CN104422969B - A kind of method for reducing electromagnetic sounding inversion result nonuniqueness - Google Patents
A kind of method for reducing electromagnetic sounding inversion result nonuniqueness Download PDFInfo
- Publication number
- CN104422969B CN104422969B CN201310397390.6A CN201310397390A CN104422969B CN 104422969 B CN104422969 B CN 104422969B CN 201310397390 A CN201310397390 A CN 201310397390A CN 104422969 B CN104422969 B CN 104422969B
- Authority
- CN
- China
- Prior art keywords
- depth
- resistivity
- point
- inversion
- regularization
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Landscapes
- Geophysics And Detection Of Objects (AREA)
Abstract
The present invention is a kind of method of the nonuniqueness for reducing electromagnetic sounding inversion result, collection exploratory area electromagnetic sounding data, obtain at least more than 2 two dimensional inversion depth resistivity models, regularization is processed and obtains regularization two-dimensional depth resistivity model inversion result, determine learning sample and target sample, set up BP Neural Networks Systems, work area interior profile pointwise is sequentially input neuroid by depth point, predicting the outcome for each each depth point of measuring point desired is obtained, and is restored and the two dimensional inversion result for optimizing is obtained for real inverting resistivity value.The present invention is a kind of processing method of the inverting nonuniqueness rapid Optimum for electromagnetic sounding data, by the Processing with Neural Network to multiple inversion results, by the study to Given information, nonuniqueness is reduced, to obtaining the true model practicability and effectiveness of complex area objective body.
Description
Technical field
The present invention relates to geophysical exploration method, is electromagnetic survey data treatment technology, specifically a kind of reduction electromagnetism
The method of the nonuniqueness of depth measurement inversion result.
Background technology
Electromagnetic sounding is a kind of change feelings changed to understand underground medium resistivity by ground observation electromagnetic field
The geophysical exploration method of condition.In recent years, theoretical research and computer technology are fast-developing, by the frequency-resistivity for gathering
Data are converted to the inversion method of depth-resistivity data and emerge in an endless stream, and two-dimensional layered model inverting, two-dimentional continuous media are anti-
Drill, or even 3-d inversion method etc. is also progressively grown up, calculating speed is greatly improved, can be than faster obtaining underground ground electricity
Model, makes electromagnetic sounding method in seismic prospecting difficulty area, such as mountain front, igneous rock exposure area, top layer gravel area, carbonate rock
Area, the loess tableland area of coverage achieve preferable exploration effects, compensate for the deficiency of seismic prospecting data.
But, either two dimensional inversion or 3-d inversion, even if error of fitting very little, the inversion result of acquisition is often
Not exclusively it is consistent with electrical structure is practically descended, traces it to its cause or caused due to the nonuniqueness of Geophysics Inversion, i.e.,
Geophysical survey data has multiple models corresponding therewith, and the nonuniqueness of electromagnetic sounding inverting cannot even more be avoided.
The nonuniqueness for how reducing inversion result is always the direction that geophysicist makes great efforts, wherein, most study
Be inversion algorithm optimization, but result can not be entirely satisfactory.This is always electromagnetic sounding inverting achievement and is difficult to meet
The reason for needs of production.
Content of the invention
It is to provide a kind of acquisition be close to the true model of complex area objective body, reduce electromagnetic sounding inverting mesh of the present invention
As a result the method for nonuniqueness.
The present invention is realized by following steps:
1)Collection work area electromagnetic sounding data, does different inversion procedures to every survey line, obtains at least more than 2 two dimensions anti-
Drill depth-resistivity models;
Described different inversion procedures are using two-dimentional Aukma(Occam)Inverting or two-dimentional fast relaxation(RRI)Inverting
Or two-dimentional conjugate gradient inversion or two dimension regard mould inverting or two dimension regularization inverting or Two-Dimensional Generalized anti-inversion or two-dimensional analog is moved back
Fiery inverting.
2)To step 1)Two dimensional inversion depth-resistivity models carry out regularization process, obtain regularization two-dimensional depth-electricity
Resistance rate model inversion result;
It is to all steps 1 that described regularization is processed)Two dimensional inversion depth-the resistivity models for obtaining are by ground to ground
In lower depth bounds, according to identical depth interval, profile intervals, interpolation processing is carried out, find out the minimum of all resistivity values
Value(RHOmin)With maximum (RHOmax), to each resistivity value(RHOij)According to formula(RHOij-RHOmin)/
(RHOmax-RHOmin)Carry out regularization.
Described subterranean depth can be adjusted depending on exploration targets body buried depth situation, be 5-20km;Most preferably 10km.
Described depth interval is 50-200m;Most preferably 100m.
Described section measuring point is at intervals of 50-200m;Most preferably 100m.
3)Determine learning sample and target sample;
Inversion result after using regularization has the position of drilling well or seismic profile data as given data point, in known money
1-3 point is chosen in shots environs as learning sample;
Described learning sample given data point environs are 100-300m.
1-3 point of described learning sample is from step 1)Different inverse models.
If well logging resistivity curve depth is less than 10km, insufficient section step 1)The mean value of different inverse models
Replace, according to step 2)Regularization process, according to step 2)Depth bounds, step 2)Depth interval enters row interpolation;Obtain
Minimum of a value RHOmin of well logging resistivity curve, maximum RHOmax, by each resistivity value RHOij according to step 2)Formula
Regularization is carried out, the target sample of given data point is obtained.
If without drilling data, the seismic profile after selecting to explain sets up strata division model, then according to exploratory area thing
Property analysis obtain formation resistivity master data, set up layering resistivity models, on seismic profile from electromagnetic sounding section
The resistivity hierarchical mode of nearest point or the point intersected with electromagnetic sounding as peg model, according to above interpolation and regularization
Process carries out regularization, obtains the target sample of given data point.
4)Set up BP Neural Networks Systems;
Described BP Neural Networks Systems of setting up are:The data of each each depth point of learning sample are sequentially inputted to
Neuroid, is accordingly input into target sample, starts BP neuroids and is trained, if error of fitting has reached setting
Error, or reached maximum iterations then terminate training, just establish BP Neural Networks Systems.
Described BP Neural Networks Systems, different two dimensional inversion depth-resistivity models that different inversion methods are obtained,
Pattern number is input layer number;Hidden nodes are 10-15, and output layer neuron number is 1.
The data of each described depth point are steps 1)The data of different inverse models.
The described error for setting is as 0.001.
The iterations of described maximum is 1000 times.
5)By a section or a plurality of section in work area, pointwise is by depth point successively according to step 2)After regularization is processed
Data, are sequentially inputted to by step 4)The neuroid of foundation, obtains the prediction knot of each each depth point of measuring point desired
Really(Kij);According to step 3)RHOmin the and RHOmax values of determination, are reduced to real inverting resistivity value, i.e. RHOij=
RHOmin+Kij*(RHOmax-RHOmin), finally give the two dimensional inversion result of optimization.
The present invention is a kind of processing method of the inverting nonuniqueness rapid Optimum for electromagnetic sounding data, by many
The Processing with Neural Network of inversion result is planted, by the study to Given information, nonuniqueness is reduced, is to obtain complex area objective body
True model provide very useful effective method.
Description of the drawings
Fig. 1 is step schematic diagram of the present invention.
Specific embodiment
Below in conjunction with accompanying drawing and the example in detail present invention.
The concrete grammar step of the model optimization method of many inversion results is as follows:
1)Collection exploratory area electromagnetic sounding data, does following 5 kinds of invertings to every survey line:1. two-dimentional Aukma(Occam)Instead
Drill, 2. two-dimentional fast relaxation(RRI)Inverting, 3. two-dimentional conjugate gradient inversion, 4. two dimension regard mould inverting, 5. Two-Dimensional Generalized is converse
Drill, obtain at least 5 two dimensional inversion depth-resistivity models;
2)To step 1)5 kinds of two dimensional inversion depth-resistivity models carry out regularization process, obtain 5 kinds of regularization
Two-dimensional depth-resistivity model inversion result;
It is to all steps 1 that described regularization is processed)The 5 kinds of two dimensional inversion depth-resistivity models for obtaining are by ground
To in the range of subterranean depth, according to identical depth interval, profile intervals, interpolation processing is carried out, find out all resistivity values
Minimum of a value(RHOmin)With maximum (RHOmax), to each resistivity value(RHOij)According to formula(RHOij-RHOmin)/
(RHOmax-RHOmin)Carry out regularization.
Described subterranean depth is chosen as 10km according to exploration targets body buried depth situation.
Described depth interval is 100m.
Described section measuring point is at intervals of 100m.
3)Determine learning sample and target sample;
Inversion result after using regularization has the position of drilling well or seismic profile data as given data point, in known money
3 points are chosen in shots environs as learning sample;
Described learning sample given data point environs are 100-300m.
3 points of described learning sample are from step 1)Different inverse models.
If well logging resistivity curve depth is less than 10km, insufficient section step 1)The mean value of different inverse models
Replace, according to step 2)Regularization process, according to step 2)Depth bounds, step 2)Depth interval enters row interpolation;Obtain
Minimum of a value RHOmin of well logging resistivity curve, maximum RHOmax, by each resistivity value RHOij according to step 2)Formula
Regularization is carried out, the target sample of given data point is obtained.
If without drilling data, the seismic profile after selecting to explain sets up strata division model, then according to exploratory area thing
Property analysis obtain formation resistivity master data, set up layering resistivity models, on seismic profile from electromagnetic sounding section
The resistivity hierarchical mode of nearest point or the point intersected with electromagnetic sounding as peg model, according to above interpolation and regularization
Process carries out regularization, obtains the target sample of given data point.
4)Set up BP Neural Networks Systems;
Described BP Neural Networks Systems of setting up are:The data of each each depth point of learning sample are sequentially inputted to
Neuroid, is accordingly input into target sample, starts BP neuroids and is trained, if error of fitting has reached setting
Error, or reached maximum iterations then terminate training, just establish BP Neural Networks Systems.
Described BP Neural Networks Systems, different two dimensional inversion depth-resistivity moulds that 5 kinds of different inversion methods are obtained
Type, pattern number are input layer number;Hidden nodes are 15, and output layer neuron number is 1.
The data of each described depth point are steps 1)The data of different inverse models.
The described error for setting is as 0.001.
The iterations of described maximum is 1000 times.
Described BP(Back Propagation)Network is 1986 by the section headed by Rumelhart and McCelland
Scholar group proposes, and is a kind of Multi-layered Feedforward Networks that trains by Back Propagation Algorithm, is most widely used nerve at present
One of network model.BP networks can learn and store substantial amounts of input-output mode map relation, and without the need for disclosing description in advance
The math equation of this mapping relations.Its learning rules are to use steepest descent method, constantly adjust net by backpropagation
The weights of network and threshold value, make the error sum of squares of network minimum.BP neural network model topology structure includes input layer
(input), hidden layer (hide layer) and output layer (output layer).There is neuron storehouse letter in Matblab built-in functions
Number, directly can quote.
5)By a section or a plurality of section in work area, pointwise is by depth point successively according to step 2)After regularization is processed
Data, are sequentially inputted to by step 4)The neuroid of foundation, obtains the prediction knot of each each depth point of measuring point desired
Really(Kij);According to step 3)RHOmin the and RHOmax values of determination, are reduced to real inverting resistivity value, i.e. RHOij=
RHOmin+Kij*(RHOmax-RHOmin), finally give the two dimensional inversion result of optimization.
Claims (13)
1. a kind of reduce electromagnetic sounding inversion result nonuniqueness method, feature through the following steps that realize:
1) gather work area electromagnetic sounding data, do different inversion procedures to every survey line, obtain more than 2 two dimensional inversion depth-
Resistivity models;
2) to step 1) two dimensional inversion depth-resistivity models carry out regularization process, obtain regularization two-dimensional depth-resistivity
Model inversion result;
It is to all steps 1 that described regularization is processed) two dimensional inversion depth-resistivity models for obtaining by ground to underground depth
In the range of degree, it is spaced according to identical depth interval, section measuring point, carries out interpolation processing, find out the minimum of all resistivity values
Value RHOmin and maximum RHOmax, carry out regularization to each resistivity value RHOij according to below equation:(RHOij-
RHOmin)/(RHOmax-RHOmin);
3) learning sample and target sample are determined;
There is the position of drilling well or seismic profile data in inversion result after using regularization as given data point, in given data
1-3 point is chosen in point environs as learning sample;
4) BP Neural Networks Systems are set up;
Described BP Neural Networks Systems of setting up are:The data of each each depth point of learning sample are sequentially inputted to nerve
Metanetwork, is accordingly input into target sample, starts BP neuroids and is trained, if error of fitting has reached the mistake for setting
Differ from, or reached maximum iterations and then terminate training, just establish BP Neural Networks Systems;
5) by a section or a plurality of section in work area, by measuring point by depth point according to step 2) regularization process after data,
It is sequentially inputted to by step 4) the BP Neural Networks Systems set up, obtain the prediction knot of each each depth point of measuring point desired
Fruit Kij;According to step 2) RHOmin the and RHOmax values that determine, real inverting resistivity value is reduced to, optimization is finally given
Two dimensional inversion result,
Step 1) described in different inversion procedures be using two-dimentional Aukma inverting or two-dimentional fast relaxation inverting or two dimension conjugation
Gradient inverting or two dimension regard mould inverting or two-dimentional regularization inverting or Two-Dimensional Generalized anti-inversion or two-dimensional analog annealing inverting,
The target sample is determined using following methods:
Step 2) resistivity curve depth less than 10km when, insufficient section step 1) different inverse model mean value generation
Replace, according to step 2) regularization process, according to step 2) depth bounds and step 2) depth interval enters row interpolation;Obtain survey
Minimum of a value RHOmin and maximum RHOmax of well resistivity curve, by each resistivity value RHOij according to step 2) formula
Regularization is carried out, the target sample of given data point is obtained;
Step 3) drilling data if it did not, select explain after seismic profile set up strata division model, then according to work area
Physical Property Analysis obtain formation resistivity master data, set up the resistivity models of layering, cuing open from electromagnetic sounding on seismic profile
The resistivity hierarchical mode of the nearest point in face or the point intersected with electromagnetic sounding section as peg model, according to above interpolation and
Regularization is processed and carries out regularization, obtains the target sample of given data point.
2. method according to claim 1, feature is step 2) described in subterranean depth be 5-20km.
3. method according to claim 2, feature is step 2) described in subterranean depth be 10km.
4. method according to claim 1, feature is step 2) described in depth interval be 50-200m.
5. method according to claim 4, feature is step 2) described in depth interval be 100m.
6. method according to claim 1, feature is step 2) described in section measuring point at intervals of 50-200m.
7. method according to claim 6, feature is step 2) described in section measuring point at intervals of 100m.
8. method according to claim 1, feature is step 3) described in given data point environs be 100-300m.
9. method according to claim 1, feature is step 3) described in learning sample 1-3 point from step 1) different anti-
Drill model.
10. method according to claim 1, feature is step 4) described in BP Neural Networks Systems obtained with different inversion methods
Data after the different two dimensional inversion depth for arriving-resistivity models regularization is processed are input, and wherein pattern number is input layer god
Through first number;Hidden nodes are 10-15, and output layer neuron number is 1.
11. methods according to claim 1, feature is step 4) described in the data of each depth point be step 1) different invertings
The data of model, the data are step 1) in two dimensional inversion depth-resistivity for obtaining of different inversion procedures.
12. methods according to claim 1, feature is step 4) described in the error for setting as 0.001.
13. methods according to claim 1, feature is step 4) described in the iterations of maximum be 1000 times.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310397390.6A CN104422969B (en) | 2013-09-04 | 2013-09-04 | A kind of method for reducing electromagnetic sounding inversion result nonuniqueness |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310397390.6A CN104422969B (en) | 2013-09-04 | 2013-09-04 | A kind of method for reducing electromagnetic sounding inversion result nonuniqueness |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104422969A CN104422969A (en) | 2015-03-18 |
CN104422969B true CN104422969B (en) | 2017-03-15 |
Family
ID=52972514
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201310397390.6A Active CN104422969B (en) | 2013-09-04 | 2013-09-04 | A kind of method for reducing electromagnetic sounding inversion result nonuniqueness |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104422969B (en) |
Families Citing this family (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104777516A (en) * | 2015-04-15 | 2015-07-15 | 国网重庆市电力公司电力科学研究院 | Apparent resistivity calculating method on basis of non-linear equation solution modular form |
CN106842344B (en) * | 2017-04-24 | 2018-10-12 | 中国科学院电子学研究所 | The method of unmanned plane boat magnetic holoaxial gradient magnetic disturbance compensation based on feedforward network |
CN109188536B (en) * | 2018-09-20 | 2020-06-05 | 成都理工大学 | Time-frequency electromagnetic and magnetotelluric joint inversion method based on deep learning |
CN109343134A (en) * | 2018-11-27 | 2019-02-15 | 中煤科工集团西安研究院有限公司 | A kind of Transient Electromagnetic Method in Mine data analysis interpretation method and system |
CN109557601A (en) * | 2019-01-22 | 2019-04-02 | 青岛海洋地质研究所 | Reservoir parameter inversion method is combined in one-dimensional ocean controllable source electromagnetism and earthquake |
CN111126591B (en) * | 2019-10-11 | 2023-04-18 | 重庆大学 | Magnetotelluric deep neural network inversion method based on space constraint technology |
CN112699596B (en) * | 2020-12-04 | 2022-08-23 | 湖南工商大学 | Wide-area electromagnetic method induced polarization information nonlinear extraction method based on learning |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101008678A (en) * | 2006-01-26 | 2007-08-01 | 中国石油天然气集团公司 | 3D resistivity cascaded automatic imaging method |
CA2659362A1 (en) * | 2006-08-22 | 2008-02-28 | Kjt Enterprises, Inc. | Fast 3d inversion of electromagnetic survey data using a trained neural network in the forward modeling branch |
CN101650443A (en) * | 2009-09-15 | 2010-02-17 | 重庆大学 | Back-propagation network calculating method of apparent resistivity |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2010151354A1 (en) * | 2009-06-26 | 2010-12-29 | Exxonmobil Upstream Research Company | Constructing resistivity models from stochastic inversion |
-
2013
- 2013-09-04 CN CN201310397390.6A patent/CN104422969B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101008678A (en) * | 2006-01-26 | 2007-08-01 | 中国石油天然气集团公司 | 3D resistivity cascaded automatic imaging method |
CA2659362A1 (en) * | 2006-08-22 | 2008-02-28 | Kjt Enterprises, Inc. | Fast 3d inversion of electromagnetic survey data using a trained neural network in the forward modeling branch |
CN101650443A (en) * | 2009-09-15 | 2010-02-17 | 重庆大学 | Back-propagation network calculating method of apparent resistivity |
Non-Patent Citations (4)
Title |
---|
Artificial Neural Networks in the Solution of Inverse Electromagnetic Field Problems;S.Ratnajeevan H. Hoole;《IEEE TRANSACTIONS ON MAGNETICS》;19930331;第29卷(第2期);第1931-1934页 * |
Neural network modeling and prediction of resistivity structures using VES Schlumberger data over a geothermal area;Upendra K. Singh,et al.;《Computers & Geosciences》;20121022;第246–257页 * |
基于BP网络的视电阻率快速计算;谢品芳等;《重庆大学学报》;20090930;第32卷(第9期);第1112-1116页 * |
瞬变电磁勘探中的人工神经网络反演法;李创社等;《西安交通大学学报》;20010630;第35卷(第6期);第604-607页及第615页 * |
Also Published As
Publication number | Publication date |
---|---|
CN104422969A (en) | 2015-03-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104422969B (en) | A kind of method for reducing electromagnetic sounding inversion result nonuniqueness | |
CN110968826B (en) | Magnetotelluric deep neural network inversion method based on spatial mapping technology | |
CN111126591B (en) | Magnetotelluric deep neural network inversion method based on space constraint technology | |
EP2069826B1 (en) | Fast 3d inversion of electromagnetic survey data using a trained neural network in the forward modeling branch | |
CN104614763B (en) | Multi-wave AVO reservoir elastic parameter inversion method and system based on reflectivity method | |
CN105651676B (en) | A kind of reservoir heterogeneity characterizing method under horizontal well rule well pattern | |
CN107589448A (en) | A kind of multitrace seismogram reflection coefficient sequence Simultaneous Inversion method | |
Wang et al. | On a new method of estimating shear wave velocity from conventional well logs | |
CN113610945B (en) | Ground stress curve prediction method based on hybrid neural network | |
CN102645670B (en) | Observation system optimization design method based on stack response analysis | |
Jha et al. | Vertical electrical sounding survey and resistivity inversion using genetic algorithm optimization technique | |
Shi et al. | Brittleness index prediction in shale gas reservoirs based on efficient network models | |
CN103454677B (en) | Based on the earthquake data inversion method that population is combined with linear adder device | |
Sarkheil et al. | Fractures distribution modeling using fractal and multi-fractal–neural network analysis in Tabnak hydrocarbon field, Fars, Iran | |
Ouadfeul et al. | Lithofacies classification using the multilayer perceptron and the self-organizing neural networks | |
Liu et al. | An advanced prediction model of shale oil production profile based on source-reservoir assemblages and artificial neural networks | |
CN105005079A (en) | Well logging curve inversion method | |
CN104297792B (en) | The phased inversion method of water channel reservoir is stacked on a kind of fan | |
Rastegarnia et al. | A comparative study of 3D FZI and electrofacies modeling using seismic attribute analysis and neural network technique: A case study of Cheshmeh-Khosh Oil field in Iran | |
CN106338764B (en) | The super wire stripping recognition methods of bio-inspired computing stratigraphic trap oil-gas reservoir | |
Itani et al. | Three-dimensional modeling of spatial soil properties via artificial neural networks | |
CN110989034B (en) | Method for inverting logging transverse wave time difference by regression-fractal interpolation method | |
CN114482995A (en) | Fine determination method for argillaceous content of fine-grain sediment | |
Pradhan et al. | Seismic inversion for reservoir facies under geologically realistic prior uncertainty with 3D convolutional neural networks | |
CN111812729A (en) | Multi-dimensional geological background model construction method for optimizing electrical prospecting inversion |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant |