensional resistivity deep learning inversion.
Further, the depth-varying feature adaptive convolutional network includes:
the self-adaptive convolution aiming at the depth change characteristic of the apparent resistivity learns the amplitude and the offset of convolution kernels at different depth positions through network training, and then the self-adaptive convolution kernel of the depth change characteristic with the flexibility matched with the depth change characteristic is as follows:
Figure BDA0002442437240000081
wherein, α (h)k，β(h)k，A(h)kAdaptation parameters for the characteristics of the depth variation to be learned, α (h)kAnd β (h)kIs the horizontal and vertical position of network node k, A (h)kIs a convolution kernel wcThe amplitude value of (a).
For the convolution kernel with the size of k and the input with the vertical length of h, the total quantity of the additional parameters is assigned to be 3 multiplied by k multiplied by h.
Further, the resistivity deep learning inversion loss function carrying the vertical weight of the data item comprises:
and applying vertical weight of data items in the deep learning inversion loss function to carry out targeted training standard deployment on the depth change characteristics of the apparent resistivity data. The data item vertical weight dw is of the form:
Figure BDA0002442437240000082
wherein m isi，jIs a predicted value of the resistivity model location (i, j.) λ is a constant related to electrode setup size and current electrode location parameter β depends on the data type and the dimension of the problem.
The resistivity deep learning inversion loss function carrying the vertical weight of the data item is as follows:
Figure BDA0002442437240000083
wherein the inversion resistivity value is
Figure BDA0002442437240000084
Model value of resistivity is mi,j。
As a typical implementation manner, the present embodiment discloses a two-dimensional resistivity deep learning inversion method with adaptive data depth variation characteristics, as shown in fig. 1, including the following steps:
step one, constructing a massive typical electric model data set through finite element forward modeling.
The earth-electricity model is shown in fig. 2 and is formed by combining one or more regular or irregular high/low resistance abnormal bodies.
The size of the model of the embodiment is 6.3 mx 1.2m, the number of electrode points is 64, the electrode spacing is 0.1m, the size of the inversion grid is 0.05 mx 0.05m, the resistivity values of the low-resistance abnormal body are respectively 10 ohm meter and 30 ohm meter, the resistivity values of the high-resistance abnormal body are respectively 950 ohm meter and 1000 ohm meter, and the background resistivity is 500 ohm meter;
forward modeling apparent resistivity data of each geoelectric model and two groups of forward modeling data under a Wennan-Schrenberger observation device;
the database of this embodiment includes 29160 sets of two-dimensional apparent resistivity profile-resistivity model data pairs, wherein the ratio of the regular anomaly calculation examples 14580 set, the irregular anomaly calculation examples 14580 set, the test set, the verification set, and the training set is 1:1: 10.
And step two, building an apparent resistivity data depth change characteristic self-adaptive convolution neural network.
The data depth variation feature adaptive convolution of the embodiment is applied to a network architecture based on U-Net, the number of network layers is 5, an input channel is 1, the sizes of convolution kernels are 3 x 3, and a depth variation mode of data is captured by constructing two vertical adaptive convolution layers, so that distinguishable features for the convolution layers are generated.
And step three, adding a data item vertical weight for adjusting the network learning capability of the resistivity detection deep region in the loss function.
The formula for calculating the loss function used in this embodiment is:
Figure BDA0002442437240000091
wherein v is a data value term; alpha is a smoothing factor and takes a value of 0.2; λ is 8; beta is 1.
And step four, training a depth inversion network.
The main network parameters and hardware conditions in this embodiment are: the calculation is realized by using a single NVIDIATITANXp. A network is built based on a PyTorch platform, the batch processing quantity (batch size) of an SGD optimizer is 5, the learning rate (learning) is 0.1, the momentum (momentum) is 0.9, the weight attenuation (weight decay) is 1e-4, and the working frequency (epoch) of a learning algorithm in the whole training data set is 500.
And step five, inputting apparent resistivity data in the trained network to obtain a relatively accurate inversion result as shown in fig. 3. The method can accurately invert and obtain the position and the form of the target body, and has a good inversion effect even in an area with deeper abnormal body burial depth.
As will be appreciated by one skilled in the art, embodiments of the present disclosure may be provided as a method, system, or computer program product. Accordingly, the present disclosure may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present disclosure may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and so forth) having computer-usable program code embodied therein.
The present disclosure is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the disclosure. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor 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 processor 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.
The above description is only a preferred embodiment of the present disclosure and is not intended to limit the present disclosure, and various modifications and changes may be made to the present disclosure by those skilled in the art. Any modification, equivalent replacement, improvement and the like made within the spirit and principle of the present disclosure should be included in the protection scope of the present disclosure.
Although the present disclosure has been described with reference to specific embodiments, it should be understood that the scope of the present disclosure is not limited thereto, and those skilled in the art will appreciate that various modifications and changes can be made without departing from the spirit and scope of the present disclosure.

Claims (10)

1. A two-dimensional resistivity inversion method with self-adaptive data depth change characteristics is characterized by comprising the following steps: the method comprises the following steps:
constructing a dataset of apparent resistivity-resistivity model data pairs for different electrical models;
constructing a self-adaptive convolution network with the amplitude and the offset of convolution kernels of different horizon depths and the self-adaptive variable characteristic according to the resistivity depth change;
constructing a resistivity deep learning inversion loss function carrying the vertical weight of the data item;
training the adaptive convolution network controlled by the inversion loss function by using the data set, and establishing a mapping relation between apparent resistivity data and a resistivity model;
and inputting observation apparent resistivity data into the trained adaptive convolution network, and outputting a resistivity imaging result through the established mapping relation to realize surface two-dimensional resistivity deep learning inversion.
2. The method of claim 1 for adaptive two-dimensional resistivity inversion of depth-varying features of data, comprising: the geoelectricity model is a single or a plurality of combinations of high-resistance/low-resistance abnormal bodies with different forms.
3. The method of claim 1 for adaptive two-dimensional resistivity inversion of depth-varying features of data, comprising: the depth variation characteristic self-adaptive convolution network comprises the following components:
the self-adaptive convolution aiming at the depth change characteristic of the apparent resistivity learns the amplitude and the offset of convolution kernels at different depth positions through network training, and then the self-adaptive convolution kernel of the depth change characteristic with the flexibility matched with the depth change characteristic is as follows:
Figure FDA0002442437230000011
wherein, α (h)k，β(h)k，A(h)kAdaptation parameters for the characteristics of the depth variation to be learned, α (h)kAnd β (h)kIs the horizontal and vertical position of network node k, A (h)kIs a convolution kernel wcThe amplitude value of (a).
4. The method of claim 3, wherein the method comprises: for a convolution kernel of size k and an input of vertical length h, the total amount of additional parameters is assigned 3 × k × h.
5. The method of claim 1 for adaptive two-dimensional resistivity inversion of depth-varying features of data, comprising: the resistivity deep learning inversion loss function carrying the vertical weight of the data item comprises the following steps:
and applying vertical weight of data items in the deep learning inversion loss function to carry out targeted training standard deployment on the depth change characteristics of the apparent resistivity data.
6. The method of claim 5, wherein the method comprises: the data item vertical weight dw is of the form:
Figure FDA0002442437230000021
wherein m isi，jIs a predictor of the resistivity model location (i, j), λ is a constant related to the electrode setup size and current electrode location, and parameter β depends on the data type and the dimensionality of the problem.
7. The method of claim 6, wherein the method comprises: the resistivity deep learning inversion loss function carrying the vertical weight of the data item is as follows:
Figure FDA0002442437230000022
wherein the inversion resistivity value is
Figure FDA0002442437230000023
Model value of resistivity is mi,j。
8. A data depth change characteristic self-adaptive two-dimensional resistivity inversion system is characterized in that: the method comprises the following steps:
a dataset construction module configured to construct a dataset of apparent resistivity-resistivity model data pairs for different electrical models;
the network model building module is configured to build convolution kernel amplitudes and offsets of different horizon depths into an adaptive convolution network which is adaptively variable according to the resistivity depth change characteristics;
an inversion loss function construction module configured to construct an inversion loss function carrying a vertical weight of the resistivity data item;
the training module is configured to train the adaptive convolution network controlled by the inversion loss function by using the data set, and establish a mapping relation between apparent resistivity data and a resistivity model;
and the inversion module is configured to input the observation apparent resistivity data into the trained adaptive convolution network, output the resistivity imaging result through the established mapping relation and realize the surface two-dimensional resistivity deep learning inversion.
9. A computer-readable storage medium characterized by: a plurality of instructions stored therein, the instructions adapted to be loaded by a processor of a terminal device and to perform the steps of a method for data depth variation feature adaptive two-dimensional resistivity inversion according to any one of claims 1 to 7.
10. A terminal device is characterized in that: the system comprises a processor and a computer readable storage medium, wherein the processor is used for realizing instructions; a computer readable storage medium for storing a plurality of instructions adapted to be loaded by a processor and to perform the steps of a method for data depth variation feature adaptive two-dimensional resistivity inversion according to any one of claims 1 to 7.
CN202010269141.9A 2020-04-08 2020-04-08 Data depth change characteristic self-adaptive two-dimensional resistivity inversion method and system Active CN111597753B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010269141.9A CN111597753B (en) 2020-04-08 2020-04-08 Data depth change characteristic self-adaptive two-dimensional resistivity inversion method and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010269141.9A CN111597753B (en) 2020-04-08 2020-04-08 Data depth change characteristic self-adaptive two-dimensional resistivity inversion method and system

Publications (2)

Publication Number Publication Date
CN111597753A true CN111597753A (en) 2020-08-28
CN111597753B CN111597753B (en) 2023-06-30

Family

ID=72188685

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010269141.9A Active CN111597753B (en) 2020-04-08 2020-04-08 Data depth change characteristic self-adaptive two-dimensional resistivity inversion method and system

Country Status (1)

Country Link
CN (1) CN111597753B (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112199887A (en) * 2020-09-30 2021-01-08 山东大学 Semi-supervised resistivity inversion method and system based on countermeasure generation network and pseudo-labeling
CN112364911A (en) * 2020-11-06 2021-02-12 东北石油大学 Resistivity imaging inversion method and device and storage medium
CN115755201A (en) * 2022-11-21 2023-03-07 深蓝感知(杭州)物联科技有限公司 Method for calculating towed transient electromagnetic apparent resistivity based on neural network model

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017113205A1 (en) * 2015-12-30 2017-07-06 中国科学院深圳先进技术研究院 Rapid magnetic resonance imaging method and apparatus based on deep convolutional neural network
CN110333543A (en) * 2019-07-03 2019-10-15 山东大学 Post non of low resistance body explanation and imaging method and system based on reflection coefficient analysis
US20200065620A1 (en) * 2018-08-24 2020-02-27 Petrochina Company Limited Method and apparatus for automatically recognizing electrical imaging well logging facies
CN110888172A (en) * 2019-11-11 2020-03-17 吉林大学 Coarse medium electromagnetic response resistivity imaging method based on neural network

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017113205A1 (en) * 2015-12-30 2017-07-06 中国科学院深圳先进技术研究院 Rapid magnetic resonance imaging method and apparatus based on deep convolutional neural network
US20200065620A1 (en) * 2018-08-24 2020-02-27 Petrochina Company Limited Method and apparatus for automatically recognizing electrical imaging well logging facies
CN110333543A (en) * 2019-07-03 2019-10-15 山东大学 Post non of low resistance body explanation and imaging method and system based on reflection coefficient analysis
CN110888172A (en) * 2019-11-11 2020-03-17 吉林大学 Coarse medium electromagnetic response resistivity imaging method based on neural network

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
BIN LIU 等: "Full waveform inversion based on inequality constraint for cross-hole radar" *
刘斌等: "基于自适应加权光滑约束与PCG算法的三维电阻率探测反演成像" *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112199887A (en) * 2020-09-30 2021-01-08 山东大学 Semi-supervised resistivity inversion method and system based on countermeasure generation network and pseudo-labeling
CN112364911A (en) * 2020-11-06 2021-02-12 东北石油大学 Resistivity imaging inversion method and device and storage medium
CN115755201A (en) * 2022-11-21 2023-03-07 深蓝感知(杭州)物联科技有限公司 Method for calculating towed transient electromagnetic apparent resistivity based on neural network model

Also Published As

Publication number Publication date
CN111597753B (en) 2023-06-30

Similar Documents

Publication Publication Date Title
CN111597753A (en) Data depth change characteristic self-adaptive two-dimensional resistivity inversion method and system
CN108563837B (en) Method and system for correcting model parameters of alluvial river water sand model in real time
Durmuş et al. Parameter identification using particle swarm optimization
Jamali et al. Probability of failure for uncertain control systems using neural networks and multi-objective uniform-diversity genetic algorithms (MUGA)
Baumstein Extended subspace method for attenuation of crosstalk in multi-parameter full wavefield inversion
Aranha et al. Optimization of oil reservoir models using tuned evolutionary algorithms and adaptive differential evolution
EP2956912A1 (en) Gridless simulation of a fluvio-deltaic environment
CN107193043B (en) A kind of subsurface structure imaging method of relief surface
CN111158059B (en) Gravity inversion method based on cubic B spline function
CN111291316B (en) Multi-scale resistivity inversion method and system based on wavelet transformation
Lee et al. Field application study on automatic history matching using particle swarm optimization
CN112199887A (en) Semi-supervised resistivity inversion method and system based on countermeasure generation network and pseudo-labeling
CN102830430B (en) A kind of horizon velocity modeling method
CN116577821A (en) Time-lapse earthquake prediction method and system based on neural network
Tang et al. Fundamental matrix estimation by multiobjective genetic algorithm with Taguchi's method
CN112016956B (en) Ore grade estimation method and device based on BP neural network
CN116011338A (en) Full waveform inversion method based on self-encoder and deep neural network
CN113642232B (en) Intelligent inversion exploration method for surface waves, storage medium and terminal equipment
US8942966B2 (en) Method for parameterizing and morphing stochastic reservoir models
CN115659773A (en) Full waveform inversion acceleration method based on depth network and related device
CN117010260A (en) Automatic history fit model prediction method, system and equipment for fractured reservoir
US9488047B2 (en) Reservoir calibration parameterization method
CN111597752A (en) Cross-hole resistivity CT deep learning inversion method and system for balancing inter-hole sensitivity
CN111126439A (en) Method for generating coal mine underground two-dimensional mine earthquake imaging training data set
Jo et al. Conditioning stratigraphic, rule-Based models with generative adversarial networks: A deepwater lobe, deep learning example

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