CN108573078A - Post-frac effect forecasting method based on data mining - Google Patents
Post-frac effect forecasting method based on data mining Download PDFInfo
- Publication number
- CN108573078A CN108573078A CN201710137665.0A CN201710137665A CN108573078A CN 108573078 A CN108573078 A CN 108573078A CN 201710137665 A CN201710137665 A CN 201710137665A CN 108573078 A CN108573078 A CN 108573078A
- Authority
- CN
- China
- Prior art keywords
- data
- effect
- fracturing
- model
- data mining
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/06—Multi-objective optimisation, e.g. Pareto optimisation using simulated annealing [SA], ant colony algorithms or genetic algorithms [GA]
Abstract
The present invention provides a kind of Post-frac effect forecasting method based on data mining, including：Fracturing effect pressure break influence factor table is designed, pressure break database is established；Data Warehouse Design is completed, data warehouse is built；Integral data resource checks the integrality and correctness of data；Selection is suitable for the data mining algorithmic method of pressure break data analysis, formation algorithm class libraries；Build the data mining system for being suitable for pressure break data analysis, mating corresponding development environment；The multidimensional of data is shown and common analysis；Using data mining algorithm, fracturing effect and model of influencing factors are established；Pressure break influence factor sensitivity analysis is carried out, determines the principal element for influencing fracturing effect；Establish fracturing effect and the model of major influence factors；Carry out the quantitative forecast of fracturing effect；Carry out the optimization of construction parameter.The accuracy that Post-frac effect forecasting can be improved using the method can improve fracturing effect by Optimum Fracturing design parameter.
Description
Technical field
The present invention relates to the technical fields of Reservoir Development, especially relate to a kind of fracturing effect based on data mining
Prediction technique.
Background technology
With going deep into for oil field development, the development difficulty of oil-gas reservoir is gradually increased.Hyposmosis, spy are hypotonic in new proved reserves
Oil-gas reservoir proportion increases year by year, and development cost is also increasing.Fracturing reform is main as hyposmosis, extra-low permeability oil reservoirs
Well stimulation, how to improve the effective percentage of pressing crack construction, be improve LOW PERMEABILITY OILFIELD DEVELOPMENT effect key link.
Currently used Post-frac effect forecasting method is mainly based upon rational mechanics model.Method master based on mechanical model
If based on all kinds of mechanical models, including crack extended model, fracturing fluid leak model, fracturing fluid motion model, branch
Support agent conveying model, oil and gas flow model etc..Method advantage based on mechanical model is that model is abundant, is suitable for scientific research.
But when practical application, the model parameter based on mechanics method is difficult to obtain, and there are many uncertain variables.This often results in reality
Border construction is difficult to carry out production capacity and actual result after design requirement, pre- pressure measurement and differs greatly, and reduces the effective percentage of pressing crack construction.
The leading technology that pressure break is developed as hyposmosis, the whole nation is annual to implement the thousands of wells of hydraulic fracturing.By constantly
Desk research and field test have formulated conceptual design, component on site construction method and measure for the transformation of various low-permeability oil deposits,
Not only it solves practical problem, but also has accumulated abundant experience and knowledge.But Post-frac effect forecasting is carried out by experience merely,
It is more rationally and more accurate than being based on theory deduction sometimes, but there is certain blindness and subjectivity.
In conclusion be based purely on the decision-making technique of theory deduction and experience, may cause after pressure effect prediction difference compared with
Greatly, construction failure is even resulted in, huge economic loss is brought to oil field.Therefore, it is necessary to study being more in line with actual conditions,
Binding isotherm derives and the advantage of two kinds of Post-frac effect forecasting methods of experience, on the basis for making full use of a large amount of history well data
On, by the method for data mining, improve the effective percentage and success rate of pressure break.
The engineering field that experience and example are used for reference, data mining technology is largely relied on to have in needs such as oil-gas field developments
Its only thick advantage.Data mining technology handles the intension of various geologic(al) factors, it will the thoroughly traditional modeling assumption of change, it
Intelligent fuzzy processing means various geologic(al) factors will be gone automatically time to deposit master, eliminate the false and retain the true, also broken away from model solution
Trouble.
Therefore, the technical issues of needing those skilled in the art urgently to solve at present be exactly：It proposes a kind of based on data
The Post-frac effect forecasting method of excavation improves the accuracy of Post-frac effect forecasting, and it is efficient to improve pressure break.Thus we have invented
A kind of new Post-frac effect forecasting method based on data mining, solves the above technical problem.
Invention content
The object of the present invention is to provide it is a kind of aim at improve fracturing technology effective percentage, to pressure break historical data into
On the basis of row analysis and modeling, achieve the purpose that effect prediction and process parameter optimizing, improves the effective percentage of fracturing technology
The Post-frac effect forecasting method based on data mining.
The purpose of the present invention can be achieved by the following technical measures：Post-frac effect forecasting method based on data mining,
The Post-frac effect forecasting method based on data mining includes：Step 1, fracturing effect influence factor table is designed, pressure break is established
Database；Step 2, design data storage builds data warehouse；Step 3, integral data resource, check data integrality and
Correctness；Step 4, suitable data mining algorithm, organized data algorithm class libraries are selected；Step 5, suitable data analysis is built
Platform, mating development environment；Step 6, multidimensional data displaying and primary data analysis are carried out to data warehouse；Step 7, number is utilized
The model of fracturing effect and pressure break influence factor is established according to mining algorithm；Step 8, the quick of fracturing effect and each influence factor is carried out
Perceptual analysis determines the principal element for influencing fracturing effect；Step 9, using data mining algorithm establish fracturing effect with it is main
The model of influence factor；Step 10, fracturing effect quantitative forecast is carried out；Step 11, construction parameter optimization is carried out.
The purpose of the present invention can be also achieved by the following technical measures：
In step 1, the factors for influencing fracturing effect are divided into individual well information, layer data, construction data, pressure break
Five class of effect and pressure break annex.
In step 1, individual well information includes individual well basic data, construction data, effect data；Layer data is individual well
Result of log interpretation, including thickness, permeability, porosity, water saturation, shale content；Construction data includes pressing crack construction
Parameter, including prepad fluid, load fluid, fracture pressure, termination of pumping pressure, sand feeding amount, additive capacity；Fracturing effect includes before pressing and pressure
Contrast on effect afterwards, including yield, the term of validity after yield, pressure before pressure；Pressure break annex includes common fracturing technology and fracturing fluid class
Type annex.
In step 1, classify according to influence factor, while considering database design specification, design table (schedule) 8 is respectively
Pressure break information table, layer data table and substratum summary sheet, construction data table and construction summary sheet, fracturing effect table, construction technology
Annex table, fracturing fluid type annex table.
In step 2, Data Warehouse Design includes data warehouse model selection, the design of pressure break data analysis dimension；Data
Warehouse is designed using star structure, including a true table and multiple dimension tables.
In step 2, true table has recorded the particulars of each affairs, and measurement stores in the table from true table
The particulars of each affairs, these measurements are the levels based on dimension carrying out prefocus；What it is around true table is one group
Dimension table, these dimension tables describe the attribute of each dimension；In order to standardize, more search is created to other attributes in dimension table
Table.
In step 3, the data resource integrated migration and update for including isomeric data interface exploitation, data；Based on pressure break
The present situation that data and data class are various, disperse, completes the combing in pressure break Various types of data source, has formulated adopting for each data source unification
Collection standard and audit system；According to wellfracturing effect statistical correlation index calculating method, is calculated and pressed with source daily output data
Split before effect, including pressure break initial stage daily oil production, pressure break oil increment after daily oil production, pressure break.
In step 4, data mining algorithm includes qualitative analysis and quantitative analysis algorithm；The characteristics of for pressure break data,
Data mining algorithm is selected to include classification, cluster, return three classes.
In steps of 5, Data Analysis Platform includes the displaying function of the storage of data, the analysis of data and data；Data
It excavates general frame and takes the scheme that layering solves to realize, be divided into four layers：First layer is data active layer, contains oil well letter
All data in six breath, formation parameter, well data, construction data, effect, indirect parameter tables；The second layer is data bins
Library layer stores in data warehouse and is carried out reunifying the source data information for defining tissue according to theme and analysis demand, and builds
Vertical flow is automatically updated and is safeguarded to it；Third layer is data mining layer, by building multi-dimensional database and data mining
Structure and model；4th layer is result presentation layer, shows tool using various, in complicated multi-dimensional database and mining model
Useful information, be presented as readily comprehensible chart interface.
In step 6, preliminary data analysis includes decision tree, neural network algorithm.
In step 7, it includes the selection of model, model verification method selection, the optimization of model parameter, model to establish model
Foundation；Algorithm of support vector machine class libraries is developed on the basis of progress conventional algorithm is integrated for pressure break data characteristics,
And it is integrated into data mining platform；The method for selecting kernel function to use cross validation, is tried out one by one, is selected and is concluded error minimum
Kernel function；The parameter of support vector machines includes C (punishment parameter), ε, ν, the value of parameter will have a direct impact on model complexity and
Performance finds optimal parameter combination by the method for cross validation and grid search.
In step 8, sensitivity analysis includes correlation between fracturing effect and each factor, correlative character；Pass through
Fracturing effect and each factor are modeled, by model, analyze the relationship between each factor and fracturing effect.
In step 8, Sensitivity Analysis is：The variable to be analyzed is selected every time, the well for selecting data complete
As sample, the relational expression of situational variables and independent variable is established, then according to the model of foundation, situational variables is investigated and becomes with single
The relationship of amount, i.e. sensitivity analysis.
In step 9, it includes the selection of model, model verification method selection, the optimization of model parameter, model to establish model
Foundation；Algorithm of support vector machine class libraries is developed on the basis of progress conventional algorithm is integrated for pressure break data characteristics,
And it is integrated into data mining platform；The method for selecting kernel function to use cross validation, is tried out one by one, is selected and is concluded error minimum
Kernel function；The parameter of support vector machines includes C (punishment parameter), ε, ν, the value of parameter will have a direct impact on model complexity and
Performance finds optimal parameter combination by the method for cross validation and grid search.
In step 10, quantitative forecast includes calculating fracturing effect.
In a step 11, construction parameter optimization includes the selection of construction parameter, the generation of construction multi-scheme, fracturing effect pair
It is determined than, construction parameter；Using physical property, result of log interpretation as independent variable, changes after different construction parameters carrys out pre- pressure measurement and produce
Oil mass, according to fracturing effect with the changing rule of construction parameter come preferred fracturing parameter.
A kind of Post-frac effect forecasting method based on data mining in the present invention, it is proposed that novel storage of pressure break data
Method.On the basis of conventional pressure break database, it is proposed that it is star-like to establish data mining OLAP CUBE for the concept of data warehouse
Model is designed by Dimensionality optimization, realizes efficient storage and the analysis of pressure break data.The present invention proposes one kind and being suitable for pressure break
The novel architectural framework of data analysis.The cleanings of data, storage, prediction, displaying function are not only realized, storage can also be utilized
The method of the programmings such as process, MDX (Multidimensional Expressions), DMX (data mining expansion plugin) realizes manual when data import
Subregion, batch prediction of analysis data, the output of prediction result comparison.The present invention proposes a kind of for non-linear, more noise pressures
Split the new types of data analysis method of data.Large sample can be directed to and small sample carries out the qualitative analysis of data and quantitative model is built
Vertical, Model Parameter Optimization, model result prediction.By Integration ofTechnology, realize the acquisition of pressure break data, storage, modeling, analysis,
A whole set of flow of prediction.
Description of the drawings
Fig. 1 is the flow chart of a specific embodiment of the Post-frac effect forecasting method based on data mining of the present invention；
Fig. 2 is pressure break database E-R figures in the specific embodiment of the present invention；
Fig. 3 is pressure break data warehouse Star Schema structure chart in the specific embodiment of the present invention；
Fig. 4 is pressure break Data Integration Organization Chart in the specific embodiment of the present invention；
Fig. 5 is data mining architectural framework figure in the specific embodiment of the present invention；
Fig. 6 is decision Tree algorithms network 1 in step 6 example 1 in the specific embodiment of the present invention；
Fig. 7 is decision Tree algorithms network 2 in step 6 example 1 in the specific embodiment of the present invention；
Fig. 8 is decision Tree algorithms network 3 in step 6 example 1 in the specific embodiment of the present invention；
Fig. 9 is decision Tree algorithms network 4 in step 6 example 1 in the specific embodiment of the present invention；
Figure 10 is neural network algorithm design sketch 1 in step 6 example 2 in the specific embodiment of the present invention；
Figure 11 is neural network algorithm design sketch 2 in step 6 example 2 in the specific embodiment of the present invention；
Figure 12 is neural network algorithm design sketch 3 in step 6 example 2 in the specific embodiment of the present invention；
Figure 13 is neural network algorithm design sketch 4 in step 6 example 2 in the specific embodiment of the present invention；
Figure 14 remotely accesses design sketch for data mining results in step 6 example 3 in the specific embodiment of the present invention；
Figure 15 models schematic diagram for data mining algorithm in the specific embodiment of the present invention；
Figure 16 be the present invention a specific embodiment in construction parameter and fracturing effect sensitivity analysis figure；
Figure 17 optimizes schematic diagram for construction parameter in the specific embodiment of the present invention.
Specific implementation mode
For enable the present invention above and other objects, features and advantages be clearer and more comprehensible, it is cited below particularly go out preferable implementation
Example, and coordinate shown in attached drawing, it is described in detail below.
As shown in FIG. 1, FIG. 1 is the structure charts of the Post-frac effect forecasting method based on data mining of the present invention.
Step 1：Fracturing effect influence factor table is designed, pressure break database is established.
The factors for influencing fracturing effect are carried out to be divided into five classes：
(1) individual well information
Individual well information includes mainly individual well basic data, construction data, effect data.
(2) layer data
The result of log interpretation, including thickness, permeability, porosity, water saturation, shale content etc. of main individual well.
(3) construction data
Including fracturing parameter, including prepad fluid, load fluid, fracture pressure, termination of pumping pressure, sand feeding amount, additive capacity
Deng.
(4) fracturing effect
Including the Contrast on effect before pressing and after pressure, including yield, term of validity etc. after yield, pressure before pressure.
(5) pressure break annex
Including common fracturing technology and fracturing fluid type annex.
Classify according to influence factor, while considering database design specification, design table (schedule) 8, be respectively pressure break information table,
Layer data table and substratum summary sheet, construction data table and construction summary sheet, fracturing effect table, construction technology annex table, pressure break
Liquid type annex table.
E-R (entity relationship) figure of database is shown in Fig. 2.
Step 2：Design data storage builds data warehouse.
Data warehouse is designed using star structure, including a true table and multiple dimension tables.The core of this pattern is
True table, the table have recorded the particulars of each affairs.Measurement stores the detailed of each affairs in the table from true table
Thin data.This this measurement is the level based on dimension to carry out (or the part prefocus) of prefocus.What it is around true table is one
Group dimension table, these dimension tables describe the attribute of each dimension.In order to standardize, other attributes in dimension table can be given to create more
Look-up table.
Pressure break data warehouse Star Schema structure chart is shown in Fig. 3.
Step 3：Integral data resource checks the integrality and correctness of data.
Various based on pressure break data and data class, dispersion present situation, completes the combing in pressure break Various types of data source, formulates
The unified acquisition standard of each data source and audit system.
According to wellfracturing effect statistical correlation index calculating method, fracturing effect, packet are calculated with source daily output data
Include before pressure break initial stage daily oil production, pressure break oil increment after daily oil production, pressure break.Write SQL (structured query language) sentences into
Row statistics calculates.
Pressure break Data Integration Organization Chart is shown in Fig. 4.
Step 4：Select suitable data mining algorithm.
Different from conventional continuous analysis data, the characteristics of pressure break data, determines that the difficulty that fracturing effect is analyzed is very
Big, it is mainly reflected in：
1. data discrete is big, numerical value changes in a certain section, is influenced by reservoir property, the ginseng that FRACTURING DESIGN uses
There is also differences for number.Discrete data hardly result in general rule using conventional analysis method.
2. independence is strong between data.They are independent mutually, between there is no mutual hierarchical relationship, analyze between them
Distribution situation is to selecting the effect reached after each pressing crack construction of well to have certain meaning.
It 3. data bulk grade differs greatly, cannot change the distribution of its objective value in analytic process, integrate institute again
There is the distribution for the attribute for participating in analyzing.
The characteristics of for pressure break data, selects data mining algorithm to include classification, cluster, return three classes.
Step 5：Build suitable Data Analysis Platform, mating development environment.
Data mining general frame takes the scheme that layering solves to realize, is broadly divided into four layers：First layer is data source
Layer, contains all data in six oil well information, formation parameter, well data, construction data, effect, indirect parameter tables.
The second layer is data warehouse layer, is stored in data warehouse and is carried out reunifying the source for defining tissue according to theme and analysis demand
Data information, and Establishing process is automatically updated and is safeguarded to it.Third layer is data mining layer, passes through SSAS (SQL
Server Analysis Services) tool builds multi-dimensional database and data mining structure and model.4th layer is result presentation layer, is utilized
It is various to show tool, the useful information in complicated multi-dimensional database and mining model, it is readily comprehensible to be presented as policymaker
Chart interface, and be presented in face of policymaker, provide strong support for decision.
Data mining architectural framework figure is shown in Fig. 5.
Step 6：Multidimensional data displaying and primary data analysis are carried out to data warehouse.
Data are shown and analysis can be divided into the following steps：
(1) show CUBE (data cube) information
Using decomposition tree, in established level, can intuitively be unfolded to check inclusion relation layer by layer.
(2) mining model is checked using AS (Analysis Service) reader
After mining model is trained by data, algorithm will be in data source for the information storage that finds in the data of input
Into mining model, these information can be checked using mining model reader.
(3) mined information is shared by RS (Report Forms Service)
Using the Report Forms Service function of RS, inquires Result using MDX (Multidimensional Expressions) sentence and issues on the net,
Majority's use information is allowed by the way that permission is arranged.
(4) analysis result displaying is realized by C/S (client/server) pattern
ADO.NET may be implemented using ADOMD.NET technologies to connect dedicated for the data adapter unit and data reader of AS
Mouthful, in order to be programmed to data mining client application.
Step 6 example 1 is shown in Fig. 6-Fig. 9.Decision tree network can examine or check the influences of the parameters to yield after pressure such as construction.It is logical
The link strength sliding block for pulling left side is crossed, the reference order for analyzing yield after influencing pressure is：Permeability (Fig. 6), sand amount (figure
7), thickness (Fig. 8), discharge capacity and liquid measure (Fig. 9).
Step 6 example 2 is shown in Figure 10-13.What neural network can quantify examines or check shadow of the different parameter sections to yield
It rings.Figure 10 can examine or check influence of whole variables to yield；Figure 11 and Figure 12 can be seen that single parameter (discharge capacity and sand amount) is right
The influence of yield；Figure 13 can analyze the influence of different displacements and sand amount section to yield simultaneously.
Step 6 example 3 is shown in Figure 14.Data mining results can be accessed by client.
Step 7：The model of fracturing effect and pressure break influence factor is established using data mining algorithm.
Algorithm of support vector machine class libraries is developed on the basis of progress conventional algorithm is integrated for pressure break data characteristics,
And it is integrated into data mining platform.
Algorithm of support vector machine solves the problems, such as that the key of Nonlinear Classification is exactly to select appropriate kernel function, Selection of kernel function
It is whether appropriate, it will directly affect the performance of grader.The method for usually selecting kernel function to use cross validation in practice, by
It is a on probation, select the kernel function for concluding error minimum.The parameter of support vector machines mainly has C (punishment parameter), ε, ν etc., parameter
Value will have a direct impact on the complexity and performance of model, optimal ginseng is found by the method for cross validation and grid search
Array is closed.
Data mining algorithm modeling schematic diagram is shown in Figure 15.
Step 8：The sensitivity analysis of fracturing effect and each influence factor determines the principal element for influencing fracturing effect.
The factor for influencing fracturing effect is various, and the relationship of these factors and fracturing effect is all nonlinear.It is logical
It crosses and models fracturing effect and each factor, by model, the relationship between each factor and fracturing effect can be analyzed.
Analysis method：The variable to be analyzed is selected every time, and such as initial stage liquid measure, the well for selecting data complete is as sample
This, establishes the relational expression of situational variables and independent variable, then according to the model of foundation, investigates the pass of situational variables and single variable
System, i.e. sensitivity analysis.
The sensitivity analysis schematic diagram of sand amount and fracturing effect is shown in Figure 16.Sand amount increases, and effect improves, and oil mass is first steady,
Then increase.
Step 9：Fracturing effect and the model of major influence factors are established using data mining algorithm.
It is modeled using the method for step 7.
Step 10：Carry out fracturing effect quantitative forecast.
Effect prediction after the model established using step 10 is pressed.
Step 11：Carry out construction parameter optimization.
The optimization method of use changes different construction parameters to predict using physical property, result of log interpretation as independent variable
Oil production after pressure, according to fracturing effect with the changing rule of construction parameter come preferred fracturing parameter.Such as other parameters not
Under conditions of change, different displacements value is set, different modeling schemes is generated, to compare construction effect.
Construction parameter optimization schematic diagram is shown in Figure 17.As discharge capacity increases, yield increases, rational discharge capacity section 5~
5.5m3/min.As sand amount increases, yield first increases, when sand feeding amount reaches 36m3, yield starts to reduce, and rational sand amount is 36
~40m3。
Claims (16)
1. the Post-frac effect forecasting method based on data mining, which is characterized in that should the Post-frac effect forecasting based on data mining
Method includes：
Step 1, fracturing effect influence factor table is designed, pressure break database is established；
Step 2, design data storage builds data warehouse；
Step 3, integral data resource checks the integrality and correctness of data；
Step 4, suitable data mining algorithm, organized data algorithm class libraries are selected；
Step 5, suitable Data Analysis Platform, mating development environment are built；
Step 6, multidimensional data displaying and primary data analysis are carried out to data warehouse；
Step 7, the model of fracturing effect and pressure break influence factor is established using data mining algorithm；
Step 8, the sensitivity analysis of fracturing effect and each influence factor is carried out, determines the principal element for influencing fracturing effect；
Step 9, fracturing effect and the model of major influence factors are established using data mining algorithm；
Step 10, fracturing effect quantitative forecast is carried out；
Step 11, construction parameter optimization is carried out.
2. the Post-frac effect forecasting method according to claim 1 based on data mining, which is characterized in that in step 1,
The factors for influencing fracturing effect are divided into individual well information, layer data, construction data, fracturing effect and pressure break annex five
Class.
3. the Post-frac effect forecasting method according to claim 2 based on data mining, which is characterized in that in step 1,
Individual well information includes individual well basic data, construction data, effect data；Layer data is the result of log interpretation of individual well, including
Thickness, permeability, porosity, water saturation, shale content；Construction data includes fracturing parameter, including prepad fluid, is taken
Sand liquid, fracture pressure, termination of pumping pressure, sand feeding amount, additive capacity；Fracturing effect includes the preceding Contrast on effect with after pressure of pressure, including
Yield, the term of validity after pressure preceding yield, pressure；Pressure break annex includes common fracturing technology and fracturing fluid type annex.
4. the Post-frac effect forecasting method according to claim 3 based on data mining, which is characterized in that in step 1,
Classify according to influence factor, while considering database design specification, design table (schedule) 8 is pressure break information table, layer data respectively
Table and substratum summary sheet, construction data table and construction summary sheet, fracturing effect table, construction technology annex table, fracturing fluid type are attached
Record table.
5. the Post-frac effect forecasting method according to claim 1 based on data mining, which is characterized in that in step 2,
Data Warehouse Design includes data warehouse model selection, the design of pressure break data analysis dimension；Data warehouse is set using star structure
Meter, including a true table and multiple dimension tables.
6. the Post-frac effect forecasting method according to claim 5 based on data mining, which is characterized in that in step 2,
True table has recorded the particulars of each affairs, and measurement stores the detailed money of each affairs in the table from true table
Material, these measurements are the levels based on dimension to carry out prefocus；What it is around true table is one group of dimension table, these dimension tables describe
The attribute each tieed up；In order to standardize, more look-up tables are created to other attributes in dimension table.
7. the Post-frac effect forecasting method according to claim 1 based on data mining, which is characterized in that in step 3,
The data resource integrated migration and update for including isomeric data interface exploitation, data；Various based on pressure break data and data class,
The present situation of dispersion completes the combing in pressure break Various types of data source, has formulated the unified acquisition standard of each data source and audit system；It presses
According to wellfracturing effect statistical correlation index calculating method, fracturing effect, including the pressure break day before yesterday are calculated with source daily output data
Initial stage daily oil production, pressure break oil increment after oil production, pressure break.
8. the Post-frac effect forecasting method according to claim 1 based on data mining, which is characterized in that in step 4,
Data mining algorithm includes qualitative analysis and quantitative analysis algorithm；The characteristics of for pressure break data, selects data mining algorithm packet
It includes classification, cluster, return three classes.
9. the Post-frac effect forecasting method according to claim 1 based on data mining, which is characterized in that in steps of 5,
Data Analysis Platform includes the displaying function of the storage of data, the analysis of data and data；Data mining general frame is taken point
The scheme that layer solves is realized, is divided into four layers：First layer is data active layer, contains oil well information, formation parameter, oil well number
According to, all data in six construction data, effect, indirect parameter tables；The second layer is data warehouse layer, is stored in data warehouse
Reunify the source data information for defining tissue according to theme and analysis demand, and Establishing process it is carried out it is automatic more
New and maintenance；Third layer is data mining layer, by building multi-dimensional database and data mining structure and model；4th layer is knot
Fruit represent layer shows tool using various, the useful information in complicated multi-dimensional database and mining model, is presented as being easy
The chart interface of understanding.
10. the Post-frac effect forecasting method according to claim 1 based on data mining, which is characterized in that in step 6
In, preliminary data analysis includes decision tree, neural network algorithm.
11. the Post-frac effect forecasting method according to claim 1 based on data mining, which is characterized in that in step 7
In, it includes the selection of model, model verification method selection, the optimization of model parameter, the foundation of model to establish model；For pressure break
Data characteristics develop algorithm of support vector machine class libraries, and be integrated into data mining on the basis of progress conventional algorithm is integrated
Platform；The method for selecting kernel function to use cross validation, is tried out one by one, selects the kernel function for concluding error minimum；Supporting vector
The parameter of machine includes punishment parameter C, ε, ν, and the value of parameter will have a direct impact on the complexity and performance of model, pass through cross validation
Optimal parameter combination is found with the method for grid search.
12. the Post-frac effect forecasting method according to claim 1 based on data mining, which is characterized in that in step 8
In, sensitivity analysis includes correlation between fracturing effect and each factor, correlative character；By by fracturing effect with it is each because
Element is modeled, and by model, analyzes the relationship between each factor and fracturing effect.
13. the Post-frac effect forecasting method according to claim 1 based on data mining, which is characterized in that in step 8
In, Sensitivity Analysis is：The variable to be analyzed is selected every time, and the well for selecting data complete is established and divided as sample
It analyses the relational expression of variable and independent variable, then according to the model of foundation, investigates the relationship of situational variables and single variable, i.e., it is sensitive
Property analysis.
14. the Post-frac effect forecasting method according to claim 1 based on data mining, which is characterized in that in step 9
In, it includes the selection of model, model verification method selection, the optimization of model parameter, the foundation of model to establish model；For pressure break
Data characteristics develop algorithm of support vector machine class libraries, and be integrated into data mining on the basis of progress conventional algorithm is integrated
Platform；The method for selecting kernel function to use cross validation, is tried out one by one, selects the kernel function for concluding error minimum；Supporting vector
The parameter of machine includes punishment parameter C, ε, ν, and the value of parameter will have a direct impact on the complexity and performance of model, pass through cross validation
Optimal parameter combination is found with the method for grid search.
15. the Post-frac effect forecasting method according to claim 1 based on data mining, which is characterized in that in step 10
In, quantitative forecast includes calculating fracturing effect.
16. the Post-frac effect forecasting method according to claim 1 based on data mining, which is characterized in that in step 11
In, construction parameter optimization includes the selection of construction parameter, construction multi-scheme generates, fracturing effect compares, construction parameter determines；It will
Physical property, result of log interpretation change different construction parameters and carry out oil production after pre- pressure measurement as independent variable, according to fracturing effect with
The changing rule of construction parameter carrys out preferred fracturing parameter.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710137665.0A CN108573078B (en) | 2017-03-09 | 2017-03-09 | Fracturing effect prediction method based on data mining |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710137665.0A CN108573078B (en) | 2017-03-09 | 2017-03-09 | Fracturing effect prediction method based on data mining |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108573078A true CN108573078A (en) | 2018-09-25 |
| CN108573078B CN108573078B (en) | 2021-11-26 |
Family
ID=63577840
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710137665.0A Active CN108573078B (en) | 2017-03-09 | 2017-03-09 | Fracturing effect prediction method based on data mining |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108573078B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110264014A (en) * | 2019-06-27 | 2019-09-20 | 北京中油瑞飞信息技术有限责任公司 | A kind of method and device for predicting old well oil production |
| CN110965977A (en) * | 2019-11-20 | 2020-04-07 | 中国石油大学(北京) | Fracturing construction analysis method |
| CN111324657A (en) * | 2020-02-12 | 2020-06-23 | 广州奥格智能科技有限公司 | Emergency plan content optimization method and computer equipment |
| CN111581771A (en) * | 2020-03-30 | 2020-08-25 | 无锡融合大数据创新中心有限公司 | Stamping workpiece cracking prediction platform based on artificial intelligence technology |
| CN112528358A (en) * | 2020-01-14 | 2021-03-19 | 枣庄科技职业学院 | Method for obtaining wall paint spraying effect by utilizing big data |
| CN113803042A (en) * | 2020-06-12 | 2021-12-17 | 中国石油化工股份有限公司 | Single-section single-cluster dense fracturing method and system |
| CN115310357A (en) * | 2022-08-09 | 2022-11-08 | 大庆正方软件科技股份有限公司 | Fracturing analysis method based on data-driven decision |
| CN117572531A (en) * | 2024-01-16 | 2024-02-20 | 电子科技大学 | Intelligent detector embedding quality testing method and system |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104018831A (en) * | 2014-06-24 | 2014-09-03 | 西南石油大学 | Method for evaluating reservoir of fractured well |
| WO2014146004A2 (en) * | 2013-03-15 | 2014-09-18 | Intelligent Solutions, Inc. | Data-driven analytics, predictive modeling & optimization of hydraulic fracturing in marcellus shale |
-
2017
- 2017-03-09 CN CN201710137665.0A patent/CN108573078B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2014146004A2 (en) * | 2013-03-15 | 2014-09-18 | Intelligent Solutions, Inc. | Data-driven analytics, predictive modeling & optimization of hydraulic fracturing in marcellus shale |
| CN104018831A (en) * | 2014-06-24 | 2014-09-03 | 西南石油大学 | Method for evaluating reservoir of fractured well |
Non-Patent Citations (2)
| Title |
|---|
| 孟庆民 等: "数据挖掘技术在压裂决策中的应用研究", 《数字石油和化工》 * |
| 孟庆民 等: "浊积岩储层压裂效果数据挖掘研究", 《石油天然气学报（江汉石油学院学报）》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110264014A (en) * | 2019-06-27 | 2019-09-20 | 北京中油瑞飞信息技术有限责任公司 | A kind of method and device for predicting old well oil production |
| CN110264014B (en) * | 2019-06-27 | 2020-04-03 | 北京中油瑞飞信息技术有限责任公司 | Method and device for predicting oil production of old well |
| CN110965977A (en) * | 2019-11-20 | 2020-04-07 | 中国石油大学(北京) | Fracturing construction analysis method |
| CN112528358A (en) * | 2020-01-14 | 2021-03-19 | 枣庄科技职业学院 | Method for obtaining wall paint spraying effect by utilizing big data |
| CN111324657A (en) * | 2020-02-12 | 2020-06-23 | 广州奥格智能科技有限公司 | Emergency plan content optimization method and computer equipment |
| CN111324657B (en) * | 2020-02-12 | 2023-09-08 | 奥格科技股份有限公司 | Emergency plan content optimization method and computer equipment |
| CN111581771A (en) * | 2020-03-30 | 2020-08-25 | 无锡融合大数据创新中心有限公司 | Stamping workpiece cracking prediction platform based on artificial intelligence technology |
| CN113803042A (en) * | 2020-06-12 | 2021-12-17 | 中国石油化工股份有限公司 | Single-section single-cluster dense fracturing method and system |
| CN115310357A (en) * | 2022-08-09 | 2022-11-08 | 大庆正方软件科技股份有限公司 | Fracturing analysis method based on data-driven decision |
| CN117572531A (en) * | 2024-01-16 | 2024-02-20 | 电子科技大学 | Intelligent detector embedding quality testing method and system |
| CN117572531B (en) * | 2024-01-16 | 2024-03-26 | 电子科技大学 | Intelligent detector embedding quality testing method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108573078B (en) | 2021-11-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108573078A (en) | Post-frac effect forecasting method based on data mining | |
| RU2496972C2 (en) | Device, method and system of stochastic investigation of formation at oil-field operations | |
| CA2640725C (en) | Methods, systems, and computer-readable media for fast updating of oil and gas field production models with physical and proxy simulators | |
| US20190251460A1 (en) | Method for predicting oil and gas reservoir production | |
| CN110533163A (en) | Sand plugging after fracturing method for early warning based on long Memory Neural Networks in short-term | |
| CN104881435A (en) | Data mining based automatic research flow well logging evaluation expert system | |
| CN112308269A (en) | Yield prediction method and device for low-permeability oil and gas reservoir | |
| US11481413B2 (en) | Systems and methods for evaluating petroleum data for automated processes | |
| Alenezi et al. | A data-driven smart proxy model for a comprehensive reservoir simulation | |
| CN114117881A (en) | Sand production risk prediction method and system | |
| Salehian et al. | Robust integrated optimization of well placement and control under field production constraints | |
| Wang et al. | Big data technique in the reservoir parameters’ prediction and productivity evaluation: A field case in western South China sea | |
| CN110717270B (en) | Oil and gas reservoir simulation method based on data | |
| Hongliang et al. | Management and instant query of distributed oil and gas production dynamic data | |
| Sankaran et al. | Data Analytics in Reservoir Engineering | |
| CN108843296B (en) | Single-well repeated fracturing effect prediction method based on multi-factor influence | |
| Zhou et al. | A CNN-BiGRU-AM neural network for AI applications in shale oil production prediction | |
| CN110489844A (en) | One kind being suitable for the uneven large deformation grade prediction technique of soft rock tunnel | |
| Mutalova et al. | Machine learning on field data for hydraulic fracturing design optimization | |
| Mohaghegh | A new methodology for the identification of best practices in the oil and gas industry, using intelligent systems | |
| CN113537706A (en) | Oil field production increasing measure optimization method based on intelligent integration | |
| Simanjuntak et al. | Applying artificial neural network and xgboost to improve data analytics in oil and gas industry | |
| Liu et al. | Optimizing construction parameters for fractured horizontal wells in shale oil | |
| Likhoded et al. | Technology for rapid well ranking and screening for pumping optimization with regard to subsurface risks and economics. A case study of a giant oil brownfield | |
| Feder | Prescriptive Analytics Aids Completion Optimization in Unconventionals |
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 |