CN112464483B - Logging curve reconstruction method based on genetic neural network algorithm - Google Patents

Logging curve reconstruction method based on genetic neural network algorithm Download PDF

Info

Publication number
CN112464483B
CN112464483B CN202011407762.5A CN202011407762A CN112464483B CN 112464483 B CN112464483 B CN 112464483B CN 202011407762 A CN202011407762 A CN 202011407762A CN 112464483 B CN112464483 B CN 112464483B
Authority
CN
China
Prior art keywords
neural network
genetic
logging curve
training
algorithm
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
Application number
CN202011407762.5A
Other languages
Chinese (zh)
Other versions
CN112464483A (en
Inventor
张强
李家金
王毛毛
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.)
Nuke Industry No216 Brigade
Original Assignee
Nuke Industry No216 Brigade
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 Nuke Industry No216 Brigade filed Critical Nuke Industry No216 Brigade
Priority to CN202011407762.5A priority Critical patent/CN112464483B/en
Publication of CN112464483A publication Critical patent/CN112464483A/en
Application granted granted Critical
Publication of CN112464483B publication Critical patent/CN112464483B/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
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01D—MEASURING NOT SPECIALLY ADAPTED FOR A SPECIFIC VARIABLE; ARRANGEMENTS FOR MEASURING TWO OR MORE VARIABLES NOT COVERED IN A SINGLE OTHER SUBCLASS; TARIFF METERING APPARATUS; MEASURING OR TESTING NOT OTHERWISE PROVIDED FOR
    • G01D21/00—Measuring or testing not otherwise provided for
    • G01D21/02—Measuring two or more variables by means not covered by a single other subclass
    • 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
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • 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
    • G06N3/086—Learning methods using evolutionary algorithms, e.g. genetic algorithms or genetic programming

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Artificial Intelligence (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Computer Hardware Design (AREA)
  • Geometry (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Physiology (AREA)
  • Geophysics And Detection Of Objects (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention relates to the field of geophysical data processing, in particular to a logging curve reconstruction method based on a genetic neural network algorithm, which mainly comprises the following steps: the first step is as follows: and (6) well logging curve standardization. The conventional well logging curves are standardized to be unified to the same dimension level. The second step: a neural network structure is established and the network is trained. And determining the input, output and network layer number of the neural network, establishing a neural network structure, and training the network. The third step: and (4) genetic manipulation. And calculating a training error and fitness function, and optimizing the network structure and the weight threshold by using three genetic operators of selection, intersection and variation. The fourth step: and reconstructing the logging curve by using the optimized network structure until the accuracy requirement is met and outputting a result. Compared with the traditional well logging curve reconstruction method, the method has higher precision, can reduce the production cost, improves the operation efficiency and improves the well logging curve reconstruction effect.

Description

Logging curve reconstruction method based on genetic neural network algorithm
Technical Field
The invention belongs to the technical field of geophysical data processing, and particularly relates to a logging curve reconstruction method based on a genetic neural network algorithm.
Background
In the field of mineral exploration, there is a need to measure multiple different types of well logs for reducing the multi-interpretation of geological interpretation.
However, in practice, it is often the case that the well logs are distorted or missing, for example, by an enlarged borehole diameter or stuck instruments, so that some of the well logs are distorted or missing.
Early boreholes may lack some important well logs, such as sonic logging, due to the incomplete logging methods, which can present difficulties in geological research.
Therefore, how to accurately reconstruct the missing or distorted well log data is a problem worth discussing. For the reconstruction problem of the logging curve, the currently more common method is the traditional BP neural network algorithm, i.e. the internal correlation between known curves is obtained through the learning training of the network, and then the reconstruction of an unknown curve is completed (Luisa Rolon et al, 2009).
However, the traditional BP neural network has its own disadvantages, such as that the optimization is easy to fall into a local minimum value, and the like, which results in poor reconstruction accuracy of a logging curve (zheng qing sheng and han da ge, 2007).
Disclosure of Invention
In view of the above disadvantages, the present invention aims to provide a method for reconstructing a logging curve based on a genetic neural network algorithm, which can achieve a better effect of reconstructing the logging curve, thereby providing more accurate curve parameters for subsequent geological interpretation.
The technical scheme of the invention is as follows:
a logging curve reconstruction method based on a genetic neural network algorithm comprises the following steps:
step (1), well logging curve standardization: because the unit and the magnitude of the logging curves of different types are different, when the method is used for a neural network, the data needs to be standardized in the following way:
Figure BDA0002817275960000021
wherein x is i A well logging curve is formed; x max And X min The maximum value and the minimum value of the logging curve are obtained; x i A normalized logging curve;
step (2), establishing a neural network structure and training a network: taking the number of the well logging curves needing to be trained as the input of a neural network, taking the well logging curves needing to be reconstructed as the output of the neural network, setting the number of the hidden layers as 1 layer, establishing a three-layer neural network structure of an input layer, the hidden layers and an output layer, and training the network;
step (3), fitness function determination: firstly, calculating the error between a prediction result and a training target by using a trained neural network:
||error||＝(|Y 1 -T 1 | 2 +|Y 2 -T 2 | 2 +…+|Y n -T n | 2 ) 1/2
wherein: y is a BP operator prediction result; t is an actual training target value; n is the number of training samples;
then, calculating the fitness value of the population individuals according to the training errors, namely the larger the error between the prediction result and the actual target is, the lower the individual fitness is, and taking the reciprocal of the training error as a function for calculating the fitness:
fitness＝1/error
step (4), genetic manipulation: establishing a connection with the neural network through a fitness function according to three genetic operators of selection, intersection and variation to realize the optimization of the weight and the threshold of the neural network;
step (5), reconstructing a curve: and (5) repeating the steps (3) and (4) until the boundary condition is met, completing the optimization of the neural network, and carrying out the reconstruction of the logging curve on the optimized neural network structure and parameters until the accuracy requirement is met and outputting the reconstruction result of the logging curve.
And (3) optimizing the initial network by the training network in the step (2) through a BP operator in a neural network toolbox in MATLAB software.
And (4) the genetic algorithm coding mode in the step (4) adopts binary gene coding.
And (4) selecting, crossing and mutating three genetic operators in the step (4) by using a genetic algorithm toolbox in MATLAB software.
The mutation operator realizes gene mutation processing by setting mutation probability.
The crossover operator randomly sets crossover points in the gene string, and exchanges individuals before and after the crossover points to form a new gene string.
The selection operator adopts a roulette selection method.
The invention has the beneficial effects that:
compared with the prior art, the method for extracting the weak information of the sandstone-type uranium deposit deep penetration geochemical ionization has the advantages that the genetic algorithm is introduced to optimize the neural network structure and the weight threshold value, automatic calculation is realized, the problems that redundant calculation and the traditional neural network method are easy to fall into the local minimum value and the like are solved, and the method has the advantages of simplicity in operation, high operation efficiency, low cost, high reconstruction precision and the like.
Drawings
FIG. 1 is a comparison graph of the reconstruction effect of an acoustic curve;
FIG. 2 is a flow chart of the method.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be obtained by a person skilled in the art without making any creative effort based on the embodiments in the present invention, belong to the protection scope of the present invention.
A logging curve reconstruction method based on a genetic neural network algorithm comprises the following steps:
step (1), well logging curve standardization: because the unit and the magnitude of the logging curves of different types are different, when the method is used for a neural network, the data needs to be standardized in the following way:
Figure BDA0002817275960000041
wherein x is i A well logging curve is formed; x max And X min The maximum value and the minimum value of the logging curve are obtained; x i Normalized well log.
Step (2), establishing a neural network structure and training a network: and taking the number of the well logging curves needing to be trained as the input of a neural network, taking the well logging curves needing to be reconstructed as the output of the neural network, setting the number of the hidden layers as 1 layer, establishing a three-layer neural network structure of an input layer, the hidden layers and an output layer, and training the network.
Step (3), fitness function determination: firstly, calculating the error between a prediction result and a training target by using a trained neural network:
||error||＝(|Y 1 -T 1 | 2 +|Y 2 -T 2 | 2 +…+|Y n -T n | 2 ) 1/2
wherein: y is a BP operator prediction result; t is an actual training target value; n is the number of training samples;
then, calculating the fitness value of the population individuals according to the training errors, namely the larger the error between the prediction result and the actual target, the lower the individual fitness, and taking the reciprocal of the training errors as a function for calculating the fitness:
fitness＝1/error
step (4), genetic manipulation: and establishing a connection with the neural network through a fitness function according to the selection, the intersection and the variation of the three genetic operators to realize the optimization of the weight and the threshold of the neural network.
Step (5), reconstructing a curve: and (5) repeating the steps (3) and (4) until the boundary condition is met, completing the optimization of the neural network, and carrying out the reconstruction of the logging curve on the optimized neural network structure and parameters until the accuracy requirement is met and outputting the reconstruction result of the logging curve.
Preferably, the training network in the step (2) is implemented by optimizing the initial network through a BP operator in a neural network toolbox in MATLAB software.
Preferably, the genetic algorithm coding mode in the step (4) adopts binary gene coding.
Preferably, the three genetic operators of selection, intersection and mutation in the step (4) are implemented by a genetic algorithm tool box in MATLAB software. Wherein the mutation operator realizes gene mutation processing by setting mutation probability. The crossover operator randomly sets crossover points in the gene string, and exchanges individuals before and after the crossover points to form a new gene string. The selection operator uses roulette selection.
Example 1:
(1) And (3) well logging curve standardization: firstly, neural network training data including natural gamma, resistivity, density, well diameter and conventional logging data of acoustic logging curves are prepared. Because different logging curve parameters are different and have different representative meanings, normalization processing is carried out on the logging curve parameters, and the logging curve parameters are unified to a 0-1 change interval range.
(2) Establishing a neural network structure and training a network: this time, four curves of natural gamma, resistivity, density and well diameter are used as input, and a single curve of acoustic logging is used as output. The number of hidden layer nodes is usually set to be twice of the number of input curves, and because the neural network structure is optimized by using a genetic algorithm at this time, redundant calculation is reduced, the number of hidden layer initial nodes at this time is set to be three times of the number of input curves, namely 12, and the neural network structure of a reconstructed curve is a three-layer structure of 4-12-1. The number of neural network iterations is set to 8000, the learning rate is 0.05, the learning target is 0.01, and the minimum gradient of descent is 0.01.
(3) Determination of fitness function: and the fitness function determination is realized by adopting an MATLAB neural network and a genetic algorithm toolbox.
(4) Genetic manipulation: and realizing selection, intersection and mutation three genetic operator operations through a genetic algorithm toolbox in MATLAB software. The genetic algebra is set as 100 generations, the initial population is set as 50, the crossover probability is set as 0.07, and the coefficient of variation is set as 0.01.
(5) And (3) curve reconstruction: and (4) reconstructing the acoustic logging curve by using the optimized neural network parameters, and outputting an acoustic curve reconstruction result through iterative calculation until the accuracy requirement is met.
In the attached drawings of the disclosed embodiment of the invention, only methods related to the disclosed embodiment are related, other methods can refer to common design, and the same embodiment and different embodiments of the invention can be combined mutually under the condition of no conflict;
the above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the present invention, and any modifications, equivalents, improvements and the like made within the spirit and principle of the present invention should be included in the scope of the present invention.

Claims (7)

1. A logging curve reconstruction method based on a genetic neural network algorithm is characterized by comprising the following steps:
step (1), well logging curve standardization: because the unit and the magnitude of the logging curves of different types are different, when the method is used for a neural network, the data needs to be standardized in the following way:
Figure FDA0002817275950000011
wherein x is i Is a logging curve; x max And X min The maximum value and the minimum value of the logging curve are obtained; x i A normalized logging curve;
step (2), establishing a neural network structure and training a network: taking the number of the well logging curves needing to be trained as the input of a neural network, taking the well logging curves needing to be reconstructed as the output of the neural network, setting the number of the hidden layers as 1 layer, establishing a three-layer neural network structure of an input layer, the hidden layers and an output layer, and training the network;
step (3), fitness function determination: firstly, calculating the error between a prediction result and a training target by using a trained neural network:
||error||＝(|Y 1 -T 1 | 2 +|Y 2 -T 2 | 2 +…+|Y n -T n | 2 ) 1/2
wherein: y is a BP operator prediction result; t is an actual training target value; n is the number of training samples;
then, calculating the fitness value of the population individuals according to the training errors, namely the larger the error between the prediction result and the actual target, the lower the individual fitness, and taking the reciprocal of the training errors as a function for calculating the fitness:
fitness＝1/error
step (4), genetic manipulation: establishing connection with the neural network through a fitness function according to three genetic operators of selection, intersection and variation to realize optimization of the weight and the threshold of the neural network;
step (5), curve reconstruction: and (5) repeating the steps (3) and (4) until the boundary condition is met, completing the optimization of the neural network, and carrying out the reconstruction of the logging curve on the optimized neural network structure and parameters until the accuracy requirement is met and outputting the reconstruction result of the logging curve.
2. The method for reconstructing a well log based on a genetic neural network algorithm as claimed in claim 1, wherein: and (3) optimizing the initial network by the training network in the step (2) through a BP operator in a neural network toolbox in MATLAB software.
3. The method for reconstructing a well log based on a genetic neural network algorithm as claimed in claim 1, wherein: and (4) the genetic algorithm coding mode in the step (4) adopts binary gene coding.
4. The method for reconstructing a log based on a genetic neural network algorithm as claimed in claim 1, wherein: and (4) selecting, crossing and mutating three genetic operators in the step (4) by using a genetic algorithm toolbox in MATLAB software.
5. The method for reconstructing a well log based on a genetic neural network algorithm as claimed in claim 4, wherein: the mutation operator realizes gene mutation processing by setting mutation probability.
6. The method for reconstructing a well log based on a genetic neural network algorithm as claimed in claim 4, wherein: the crossover operator randomly sets crossover points in the gene string, and exchanges individuals before and after the crossover points to form a new gene string.
7. The method for reconstructing a well log based on a genetic neural network algorithm as claimed in claim 4, wherein: the selection operator adopts a roulette selection method.
CN202011407762.5A 2020-12-04 2020-12-04 Logging curve reconstruction method based on genetic neural network algorithm Active CN112464483B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011407762.5A CN112464483B (en) 2020-12-04 2020-12-04 Logging curve reconstruction method based on genetic neural network algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011407762.5A CN112464483B (en) 2020-12-04 2020-12-04 Logging curve reconstruction method based on genetic neural network algorithm

Publications (2)

Publication Number Publication Date
CN112464483A CN112464483A (en) 2021-03-09
CN112464483B true CN112464483B (en) 2022-12-20

Family

ID=74805524

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011407762.5A Active CN112464483B (en) 2020-12-04 2020-12-04 Logging curve reconstruction method based on genetic neural network algorithm

Country Status (1)

Country Link
CN (1) CN112464483B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113688572B (en) * 2021-08-30 2023-08-15 桂林电子科技大学 Coal-rock mass three-dimensional perception reconstruction method based on active excitation infrared temperature characterization

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104533400A (en) * 2014-11-12 2015-04-22 中国海洋石油总公司 Method for reconstructing logging curve
CN110445581A (en) * 2019-08-10 2019-11-12 南京理工大学 The method for reducing the channel decoding bit error rate based on convolutional neural networks
CN110927798A (en) * 2018-09-20 2020-03-27 中国石油化工股份有限公司 Logging curve prediction method and system based on deep learning

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104533400A (en) * 2014-11-12 2015-04-22 中国海洋石油总公司 Method for reconstructing logging curve
CN110927798A (en) * 2018-09-20 2020-03-27 中国石油化工股份有限公司 Logging curve prediction method and system based on deep learning
CN110445581A (en) * 2019-08-10 2019-11-12 南京理工大学 The method for reducing the channel decoding bit error rate based on convolutional neural networks

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
automatic recognition and reconstruction of abnormal density and acoustic log curves: a case study from Bbasin,Chad;he miao et al.;《Progress in Geophysics》;20181020;第33卷(第5期);1911-1918 *
BP神经网络技术在声波测井曲线重构中的运用;杨志力 等;《西南石油大学学报（自然科学版）》;20080220;第30卷(第1期);63-66 *
一种基于深度神经网络模型的测井曲线生成方法;张宝录 等;《电子测量技术》;20200615;第43卷(第11期);107-111 *

Also Published As

Publication number Publication date
CN112464483A (en) 2021-03-09

Similar Documents

Publication Publication Date Title
US20210026031A1 (en) Dolomite reservoir prediction method and system based on well and seismic combination, and storage medium
CN111458748B (en) Performance earthquake motion risk analysis method based on three-layer data set neural network
CN110346831B (en) Intelligent seismic fluid identification method based on random forest algorithm
CN106405640A (en) Automatic microseismic signal arrival time picking method based on depth belief neural network
AU2017424316A1 (en) Deep learning based reservoir modeling
CN109447249B (en) Antagonistic neural network well logging data reconstruction method based on deep convolution
CN112712025A (en) Complex lithology identification method based on long-term and short-term memory neural network
CN114723095A (en) Missing well logging curve prediction method and device
CN111027882A (en) Method for evaluating brittleness index by utilizing conventional logging data based on high-order neural network
CN111914478A (en) Comprehensive geological drilling well logging lithology identification method
CN113297174B (en) Land utilization change simulation method based on deep learning
CN113610945A (en) Ground stress curve prediction method based on hybrid neural network
CN111058840A (en) Organic carbon content (TOC) evaluation method based on high-order neural network
CN112464483B (en) Logging curve reconstruction method based on genetic neural network algorithm
CN113947198A (en) Logging curve reconstruction method based on nonlinear autoregressive neural network model
CN115809601A (en) Sedimentary rock structure background distinguishing method
CN114114414A (en) Artificial intelligence prediction method for 'dessert' information of shale reservoir
CN110988997A (en) Hydrocarbon source rock three-dimensional space distribution quantitative prediction technology based on machine learning
Xu et al. Stability Analysis of Geotechnical Landslide Based on GA‐BP Neural Network Model
CN110633868A (en) Method for predicting properties of oil and gas in exploration well oil testing layer by optimizing neural network through genetic algorithm
Li et al. Digital construction of geophysical well logging curves using the LSTM deep-learning network
CN115983094A (en) Logging curve generation method based on S-CNN-Bi-GRU network, processing terminal and readable storage medium
CN110552693A (en) layer interface identification method of induction logging curve based on deep neural network
CN116167288A (en) UA2FT semi-supervised transfer learning modeling method, medium and equipment
CN115936234A (en) Thin reservoir space distribution prediction method based on deep learning

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