CN107292406A - Seismic properties method for optimizing based on vector regression and genetic algorithm - Google Patents
Seismic properties method for optimizing based on vector regression and genetic algorithm Download PDFInfo
- Publication number
- CN107292406A CN107292406A CN201610191786.9A CN201610191786A CN107292406A CN 107292406 A CN107292406 A CN 107292406A CN 201610191786 A CN201610191786 A CN 201610191786A CN 107292406 A CN107292406 A CN 107292406A
- Authority
- CN
- China
- Prior art keywords
- mrow
- vector regression
- seismic properties
- attribute
- msub
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 52
- 230000002068 genetic effect Effects 0.000 title claims abstract description 43
- 238000000034 method Methods 0.000 title claims abstract description 42
- 238000012360 testing method Methods 0.000 claims abstract description 31
- 238000012549 training Methods 0.000 claims description 23
- 238000009395 breeding Methods 0.000 claims description 9
- 230000001488 breeding effect Effects 0.000 claims description 9
- 238000005457 optimization Methods 0.000 claims description 7
- 238000011158 quantitative evaluation Methods 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000002790 cross-validation Methods 0.000 claims description 6
- 238000011156 evaluation Methods 0.000 claims description 5
- 108090000623 proteins and genes Proteins 0.000 claims description 4
- 101150071716 PCSK1 gene Proteins 0.000 claims description 3
- 238000012952 Resampling Methods 0.000 claims description 2
- 230000008901 benefit Effects 0.000 description 7
- 238000010353 genetic engineering Methods 0.000 description 6
- 230000008859 change Effects 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 230000001850 reproductive effect Effects 0.000 description 4
- 210000000349 chromosome Anatomy 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 230000006978 adaptation Effects 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000010429 evolutionary process Effects 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 238000010845 search algorithm Methods 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000006866 deterioration Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000004043 dyeing Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000008303 genetic mechanism Effects 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 230000004083 survival effect Effects 0.000 description 1
- 230000035899 viability Effects 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/02—Agriculture; Fishing; Forestry; Mining
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Theoretical Computer Science (AREA)
- Strategic Management (AREA)
- Biophysics (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Tourism & Hospitality (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Agronomy & Crop Science (AREA)
- Animal Husbandry (AREA)
- Marine Sciences & Fisheries (AREA)
- Mining & Mineral Resources (AREA)
- Molecular Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Primary Health Care (AREA)
- Biomedical Technology (AREA)
- Genetics & Genomics (AREA)
- Physiology (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
Abstract
The present invention provides a kind of seismic properties method for optimizing based on vector regression and genetic algorithm, and being somebody's turn to do the seismic properties method for optimizing based on vector regression and genetic algorithm includes:Step 1, log data and seismic properties are pre-processed;Step 2, sample is extracted, and is grouped sample set in time-domain;Step 3, seismic properties are carried out preferably, preferred seismic properties subset and the nonlinear support vector regression model of reservoir is obtained;Step 4, the support vector regression model of foundation is tested using test data, when error satisfaction is required, obtains optimum attributes combination;Step 5, output optimum attributes combination.The seismic properties method for optimizing based on vector regression and genetic algorithm is directed to specific reservoir characteristic parameter, and the Sensitive Attributes combination of reservoir characteristic can most be reflected by searching out.
Description
Technical field
The present invention relates to In Oil Field Exploration And Development technical field, especially relate to a kind of based on vector regression
With the seismic properties method for optimizing of genetic algorithm.
Background technology
When carrying out Seismic Reservoir Prediction, the various seismic properties relevant with reservoir prediction are usually introduced.Ground
The introducing of shake attribute is typically passed through one from less to more, and from more to few process.It is so-called from less to more,
Refer to design prediction scheme initial stage should as far as possible it is many include it is various may be with reservoir prediction
Relevant attribute.Various useful information can be so made full use of, the experience of each side expert is absorbed,
Improve the effect of reservoir prediction.But, it is unfavorable that the unlimited increase of attribute is also brought along for reservoir prediction
Influence, because:
(1) some seismic properties may be unrelated in itself with target zone, and reflects the change of shallow-layer interference,
If being not added with differentiating to input attribute, some attributes can only cause confusion;
(2) increase attribute can bring difficulty to calculating, because excessive data take substantial amounts of storage sky
Between and calculate the time;
(3) it is certain to include many factors being relative to each other in substantial amounts of attribute, causes the weight of information
Multiple and waste;
(4) attribute number is relevant with number of training.For pattern-recognition, when sample number is fixed
When, attribute number can excessively cause the deterioration of classifying quality.
Therefore, for particular problem, it is necessary to select some best earthquake category from numerous seismic properties
Property or combinations of attributes.
Conventional attribute method for optimizing has attribute comparison method, sequential advancement method, order retrogressing method, heredity to calculate
Method etc..
Genetic algorithm (GA) is a kind of full search algorithm for solving combination optimal problem.Introduce genetic algorithm
The overall situation or approximate optimal solution of combinatorial optimization problem can be found, that is, is found out optimal or secondary in reservoir prediction
Excellent seismic properties combination.The algorithm is a kind of based on living nature natural selection and natural genetic mechanism solution combination
The full search algorithm of optimization problem, but it is different from common searching algorithm.Common searching algorithm one
As simply from a solution be improved to another preferably solution and genetic algorithm be then from former problem one
Group solution, which is set out, to be improved to another group and preferably solves.
Genetic algorithm is mainly made up of following problem:
(1) encoded question
Genetic algorithm can not direct process problem space parameter, it is necessary to convert them into hereditary space
By certain structure and by the chromosome or individual of genomic constitution.This conversion operation is referred to as coding, can also
The referred to as expression of problem.Mapping generally from from problem space to GA spaces is referred to as coding, and from GA spaces to
The mapping of problem space referred to as " is decoded ".Index of the code levels frequently with three aspects is assessed, i.e.,
Problem of completeness (all candidate solutions in space can be expressed as the chromosome in GA spaces), viability (GA
The candidate solution in dyeing physical efficiency correspondence all problems space in space) and nonredundancy (chromosome and time
Choosing solution is corresponded).
(2) colony sets
Genetic manipulation is that numerous individuals are carried out simultaneously, and these numerous individuals constitute colony.In heredity
Task after the design of algorithm process process encoding is the setting of initial population, an and generation one with this as the starting point
In generation, ground evolved, and terminates evolutionary process until meeting certain evolution stopping criterion, now obtains last generation
Or colony.The setting of initial population can take following strategy in genetic algorithm:1. inherently known according to problem
Know, try to hold the distribution that optimal solution takes up space in whole problem space, and in this scope
Interior setting initial population;2. the individual of random generation certain amount, and therefrom choose best individual plus people
Into initial population.This continuous iteration of process, until number reaches predetermined rule in initial population
Mould.
(3) genetic manipulation
Genetic manipulation is the operation for simulating biological gene heredity.In genetic algorithm by encoding composition initial population
After body, the task of genetic manipulation is exactly, to the individual in colony, to be fitted according to their adaptednesses to environment
Response, which is assessed, completes certain operation, so as to realize the evolutionary process of the survival of the fittest.From the angle of Optimizing Search
For degree, genetic manipulation can be such that the solution of problem optimizes by generation, and approach optimal solution.Genetic manipulation includes choosing
Select, intersect and make a variation three basic genetic operators, they have randomization features of operation, its effect with
Evolutionary operator probability, coding method, group size, initial population and adaptation acquired by three genetic operators
The setting for spending function is closely related.
Genetic algorithm have one it is very crucial the problem of:The evaluation of individual in population fitness.2002,
Wang Yonggang professor delivers《The GA-BP optimization methods of seismic properties》, by genetic algorithm and neural network algorithm
It is combined, plays the advantage of two kinds of algorithms, finds global optimum, but neural network algorithm is easy in itself
Sink into local optimum, and sample data requirement is big.2006, teacher Wei Zhenzhong delivered《Based on branch
Hold the feature selecting of vector machine and genetic algorithm》, in fitness evaluation, using support vector classification,
The qualitative classification to attribute is realized, the height of quantitative assessment is not risen to., teacher Zhang Changkai in 2012
Deliver《Attribute based on SVMs is preferably and reservoir prediction》, introducing support vector regression can be with
The quantitative calculated relationship of optimization set up between reservoir characteristic parameter and preferred attribute, can be preferred to attribute
Realize quantitative evaluation.For this, we have invented a kind of new ground based on vector regression and genetic algorithm
Attribute method for optimizing is shaken, above technical problem is solved.
The content of the invention
Support vector regression is incorporated among genetic algorithm it is an object of the invention to provide one kind, passed through
To the quantitative evaluation of preferred attribute, come improve the preferred accuracy rate of attribute based on vector regression and something lost
The seismic properties method for optimizing of propagation algorithm.
The purpose of the present invention can be achieved by the following technical measures:Based on vector regression and genetic algorithm
Seismic properties method for optimizing, should seismic properties method for optimizing bag based on vector regression and genetic algorithm
Include:Step 1, log data and seismic properties are pre-processed;Step 2, extract sample, and when
Between sample set is grouped on domain;Step 3, seismic properties are carried out preferably, preferred seismic properties is obtained
Collection and the nonlinear support vector regression model of reservoir;Step 4, branch of the test data to foundation is utilized
Hold vector regression model to test, when error satisfaction is required, obtain optimum attributes combination;Step
5, output optimum attributes combination.
The purpose of the present invention can be also achieved by the following technical measures:
In step 1, resampling is carried out to log data, makes its sample rate identical with geological data;
Extract the various prestack attributes and poststack attribute of angle domain common image gathers, prestack attribute bag in the well location place of putting
Intercept and gradient attribute are included, then prestack poststack attribute is standardized.
In step 2, pretreated log data and well location are put to the seismic properties composition one a pair at place
The sample set answered, is divided into two groups in time-domain by sample set:Training data and test data, train number
According to for preferred attribute and setting up support vector regression model, test data be used for obtained support to
Amount regression machine model is tested.
In step 3, it is combined, training data is carried out preferably using vector regression and genetic algorithm,
Preferred seismic properties subset and the nonlinear vector regression model of reservoir are obtained, passes through vector regression
Algorithm realizes the preferred quantitative evaluation of attribute, improves the preferred accuracy rate of attribute.
In step 3, binary coding is carried out to seismic properties, and is randomly formed initial population, i.e.
Binary coding, coded strings h=h are carried out to property set { att1, att2 ..., attN }1h2…hNRepresent to property set
The once selection done, wherein hi=1 represents that ith attribute is chosen, wherein hi=0 represents ith attribute
Not selected, the combination of each selected attribute of binary coding string is in an initial population
Body.
In step 3, cross validation is carried out using support vector regression and genetic algorithm, in colony
The fitness of individual is detected and assessed, so that preferred attribute, cross validation is with the training in step 2
Based on data, using the crosscheck mean square deviation of support vector regression model as genetic algorithm adaptation
Spend evaluation criterion;The individual of training data is divided into m parts from time-domain, built every time with 1 part therein
Mould, remaining m-1 parts of test, according to that 1 number according to utilization;Support vector regression obtains log
The quantitative calculated relationship of optimization between attribute, the root mean square for then calculating remaining m-1 number evidence is pre-
Survey error eRMSj;The root-mean-square prediction error tested every time after Repeated m time modeling, test, so
The average value of m root-mean-square prediction error is calculated afterwards, that is, obtains the intersection inspection of support vector regression model
Test mean square deviation;Wherein, crosscheck mean square deviation RMSECV calculation formula is as follows:
In formula 1, RMSECV represents the crosscheck mean square deviation of support vector regression model, and m is represented
Training data individual has divided how many part, e in time-domainRMSjRepresent the root mean square prediction of every part of test data
Error;Formula 2 is eRMSjExpression, wherein, n is sample number, yiFor actual value,For
Predicted value;In formula 3, xj(j=1 ..., k) represents the attribute included in individual, and f is to be gone to intend with attribute
Close actual value yiWhen function expression.
In step 3, according to the acquired results of formula 1, the half individual of fitness difference is rejected, to fitness
Good half individual is bred, and intersection and the variation of gene are carried out in breeding.
In step 3, judge the preferred attribute got using training data whether meet error requirements or
The algebraically of maximum breeding is reached, if error is bigger than setting error and is not up to maximum reproductive order of generation, after
The continuous breeding and evolution for carrying out individual;If error is less than setting error or reaches maximum reproductive order of generation,
Then enter step 4.
In step 4, error calculation formula is as follows:
In formula, e represents the error amount of test data, yiFor actual value,For predicted value, l is test
The sample number of data;
When error is unsatisfactory for requiring, return to step 3, re-start seismic properties preferably with modeling.
In steps of 5, preserve optimum attributes combination and based on vector regression model parameter, be easy to prediction
The log data distribution of even whole work area scope at well point.
The seismic properties method for optimizing based on vector regression and genetic algorithm in the present invention, by heredity calculation
Method is combined with support vector regression, merges the advantage of two kinds of technologies.The advantage of genetic algorithm is can be with
Preferably go out another group of attribute from one group of attribute, rather than preferably go out another attribute from an attribute, and
Global it can find optimal solution;The advantage of support vector regression is the learning ability with small sample, general
Change ability is strong, and can realize the quantitative evaluation to attribute, improves the preferred accuracy rate of attribute.Base
Above advantage is merged in SVR-GA seismic properties method for optimizing, so that for specific reservoir characteristic ginseng
Number, the Sensitive Attributes combination of reservoir characteristic can most be reflected by searching out.
Brief description of the drawings
Fig. 1 is a tool of the seismic properties method for optimizing based on vector regression and genetic algorithm of the present invention
The flow chart of body embodiment;
Fig. 2 is calculated in the specific embodiment for the present invention using SVR-GA methods for training data
GR prediction curves and actual curve comparison diagram;
Fig. 3 is tested in the specific embodiment for the present invention with test data to the SVR models of foundation,
The GR prediction curves and the comparison diagram of actual curve drawn.
Embodiment
For enable the present invention above and other objects, features and advantages become apparent, it is cited below particularly go out
Preferred embodiment, and coordinate shown in accompanying drawing, it is described in detail below.
As shown in figure 1, Fig. 1 is preferred for the seismic properties based on vector regression and genetic algorithm of the present invention
The flow chart of method.
In a step 101, log data and seismic properties are pre-processed.Log data is weighed
Sampling, makes its sample rate identical with geological data;Angle domain common image gathers are extracted at the well location place of putting
Various prestack attributes (including intercept and gradient attribute) and poststack attribute, then enter to prestack poststack attribute
Row standardization.
In a step 102, sample extraction is carried out.Log data and well location after arrangement is put to the earthquake at place
Attribute constitutes one-to-one sample set, and sample set is divided into two groups in time-domain:Training data and survey
Data are tried, training data is used for preferred attribute and sets up SVR models, and test data is used for the SVR to obtaining
Model is tested.
In step 103, carry out seismic properties preferably with modeling.It is combined using GA with SVR, to training
Data carry out preferably, obtaining preferred seismic properties subset and the non-linear SVR models of reservoir.Step 103 can
To realize the preferred quantitative evaluation of attribute by SVR algorithms, the preferred accuracy rate of attribute is improved.
In one embodiment, binary coding is carried out to seismic properties, and is randomly formed initial population.
That is, binary coding, coded strings h=h are carried out to property set { att1, att2 ..., attN }1h2…hNRepresent to category
The once selection that property collection is done, wherein hi=1 represents that ith attribute is chosen, wherein hi=0 represents i-th
Attribute is not selected, during the combination of each selected attribute of binary coding string is an initial population
Individual.
Cross validation is carried out using support vector regression (SVR) and genetic algorithm (GA), to colony
Middle individual fitness is detected and assessed, so that preferred attribute.Cross validation mentioned here be with
It is based on training data described in step 2, the crosscheck mean square deviation (RMSECV) of SVR models is (public
Formula 1) it is used as GA fitness evaluation standard.The individual of training data is divided into m parts from time-domain, often
Secondary to be modeled with 1 part therein, remaining m-1 parts of test is bent according to well logging is obtained using SVR according to that 1 number
The quantitative calculated relationship of optimization between line and attribute, then calculates the root mean square of remaining m-1 number evidence
Predicated error (eRMSj) (formula 2);The root mean square tested every time after Repeated m time modeling, test
Predicated error, then calculates the average value of m root-mean-square prediction error, that is, obtains the intersection of SVR models
Examine mean square deviation (formula 1).
Expression is as follows:
In formula 1, RMSECV represents the crosscheck mean square deviation of SVR models, and m represents training data individual
How many part, e divided in time-domainRMSjRepresent the root-mean-square prediction error of every part of test data;Formula 2
It is eRMSjExpression, wherein, n is sample number, yiFor actual value,It is (specific for predicted value
Expression formula such as formula 3);In formula 3, xj(j=1 ..., k) represents the attribute included in individual, and f is to use
Attribute removes fitting actual value yiWhen function expression.
According to the acquired results of formula 1, the half individual of fitness difference is rejected, good to fitness 1 half
Body is bred, and intersection and the variation of gene are carried out in breeding.
Judge whether the preferred attribute got using training data is met error requirements or reached maximum numerous
The algebraically grown, if error is bigger than setting error and is not up to maximum reproductive order of generation, proceeds individual
Breeding and evolution;If error is less than setting error or reaches maximum reproductive order of generation, into step
104.As shown in Fig. 2 Fig. 2 is trained to be directed in the specific embodiment of the present invention using SVR-GA methods
GR prediction curves and the comparison diagram of actual curve that data are calculated.Figure it is seen that GR is predicted
Value is compared with actual value coincide, and crosscheck mean square deviation only has 0.9861.Using training data get it is excellent
Attribute is selected to meet error requirements
At step 104, error-tested is carried out using test data.As shown in figure 3, the tool of the present invention
The SVR models of foundation are tested with test data in body embodiment, the GR prediction curves and reality drawn
The comparison diagram of border curve.From figure 3, it can be seen that in test data, GR predicted values and actual value difference
Larger, root-mean-square prediction error is 9.6243.The SVR models obtained with test data to step 3 are examined
Test, with actual comparison, required if error is met, obtain optimum attributes combination, into step 5,
If error be unsatisfactory for require, into step 3, re-start seismic properties preferably with modeling.In step 4
In error calculation formula it is as follows:
In formula, e represents the error amount of test data, yiFor actual value,For predicted value, l is test
The sample number of data.
In step 105, output optimum attributes combination.Optimum attributes combination and SVR model parameters are preserved,
It is easy to the log data distribution of even whole work area scope at pre- logging point.
The optimum attributes combination table of table 1
The preferred result of attribute is carried out for gamma curve as shown in Table 1, prestack angle in figure in certain work area
Degree is from 1 degree to 35 degree, and attribute is from left to right followed successively by:It is instantaneous phase, centre frequency rate of change, instantaneous
Frequency change rate, centre frequency, instantaneous frequency, instantaneous amplitude, Amplitude-squared.As it can be seen from table 1
It is small, in, big all angles there is attribute preferably to be come out, and centre frequency rate of change, instantaneous frequency become
It is more that these three attributes of rate, Amplitude-squared are preferably gone out, and other 4 attribute are relatively fewer.Flow
Terminate.
The seismic properties method for optimizing based on SVR-GA in the present invention, fusion two kinds of technologies of SVR and GA
Advantage, compared with previous methods, by the quantitative evaluation to attribute, improves preferred accurate of attribute
Rate.So as to which for specific reservoir characteristic parameter, can search out can most reflect the sensitivity of reservoir characteristic
Combinations of attributes.
Claims (10)
1. the seismic properties method for optimizing based on vector regression and genetic algorithm, it is characterised in that should be based on to
The seismic properties method for optimizing of amount regression machine and genetic algorithm includes:
Step 1, log data and seismic properties are pre-processed;
Step 2, sample is extracted, and is grouped sample set in time-domain;
Step 3, seismic properties are carried out preferably, preferred seismic properties subset is obtained and reservoir is nonlinear
Support vector regression model;
Step 4, the support vector regression model of foundation is tested using test data, in error
When satisfaction is required, optimum attributes combination is obtained;
Step 5, output optimum attributes combination.
2. the seismic properties method for optimizing according to claim 1 based on vector regression and genetic algorithm,
Characterized in that, in step 1, carrying out resampling to log data, making its sample rate and geological data
It is identical;The various prestack attributes and poststack attribute of angle domain common image gathers are extracted at the well location place of putting, are folded
Preceding attribute includes intercept and gradient attribute, and then prestack poststack attribute is standardized.
3. the seismic properties method for optimizing according to claim 1 based on vector regression and genetic algorithm,
Characterized in that, in step 2, the seismic properties that log data and well location are put into place, which are constituted, to be corresponded
Sample set, sample set is divided into two groups in time-domain:Training data and test data, training data
It is used for the supporting vector to obtaining with support vector regression model, test data is set up for preferred attribute
Regression machine model is tested.
4. the seismic properties method for optimizing according to claim 1 based on vector regression and genetic algorithm,
Characterized in that, in step 3, being combined using vector regression and genetic algorithm, to training data
Preferably, obtain preferred seismic properties subset and the nonlinear vector regression model of reservoir, pass through
Vector regression algorithm realizes the preferred quantitative evaluation of attribute, improves the preferred accuracy rate of attribute.
5. the seismic properties method for optimizing according to claim 4 based on vector regression and genetic algorithm,
Characterized in that, in step 3, seismic properties are carried out with binary coding, and be randomly formed initial
Colony, i.e. binary coding, coded strings h=h are carried out to property set { att1, att2 ..., attN }1h2…hNRepresent
The once selection done to property set, wherein hi=1 represents that ith attribute is chosen, wherein hi=0 represents
Ith attribute is not selected, and the combination of each selected attribute of binary coding string is one initial
Individual in colony.
6. the seismic properties method for optimizing according to claim 5 based on vector regression and genetic algorithm,
Characterized in that, in step 3, cross validation is carried out using support vector regression and genetic algorithm,
The fitness of individual in population is detected and assessed, so that preferred attribute, cross validation is with step 2
In training data based on, using the crosscheck mean square deviation of support vector regression model be used as heredity calculate
The fitness evaluation standard of method;The individual of training data is divided into m parts from time-domain, every time with wherein
1 part modeling, remaining m-1 part test, according to that 1 number according to utilization;Support vector regression is surveyed
The quantitative calculated relationship of optimization between well curve and attribute, then calculates the equal of remaining m-1 number evidence
Root predicated error eRMSj;The root mean square prediction tested every time after Repeated m time modeling, test is missed
Difference, then calculates the average value of m root-mean-square prediction error, that is, obtains support vector regression model
Cross-check mean square deviation;Wherein, crosscheck mean square deviation RMSECV calculation formula is as follows:
<mrow>
<mi>R</mi>
<mi>M</mi>
<mi>S</mi>
<mi>E</mi>
<mi>C</mi>
<mi>V</mi>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mi>m</mi>
</mfrac>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>m</mi>
</munderover>
<mo>&lsqb;</mo>
<mfrac>
<mn>1</mn>
<mrow>
<mi>m</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</mfrac>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mrow>
<mi>m</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</munderover>
<msub>
<mi>e</mi>
<mrow>
<mi>R</mi>
<mi>M</mi>
<mi>S</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>&rsqb;</mo>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<msub>
<mi>e</mi>
<mrow>
<mi>R</mi>
<mi>M</mi>
<mi>S</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>=</mo>
<msqrt>
<mrow>
<mfrac>
<mn>1</mn>
<mi>n</mi>
</mfrac>
<munderover>
<mo>&Sigma;</mo>
<mi>i</mi>
<mi>n</mi>
</munderover>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mover>
<mi>y</mi>
<mo>^</mo>
</mover>
<mi>i</mi>
</msub>
<mo>-</mo>
<msub>
<mi>y</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
</mrow>
</msqrt>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>2</mn>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<msub>
<mover>
<mi>y</mi>
<mo>^</mo>
</mover>
<mi>i</mi>
</msub>
<mo>=</mo>
<mi>f</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mn>1</mn>
</msub>
<mo>,</mo>
<msub>
<mi>x</mi>
<mn>2</mn>
</msub>
<mo>,</mo>
<mo>...</mo>
<mo>,</mo>
<msub>
<mi>x</mi>
<mi>k</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>,</mo>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
<mo>,</mo>
<mo>...</mo>
<mo>,</mo>
<mi>n</mi>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>3</mn>
<mo>)</mo>
</mrow>
</mrow>
In formula 1, RMSECV represents the crosscheck mean square deviation of support vector regression model, and m is represented
Training data individual has divided how many part, e in time-domainRMSjRepresent the root mean square prediction of every part of test data
Error;Formula 2 is eRMSjExpression, wherein, n is sample number, yiFor actual value,For
Predicted value;In formula 3, xj(j=1 ..., k) represents the attribute included in individual, and f is to be gone to intend with attribute
Close actual value yiWhen function expression.
7. the seismic properties method for optimizing according to claim 6 based on vector regression and genetic algorithm,
Characterized in that, in step 3, according to the acquired results of formula 1, the half individual of fitness difference is rejected,
The half individual good to fitness is bred, and intersection and the variation of gene are carried out in breeding.
8. the seismic properties method for optimizing according to claim 7 based on vector regression and genetic algorithm,
Characterized in that, in step 3, judging whether the preferred attribute got using training data meets error
It is required that or reach the algebraically of maximum breeding, if error is bigger than setting error and not up to maximum breeding generation
Number, then proceed the breeding and evolution of individual;If error is less than setting error or reached maximum numerous
Algebraically is grown, then into step 4.
9. the seismic properties method for optimizing according to claim 1 based on vector regression and genetic algorithm,
Characterized in that, in step 4, error calculation formula is as follows:
<mrow>
<mi>e</mi>
<mo>=</mo>
<msqrt>
<mrow>
<mfrac>
<mn>1</mn>
<mi>l</mi>
</mfrac>
<munderover>
<mo>&Sigma;</mo>
<mi>i</mi>
<mi>l</mi>
</munderover>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mover>
<mi>y</mi>
<mo>^</mo>
</mover>
<mi>i</mi>
</msub>
<mo>-</mo>
<msub>
<mi>y</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
</mrow>
</msqrt>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>4</mn>
<mo>)</mo>
</mrow>
</mrow>
In formula, e represents the error amount of test data, yiFor actual value,For predicted value, l is test
The sample number of data;
When error is unsatisfactory for requiring, return to step 3, re-start seismic properties preferably with modeling.
10. the seismic properties method for optimizing according to claim 1 based on vector regression and genetic algorithm,
Characterized in that, in steps of 5, preserve optimum attributes combination and based on vector regression model parameter,
It is easy to the log data distribution of even whole work area scope at pre- logging point.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610191786.9A CN107292406A (en) | 2016-03-30 | 2016-03-30 | Seismic properties method for optimizing based on vector regression and genetic algorithm |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610191786.9A CN107292406A (en) | 2016-03-30 | 2016-03-30 | Seismic properties method for optimizing based on vector regression and genetic algorithm |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107292406A true CN107292406A (en) | 2017-10-24 |
Family
ID=60087892
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610191786.9A Pending CN107292406A (en) | 2016-03-30 | 2016-03-30 | Seismic properties method for optimizing based on vector regression and genetic algorithm |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107292406A (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109271734A (en) * | 2018-10-11 | 2019-01-25 | 河南工业大学 | A kind of earthquake-induced site liquefaction potential evaluation method based on machine learning techniques |
CN109633748A (en) * | 2018-11-12 | 2019-04-16 | 中国石油大学(华东) | A kind of seismic properties preferred method based on improved adaptive GA-IAGA |
CN109858093A (en) * | 2018-12-28 | 2019-06-07 | 浙江工业大学 | The air source heat pump multi-objective optimization design of power method of the non-dominated sorted genetic algorithm of SVR neural network aiding |
CN110276497A (en) * | 2019-06-27 | 2019-09-24 | 河南工业大学 | A kind of point prediction method of barn building air-tightness index |
CN110533150A (en) * | 2019-07-05 | 2019-12-03 | 江西财经大学 | Self -adaptive and reuse system and method based on Support vector regression model |
CN113945977A (en) * | 2021-10-26 | 2022-01-18 | 中国石油大学(华东) | Binary data permutation and combination based seismic attribute multi-element fusion display and storage method |
CN116340757A (en) * | 2023-04-25 | 2023-06-27 | 中国地震局地震研究所 | Characteristic self-adaptive earthquake early-warning magnitude prediction method and system thereof |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102005135A (en) * | 2010-12-09 | 2011-04-06 | 上海海事大学 | Genetic algorithm-based support vector regression shipping traffic flow prediction method |
-
2016
- 2016-03-30 CN CN201610191786.9A patent/CN107292406A/en active Pending
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102005135A (en) * | 2010-12-09 | 2011-04-06 | 上海海事大学 | Genetic algorithm-based support vector regression shipping traffic flow prediction method |
Non-Patent Citations (4)
Title |
---|
TING XIONG ET AL.: "The research on the joint application of the seismic attributes optimizing based on the genetic algorithm and the neural network methods", 《2011 SEVENTH INTERNATIONAL CONFERENCE ON NATURAL COMPUTATION》 * |
杨翔: "数据驱动法预测储层参数的研究及应用", 《中国优秀博硕士学位论文全文数据库(硕士)基础科学辑》 * |
赵龙飞 等: "基于遗传算法和支持向量机的地震属性优选方法研究", 《中国地球科学联合学术年会 2014》 * |
韦振中 等: "基于支持向量机和遗传算法的特征选择", 《广西工学院学报》 * |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109271734A (en) * | 2018-10-11 | 2019-01-25 | 河南工业大学 | A kind of earthquake-induced site liquefaction potential evaluation method based on machine learning techniques |
CN109271734B (en) * | 2018-10-11 | 2023-04-07 | 河南工业大学 | Site earthquake liquefaction potential evaluation method based on machine learning technology |
CN109633748A (en) * | 2018-11-12 | 2019-04-16 | 中国石油大学(华东) | A kind of seismic properties preferred method based on improved adaptive GA-IAGA |
CN109633748B (en) * | 2018-11-12 | 2020-05-01 | 中国石油大学(华东) | Seismic attribute optimization method based on improved genetic algorithm |
CN109858093A (en) * | 2018-12-28 | 2019-06-07 | 浙江工业大学 | The air source heat pump multi-objective optimization design of power method of the non-dominated sorted genetic algorithm of SVR neural network aiding |
CN110276497A (en) * | 2019-06-27 | 2019-09-24 | 河南工业大学 | A kind of point prediction method of barn building air-tightness index |
CN110533150A (en) * | 2019-07-05 | 2019-12-03 | 江西财经大学 | Self -adaptive and reuse system and method based on Support vector regression model |
CN110533150B (en) * | 2019-07-05 | 2023-05-23 | 江西财经大学 | Test generation and reuse system and method based on support vector machine regression model |
CN113945977A (en) * | 2021-10-26 | 2022-01-18 | 中国石油大学(华东) | Binary data permutation and combination based seismic attribute multi-element fusion display and storage method |
CN113945977B (en) * | 2021-10-26 | 2023-08-18 | 中国石油大学(华东) | Seismic attribute fusion display and storage method based on binary permutation and combination |
CN116340757A (en) * | 2023-04-25 | 2023-06-27 | 中国地震局地震研究所 | Characteristic self-adaptive earthquake early-warning magnitude prediction method and system thereof |
CN116340757B (en) * | 2023-04-25 | 2023-08-08 | 中国地震局地震研究所 | Characteristic self-adaptive earthquake early-warning magnitude prediction method and system thereof |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107292406A (en) | Seismic properties method for optimizing based on vector regression and genetic algorithm | |
Lino et al. | Virtual teaching and learning environments: Automatic evaluation with symbolic regression | |
Roodposhti et al. | Towards automatic calibration of neighbourhood influence in cellular automata land-use models | |
CN108763857A (en) | A kind of process soft-measuring modeling method generating confrontation network based on similarity | |
CN106528433A (en) | Testing case priority ranking method for white-box testing | |
CN105975573A (en) | KNN-based text classification method | |
CN111695042B (en) | User behavior prediction method and system based on deep walking and ensemble learning | |
CN104536881A (en) | Public testing error report priority sorting method based on natural language analysis | |
US20210381362A1 (en) | Method and apparatus for estimating lithofacies by learning well logs | |
CN105893876A (en) | Chip hardware Trojan horse detection method and system | |
Kangas et al. | Applying stochastic multicriteria acceptability analysis to forest ecosystem management with both cardinal and ordinal criteria | |
McCulloch et al. | Calibrating agent-based models using uncertainty quantification methods | |
CN111639803A (en) | Prediction method applied to future vegetation index of area under climate change scene | |
CN109472346B (en) | Emergency material demand prediction method considering fuzzy and missing of partial data | |
CN109633748A (en) | A kind of seismic properties preferred method based on improved adaptive GA-IAGA | |
CN114757433B (en) | Method for rapidly identifying relative risk of drinking water source antibiotic resistance | |
CN112529684A (en) | Customer credit assessment method and system based on FWA _ DBN | |
CN115018357A (en) | Farmer portrait construction method and system for production performance improvement | |
CN106935038A (en) | One kind parking detecting system and detection method | |
Ek et al. | Development of individual tree based stand growth simulators: progress and applications. | |
Das et al. | Evaluation of liquefaction potential of soil using genetic programming | |
CN105930272A (en) | Important statement-based branch coverage testing data generation method | |
Helidoniotis et al. | Growth models for fisheries: The effect of unbalanced sampling error on model selection, parameter estimation, and biological predictions | |
CN111582634B (en) | Multi-factor safety grading method and system for underground large-space construction | |
Putri et al. | Optimization of analytic hierarchy process using genetic algorithm for selecting tutoring agencies in Kampung Inggris Pare |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20171024 |