CN106869990A - Coal gas Permeability Prediction method based on LVQ CPSO BP algorithms - Google Patents
Coal gas Permeability Prediction method based on LVQ CPSO BP algorithms Download PDFInfo
- Publication number
- CN106869990A CN106869990A CN201710121485.3A CN201710121485A CN106869990A CN 106869990 A CN106869990 A CN 106869990A CN 201710121485 A CN201710121485 A CN 201710121485A CN 106869990 A CN106869990 A CN 106869990A
- Authority
- CN
- China
- Prior art keywords
- lvq
- cpso
- layer
- sample
- effective stress
- 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
- 230000035699 permeability Effects 0.000 title claims abstract description 82
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 75
- 239000003034 coal gas Substances 0.000 title claims abstract description 69
- 238000000034 method Methods 0.000 title claims abstract description 30
- 238000013528 artificial neural network Methods 0.000 claims abstract description 41
- 239000003245 coal Substances 0.000 claims abstract description 40
- 238000012549 training Methods 0.000 claims abstract description 24
- 238000005457 optimization Methods 0.000 claims abstract description 21
- 239000002245 particle Substances 0.000 claims abstract description 18
- 239000013598 vector Substances 0.000 claims abstract description 16
- 230000007935 neutral effect Effects 0.000 claims abstract description 8
- 238000007906 compression Methods 0.000 claims description 21
- 230000006835 compression Effects 0.000 claims description 18
- 210000002569 neuron Anatomy 0.000 claims description 16
- 230000008859 change Effects 0.000 claims description 12
- 230000006870 function Effects 0.000 claims description 9
- 241001269238 Data Species 0.000 claims description 7
- 238000011156 evaluation Methods 0.000 claims description 7
- 239000011159 matrix material Substances 0.000 claims description 6
- 230000000694 effects Effects 0.000 claims description 4
- 230000000739 chaotic effect Effects 0.000 claims description 3
- 239000006185 dispersion Substances 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 241000894007 species Species 0.000 claims description 3
- 238000001764 infiltration Methods 0.000 claims description 2
- 230000008595 infiltration Effects 0.000 claims description 2
- 238000009738 saturating Methods 0.000 claims 1
- 210000003127 knee Anatomy 0.000 abstract description 5
- 238000013139 quantization Methods 0.000 abstract description 4
- 238000010835 comparative analysis Methods 0.000 abstract description 2
- 238000005065 mining Methods 0.000 description 5
- 238000012360 testing method Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 239000011148 porous material Substances 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- VNWKTOKETHGBQD-UHFFFAOYSA-N methane Chemical compound C VNWKTOKETHGBQD-UHFFFAOYSA-N 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 244000287680 Garcinia dulcis Species 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000000205 computational method Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 230000002687 intercalation Effects 0.000 description 1
- 238000009830 intercalation Methods 0.000 description 1
- 238000004452 microanalysis Methods 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 210000004205 output neuron Anatomy 0.000 description 1
- 230000002028 premature Effects 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 239000011435 rock Substances 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
Classifications
-
- E—FIXED CONSTRUCTIONS
- E21—EARTH OR ROCK DRILLING; MINING
- E21F—SAFETY DEVICES, TRANSPORT, FILLING-UP, RESCUE, VENTILATION, OR DRAINING IN OR OF MINES OR TUNNELS
- E21F7/00—Methods or devices for drawing- off gases with or without subsequent use of the gas for any purpose
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- Mining & Mineral Resources (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Computer Hardware Design (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Geology (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Geometry (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Geochemistry & Mineralogy (AREA)
- Software Systems (AREA)
- General Life Sciences & Earth Sciences (AREA)
Abstract
Coal gas Permeability Prediction method based on LVQ CPSO BP algorithms, proposes the LVQ CPSO BP coal gas Permeability Prediction methods predicted based on learning vector quantization neutral net (LVQ) classification, Chaos particle swarm optimization algorithm (CPSO) optimization, BP neural network.Determine that coal seam buried depth is divided into two-layer by critical value；Based on there is flex point relation between effective stress and gas permeation rate, determine that effective stress is divided into two sections by knee value；4 microcosmic sample parameters are carried out by Classification and Identification according to corner feature using LVQ, learning training is carried out using BP neural network and output is predicted the outcome, and the weights and threshold value of BP neural network are optimized with CPSO；Finally, the checking that predicts the outcome has been carried out to the LVQ CPSO BP algorithms that the present invention builds based on sample cases, and the result predicted with BP algorithm, GA BP algorithms and PSO BP algorithms has carried out comparative analysis.
Description
Technical field
The present invention relates to coal gas Permeability Prediction field, the more particularly to coal body watt based on LVQ-CPSO-BP algorithms
This Permeability Prediction method.
Background technology
Coal mine gas disaster is one of major casualty in process of coal mining, is influenceed by geological conditions so that mine
Gas is more in coal seam to be moved with instability status seepage flow.Because gas permeation rate is residing with crustal stress, temperature and gas pressure etc.
Earth time spatial character it is closely bound up, the features such as cause gas permeation rate to show time variation, non-linear, ambiguity, such as
The dynamic change of what accurate prediction coal body gas permeation rate, has important meaning to coal mine gas disaster in prevention mining
Justice.
Domestic and international researcher is studied for the change of coal gas permeability, and intercalation life of Lv etc. have studied different knots
The otherness of coal gas permeability under structure, discloses influence of the Coal Pore Structure otherness to gas permeation rate；Wei builds equality and grinds
The Permeability factor of coal containing methane gas under moisture and loaded condition is studied carefully；Yin Guangzhi, Jiang Changbao etc. have studied gas pressure,
The relation of stress and permeability；Wang Dengke etc. have studied the seepage characteristic of gas, draw the new computational methods of permeability；Xu Jiang
Etc. the influence that have studied in the environment of temperature, low pore pressure and different gas pressure to coal gas permeability characteristics；
Perera etc. have studied permeability of the coal gas gas under the conditions of non-zero transverse strain；Zhaoyang's liter etc. have studied triaxiality
Lower suction-operated is to coal and rock laws of gas flow.Due in recovery process coal gas permeability it is uncertain and fuzzy
Property feature, domestic and foreign scholars set up theoretical model and empirical equation be not suitable for coal gas Permeability Prediction.Therefore,
Forecasting Methodology based on intelligent algorithm starts to be applied, wherein BP neural network (back propagation neural
Network) it is one of main method for coal gas Permeability Prediction, mine safety prediction is solved to a certain extent and is asked
Topic.But the single predictive model algorithm based on BP neural network can only be answered for particular problem, therefore to combine
Algorithm is that the hybrid prediction model research of main framework is developed rapidly, that is, consider the advantage and disadvantage of Individual forecast model and be applicable
Property, by building intelligent optimization algorithm combination analysis of neural network coal gas permeability Variation Features at different conditions,
Double combinational algorithms that a kind of optimized algorithm is combined with neutral net are set up, precision of prediction is improve, however, obtaining higher precision
Coal gas Permeability Prediction algorithmically needs more combinations and improves.
In sum, the present invention is proposed using learning vector quantization neural network algorithm (Learning Vector
Quantization, i.e. LVQ) classification, Chaos particle swarm optimization algorithm (the Chaos Particle of chaos and population combination producing
Swarm Optimization, CPSO) optimization, BP neural network prediction many combinational algorithms coal gas permeability is carried out it is pre-
Survey, by coal seam buried depth macro-laminate, the microcosmic segmentation of effective stress, weights and threshold value parallel optimization, sample data learning training,
Test samples prediction checking, and with the single algorithms of BP, heredity and BP pair combinational algorithm (Genetic AlgorithmGA-BP), grains
The result of subgroup and double combinational algorithm (Particle Swarm Optimization, the PSO-BP) predictions of BP is analyzed,
Fully demonstrate many combinational algorithms of LVQ-CPSO-BP feasible, effective in terms of coal gas Permeability Prediction precision raising.
The content of the invention
To solve the problems, such as above-mentioned prior art, it is an object of the invention to provide based on LVQ-CPSO-BP algorithms
Coal gas Permeability Prediction method.Macro-forecast value is high with the measured value goodness of fit, especially when effective stress reduces, prediction
Precision is higher.
To reach above-mentioned purpose, the technical scheme is that：
Coal gas Permeability Prediction method based on LVQ-CPSO-BP algorithms, is made up of following steps：
Step one, coal seam buried depth are divided and selected：Coal seam buried depth is studied：Coal seam buried depth be divided into shallow embedding layer and it is non-shallow
Buried regions, and shallow embedding layer is general in more than 150m, in gas weathered zone, the present invention is not discussed；Rather than shallow embedding layer refers to 150m
Hereinafter, most as deep as 1500m or so is arrived, most of coal seam is between 200m-700m, therefore the present invention ignores shallow embedding layer, will be non-
Shallow embedding layer it is self-defined is divided into two-layer, if critical value be 700m, with it is shallow be middle buried regions, be buried layer with depth；
Step 2, major influence factors selection：Selection effective stress, temperature, gas pressure, compression strength are used as main shadow
The factor of sound；
There is flex point in relation between effective stress and coal gas permeability, when effective stress is less than flex point, coal body watt
This permeability variation is very fast, and during more than flex point, coal gas permeability variation is gentle, and 2MPa is set into effective stress flex point；
Step 3, on the basis of coal seam buried depth non-shallow embedding layer macroscopic view is divided, using LVQ algorithms to effective stress sample data
Microscopic classification is carried out by flex point；
The coal gas Permeability Prediction submodel of step 4, foundation based on CPSO-BP algorithms；
Step 5, the training by LVQ-CPSO-BP forecasting systems, obtain coal gas Permeability Prediction result.
Further, in order to evaluate the confidence level and accuracy of set up forecast model, error and precision are quantified,
Coal gas Permeability Prediction result qualitative assessment is carried out using formula (1)-formula (4) predicated error and precision evaluation index：
1) the reliability average relative error of predicated error:(mean relative error,MRE)：
2) size of predicted value deviation true value error is with mean absolute error (mean absolute error, MAE)：
3) sample data dispersion degree is with root-mean-square error (root mean square error, RMSE)：
4) forecast model accuracy (forecast model accuracy, FMA)：
Wherein：XiIt is actual value, Xi' it is and predicted value that M is sample size.
Further, in the step 3, LVQ is used to be classified as coal gas permeability by boundary of effective stress flex point
Change high order smooth pattern and precipitous type two types, average value of the effective stress as (1-5) MPa with sample dataTemperature-averaging
ValueGas pressure average valueCompression strength average valueAs the input of LVQ, by after the training of LVQ Classification and Identifications, obtaining
Obtain two class classification samples data.
Further, in the step 3, coal gas Permeability Prediction learning sample vector represents that sample set is with X：
[X]=[x1,x2,x3,x4]T=[σ, T, P, σm]T (5)
Wherein σ is effective stress (MPa), T is temperature (DEG C), P is gas pressure (MPa), σmIt is compression strength (MPa)；
Input layer input set is：
The sample of competition layer is expressed as with Y：
Y=(y1, y2... ym)T, yi∈ { 0,1 }, i=1,2 ..., m (7)
Weight matrix W between input layer and competition layer1：
Competition layer is to the weight matrix W between output layer2：
The desired value output of network is expressed as with d：
D=(d1, d2... dl)T (10)
The output of output layer represents that output collection is with O：
[O]=[o1, o2]T=[1,0]T (11)
Wherein, o1It is precipitous type classification, o2It is high order smooth pattern classification.
Further, in the step 4, after sample data to be divided into LVQ two major classes, predicted using CPSO-BP,
CPSO algorithms map chaos system using Logistic, and iterative formula is：
zi+1=μ zi(1-zi), i=0,1 ..., μ ∈ (2,4] (12)
In formula：μ is control variables；
When μ=4,0≤z0When≤1, Logistic is completely in chaos state, and chaotic motion characteristic during using μ=4 changes
Enter standard particle group's algorithm, i.e.,
zi+1=4zi(1-zi), i=0,1 ... (13)
Population initial phase produces excellent initial population as primary group by the use of formula (13), to mitigate
Influence of the random initializtion population to algorithm optimization performance.
Further, in the step 4, the number of plies of BP neural network structure is set to 3 layers, if R, S1, S2 are respectively defeated
Enter layer, hidden layer, output layer neuron number, input layer therein is the sorted sample datas of LVQ, and hidden layer is
S function, most preferably implicit nodes are 3, and output layer is coal gas permeability.The weights and threshold value and particle of BP neural network
Shown in mapping relations such as formula (14) between the dimension N of the individual position vector P of group：
N=S1 × R+S1 × S2+S1+S2 (14)
Further, in the step 3, by influence factor screen, with effective stress, temperature, gas pressure, resistance to compression
The corresponding data of intensity and permeability is learnt for sample；
BP, GA-BP, PSO-BP, CPSO-BP have been write using Matlab2015 Neural Network Toolbox combination sample datas
Four kinds of combination Permeability Prediction model codes, setting BP neural network iterations 1500 times, learning rate 0.01 is minimum equal
Square error 10-5；The particle of CPSO optimization neural networks takes W=0.7298, c1=c2=1.4962, the upper dividing value of particle is 5,
Floor value is -5.
Further, in the step 5, by sample data input LVQ it is trained, netinit is carried out first：If
Fixed smaller random number is used as initial weight W1And W2；It is determined that initial learning efficiency η (0)=0.9 and frequency of training tm=100 and point
Class species number K=2.It is four factors, 40 groups of average vectors of sample vector in table 1 to be input into.The god that wins is found by the way that desired value is minimum
Through unit.It is no be divided into correct classification of the neuron less than the effective stress average value (about 2.5MPa) that the present invention is specified of winning
Weights when then adjustment is classified, renewal learning efficiency, until t=tm, obtain classification results.
Further, in the step 5, using effective stress, temperature, gas pressure, compression strength average value as defeated
Enter sample, it is stipulated that the output type on the critical value left side and the right with effective stress average value as flex point is followed successively by 1,0, is divided
Class recognition result；
Sample data input LVQ is carried out into sorted two groups of sample datas input CPSO-BP to be trained, first initially
Change BP neural network 3-tier architecture, input layer is 4：Effective stress, temperature, gas pressure, compression strength；Hidden layer god
It is 5 through unit, output layer neuron is 1：Coal gas permeability；By chaos intialization Population Size, dimension D and particle S speed
, from optimizing, the optimal solution that will be searched out is assigned to the weights and threshold of BP networks for degree and position and mean square error population fitness function
Value, continues the learning training prediction of BP networks.
Further, the present invention has carried out 40 groups of learning trainings of sample data, forecast model using four kinds of forecast models
One predicts for BP neural network, and forecast model two is GA-BP neural network predictions, and forecast model three is that PSO-BP neutral nets are pre-
Survey, forecast model four is CPSO-BP；Input data in forecast model is the sorted data of LVQ, wherein Class1 training
Sample is 26 groups；The training sample of type 0 is 14 groups.
Further, effective stress flex point is divided using LVQ models has precision higher, and Classification and Identification accuracy is most
It is high by reachable 96.15%.
Further, four kinds of coal gas permeability are predicted the outcome according to the error evaluation method set up true with experiment
Real-valued difference carries out error calculation, and the coal gas Permeability Prediction precision of LVQ-CPSO-BP algorithms is 95.26%, better than it
His three kinds of models, the LVQ-CPSO-BP forecast models set up are effective, adaptability is stronger, precision of prediction is higher.
Relative to prior art, beneficial effects of the present invention are：
Coal gas Permeability Prediction method of the present invention based on LVQ-CPSO-BP algorithms, from macroscopic view and microcosmic combination angle
Degree, filter out influence precision of prediction 5 principal element -1 macroscopic views (coal seam buried depth) and 4 it is microcosmic (effective stress, temperature,
Gas pressure, compression strength), propose a kind of based on learning vector quantization neutral net (LVQ) classification, Chaos particle swarm optimization algorithm
(CPSO) optimization, the LVQ-CPSO-BP coal gas Permeability Prediction methods of BP neural network prediction.First, based on China's coal
The statistics of layer buried depth value, two-layer is divided into from macroscopically determination critical value by coal seam buried depth；Based on effective stress and gas
There is flex point relation between permeability, effective stress is divided into two sections from microcosmic upper determination knee value；Then, using LVQ by 4
Individual microcosmic sample parameter carries out Classification and Identification according to corner feature, carries out learning training using BP neural network and exports prediction knot
Really, and with CPSO the weights and threshold value of BP neural network are optimized；Finally, the present invention is built based on sample cases
LVQ-CPSO-BP algorithms have carried out the checking that predicts the outcome, and the result predicted with BP algorithm, GA-BP algorithms and PSO-BP algorithms
Comparative analysis is carried out.Result shows：LVQ classification correct recognition ratas are higher, and CPSO-BP algorithms precision of prediction preferably, and is better than
Other three kinds of algorithms.LVQ-CPSO-BP algorithm macro-forecast values are high with the measured value goodness of fit, especially when effective stress reduces,
Precision of prediction is higher.
Brief description of the drawings
Fig. 1 be permeability with effective stress variation diagram, wherein, (a) flex point be 1.5MPa, (b) flex point be 3MPa, (c) is turned
Point is 2MPa.
Fig. 2 is coal gas Permeability Prediction general frame block diagram.
Fig. 3 LVQ network structures.
Fig. 4 CPSO-BP algorithm flow block diagrams.
The classification of Fig. 5 effective stresses flex point predicts the outcome figure, wherein：A () Class1 (b) type 0 that predicts the outcome predicts the outcome.
Specific embodiment
Technical solution of the present invention is described in further detail with reference to the accompanying drawings and detailed description：
Experimental example：
As shown in figure 1,
Coal gas Permeability Prediction method based on LVQ-CPSO-BP algorithms, is made up of following steps：
Major influence factors are screened and layering critical value determines with stress knee value
Coal gas permeability is closely bound up with Geologic Time spatial character, with the change of process of coal mining, presents
The features such as going out time variation, non-linear, ambiguity, shows dynamic change characterization.Its influence factor is macroscopically mainly reflected in ground
The aspects such as texture is made, geology coal bed texture, seam mining depth, Coal Characteristics, it is microcosmic on be mainly reflected in effective stress, temperature
The aspects such as degree, gas pressure, compression strength, pore structure, ignore the relation between factors above and permeability, then can not be true
Reflection Inherent relation and regularity, comprehensively consider factors above can cause again various parameter, cross jamming, model complexity, be difficult to it is pre-
Survey.Therefore influence factor is divided and is screened and classified, will be the effective hand for improving coal gas Permeability Prediction precision
Section.
In terms of macroscopic view, knowable to the Research Literature delivered：Buried depth change in coal seam will produce different crustal stress, to watt
The influence of this permeability is larger, and geological structure, Coal Characteristics relative effect are weaker, therefore the present invention only grinds to coal seam buried depth
Study carefully：Coal seam buried depth is divided into shallow embedding layer and non-shallow embedding layer, and shallow embedding layer it is general in 150m with shallow, in gas weathered zone, the present invention
It is not discussed.Rather than shallow embedding layer refers to 150m with depth, most as deep as arriving 1500m or so, most of coal seam be in 200m-700m it
Between, thus the present invention ignore shallow embedding layer, by non-shallow embedding layer it is self-defined be divided into two-layer, if critical value be 700m, with it is shallow be middle buried regions,
It is buried layer with depth.
At microcosmic aspect, knowable to the Research Literature delivered：Effective stress, temperature, gas pressure, compression strength, hole
Gap structure influence is larger, but document is pointed out, hole changes with STRESS VARIATION, therefore selection effective stress, temperature, gas pressure
Power, compression strength are used as major influence factors.Effective stress therein has spy as shown in Figure 1 according to the test data of document
Levy：There is flex point in relation between effective stress and coal gas permeability, when effective stress is less than flex point, coal gas infiltration
Rate change is very fast, and during more than flex point, coal gas permeability variation is gentle.Because experimental condition is different, there is the position of flex point
Also different, flex point is 1.5MPa in Fig. 1 (a), and flex point is 3MPa in Fig. 1 (b), and in order to embody versatility, the present invention is proposed two
The mean value definition of the effective stress flex point in group test data is flex point, for the flex point of 2MPa has fully demonstrated this in Fig. 1 (c)
Invention defines the reasonability of knee value.
LVQ-CPSO-BP forecasting system frameworks
LVQ has the function that Classification and Identification is carried out to large sample amount, and CPSO has the power to uncertain and ambiguity amount
Value and threshold value carry out the function of intelligent optimization, and BP neural network has carries out self study forecast function to random and amount of nonlinearity.
Therefore, on the basis of buried depth non-shallow embedding layer macroscopic view in coal seam is divided, effective stress sample data is pressed using LVQ algorithms
Flex point carries out microscopic classification, sets up the coal gas Permeability Prediction submodel based on CPSO-BP algorithms.By LVQ-CPSO-
The training of BP forecasting systems, obtains coal gas Permeability Prediction result.
Precision of prediction appraisal procedure
In order to evaluate the confidence level and accuracy of set up forecast model, error and precision are quantified, using formula
(1)-formula (4) predicated error and precision evaluation index carry out coal gas Permeability Prediction result qualitative assessment.
1) the reliability average relative error of predicated error:(mean relative error,MRE)：
2) size of predicted value deviation true value error is with mean absolute error (mean absolute error, MAE)：
3) sample data dispersion degree is with root-mean-square error (root mean square error, RMSE)：
4) forecast model accuracy (forecast model accuracy, FMA)：
Wherein：XiIt is actual value, Xi' it is and predicted value that M is sample size.
LVQ-CPSO-BP algorithms
LVQ algorithms
LVQ belongs to multilayer feedforward network, for input sample Classification and Identification.Network by input layer, competition layer and
Output layer up of three layers, when input of the sample data as LVQ, the neuron of competition layer by the way that the victor is a king, produce by the rules of competition
Raw triumph neuron, is output as 1, and loser is output as 0.The output neuron being connected with group where triumph neuron is output as
1, other are output as 0, to reach the purpose that output sample classification is recognized.LVQ network structures are as shown in Figure 3.
CPSO-BP algorithms
BP algorithm is that one kind common are tutor's supervision type algorithm, by the error between actual value and predicted value, reversely
The weights and threshold value being broadcast between input layer, hidden layer and output layer, because BP algorithm is essentially gradient descent method, computing
Later stage is easily absorbed in local optimum, convergence rate slowly, and PSO algorithms are a kind of optimization tools of iteration, the spy with global optimizing
Point, it, without derivative constant gradient information, can well solve the structure and right-value optimization of neutral net in optimization neural network
Problem, but population diversity is lost during PS0 algorithms have searching process, is easily caused algorithm and showing for premature convergence occurs
As, therefore chaos thought is combined with PS0 algorithms, the particle of PSO algorithms is scanned for using chaos algorithm so that two kinds
The advantage of algorithm is blended, and forms CPSO algorithms.Using the weights and threshold value of CPSO algorithm optimization BP neural networks, both reached
The purpose of global optimizing, improves convergence rate again.CPSO-BP algorithm flows are as shown in Figure 4.
LVQ-CPSO-BP forecast models and checking
LVQ-CPSO-BP forecast models
LVQ Classification and Identification models
Coal gas permeability shows two kinds of Changing Patterns with the change of effective stress, changes precipitous before flex point,
Change after flex point gentle, it is therefore apparent that the big and precipitous coal gas permeability potential safety hazard of value is bigger.Therefore the present invention is adopted
With effective stress flex point it is that boundary is classified as coal gas permeability variation high order smooth pattern and precipitous type two types with LVQ, with sample
Effective stress is the average value of (1-5) MPa in notebook dataTemperature averagesGas pressure average valueCompression strength
Average valueAs the input of LVQ, by after the training of LVQ Classification and Identifications, obtaining two class classification samples data.
Coal gas Permeability Prediction learning sample vector represents that sample set is with X：
[X]=[x1,x2,x3,x4]T=[σ, T, P, σm]T (5)
Wherein σ is effective stress (MPa), T is temperature (DEG C), P is gas pressure (MPa), σmIt is compression strength (MPa)
Input layer input set is：
The sample of competition layer is expressed as with Y：
Y=(y1, y2... ym)T, yi∈ { 0,1 }, i=1,2 ..., m (7)
Weight matrix W between input layer and competition layer1：
Competition layer is to the weight matrix W between output layer2：
The desired value output of network is expressed as with d：
D=(d1, d2... dl)T (10)
The output of output layer represents that output collection is with O：
[O]=[o1, o2]T=[1,0]T (11)
Wherein, o1It is precipitous type classification, o2It is high order smooth pattern classification.
CPSO-BP forecast models
After sample data to be divided into LVQ two major classes, predicted using CPSO-BP, CPSO algorithms are mapped using Logistic
Chaos system, iterative formula is：
zi+1=μ zi(l-zi), i=0,1 ..., μ ∈ (2,4] (12)
In formula：μ is control variables.
When μ=4,0≤z0When≤1, Logistic is completely in chaos state, and chaotic motion characteristic during using μ=4 changes
Enter standard particle group's algorithm, i.e.,
zi+1=4zi(1-zi), i=0,1 ... (13)
Population initial phase produces excellent initial population as primary group by the use of formula (13), to mitigate
Influence of the random initializtion population to algorithm optimization performance.
The number of plies of BP neural network structure is 3 layers, if R, S1, S2 are respectively input layer, hidden layer, output layer neuron
Number, input layer therein is the sorted sample datas of LVQ, and hidden layer is S function, and most preferably implicit nodes are 3,
Output layer is coal gas permeability.Between the dimension N of weights and threshold value and population individuality the position vector P of BP neural network
Mapping relations such as formula (14) shown in：
N=S1 × R+S1 × S2+S1+S2 (14)
Learning sample data
Screened by influence factor, present invention obtains main " four factors ", its part sample data comes from Tang Guoshui,
Zhang Hongwei, Han Jun, wait to be based on coal seam with gas Permeability Prediction model [J] China safety in production science skill of MABC-SVM
Art, 2015,02:Test data in 11-16., is shown in Table 1.
The coal gas permeability learning sample master data of table 1
Learning sample is trained and predicted the outcome
The present invention using the data of Matlab2015 Neural Network Toolbox combinations table 1 write BP, GA-BP, PSO-BP,
Tetra- kinds of combination Permeability Prediction model codes of CPSO-BP, setting BP neural network iterations 1500 times, learning rate
0.01, lowest mean square root error 10-5；The particle of CPSO optimization neural networks takes W=0.7298, c1=c2=1.4962, particle
Upper dividing value be 5, floor value be -5.
LVQ is trained and classification results
The sample data of table 1 input LVQ is trained, netinit is carried out first：The smaller random number of setting is used as first
Beginning weights W1And W2；It is determined that initial learning efficiency η (0)=0.9 and frequency of training tm=100 and classification species number K=2.It is input into and is
Four factors, 40 groups of average vectors of sample vector in table 1.Triumph neuron is found by the way that desired value is minimum.It is small with neuron of winning
Being divided into for the effective stress average value (about 2.5MPa) specified in the present invention is correctly classified, weights when otherwise adjustment is classified,
Renewal learning efficiency, until t=tm, obtain classification results.
Using effective stress, temperature, gas pressure, compression strength average value as input sample, it is stipulated that with effective stress
Average value is that the critical value left side of flex point and the output type on the right are followed successively by 1,0, and its Classification and Identification result is as shown in table 2.
The recognition correct rate of table 2 is contrasted
From table 2 it can be seen that effective stress flex point is divided using LVQ models having precision higher, Classification and Identification is just
True rate reaches as high as 96.15%.
CPSO-BP is trained and predicted the outcome
The sample data of table 1 input LVQ is carried out into sorted two groups of sample datas input CPSO-BP to be trained, first
Initialization BP neural network 3-tier architecture, input layer is 4 (effective stress, temperature, gas pressure, compression strength), is implied
Layer neuron is 5, and output layer neuron is 1 (coal gas permeability).By chaos intialization Population Size, dimension D and grain
, from optimizing, the optimal solution that will be searched out is assigned to the power of BP networks for sub- S speed and position and mean square error population fitness function
Value and threshold value, continue the learning training prediction of BP networks.
The present invention has carried out 40 groups of learning trainings of sample data using four kinds of forecast models, and forecast model one is BP nerves
Neural network forecast, forecast model two is GA-BP neural network predictions, and forecast model three is PSO-BP neural network predictions, predicts mould
Type four is CPSO-BP.Input data in forecast model is the sorted data of LVQ, and wherein Class1 training sample is 26
Group；The training sample of type 0 is 14 groups.Four kinds of class prediction curves of model two and Tang Guoshui, Zhang Hongwei, Han Jun, wait to be based on MABC-
Coal seam with gas Permeability Prediction model [J] China safety in production science and technology of SVM, 2015,02:The experiment of 11-16. is surveyed
Examination curve is as shown in Figure 5.
The contrast simulation result of Fig. 5 fully shows：The coal gas Permeability Prediction model-fitting degree that the present invention sets up is high,
Prediction effect is good.
The coal gas permeability of Class1 is high and variable quantity is big, but precision of prediction is high, can more effectively avoid potential safety hazard,
Therefore the method that sample data is classified and prediction is optimized with reference to CPSO-BP is carried out using LVQ according to effective stress flex point, is had
Beneficial to the potential safety hazard early warning of gas in process of coal mining.
Analysis and assessment
The four kind differences that predict the outcome and test actual value of the error evaluation method that foundation is set up to coal gas permeability
Error calculation is carried out, as shown in table 3.
3 four kinds of error prediction model evaluation index results of table
As shown in Table 3, the coal gas Permeability Prediction precision of LVQ-CPSO-BP algorithms be 95.26%, better than other three
Model is planted, the LVQ-CPSO-BP forecast models set up are effective, adaptability is stronger, precision of prediction is higher.
Operation principle of the invention is：
(1) the coal gas Permeability Prediction method of LVQ-CPSO-BP algorithms is proposed, effective stress is obtained less than turning
The point section coal gas more preferable effect of Permeability Prediction precision, is conducive to the early warning of coal mine gas potential safety hazard.
(2) macro and micro analysis method is used, the main macroscopic factor for having filtered out influence coal gas permeability is coal
Layer buried depth, main microcosmic influence factors are effective stress, temperature, gas pressure, compression strength.
(3) according to the distribution character of coal seam buried depth, it is proposed that critical point concept hierarchy；According to coal gas permeability with have
There is knee characteristic in the relation between efficacy, it is proposed that flex point classification concept.
(4) the coal gas Permeability Prediction of LVQ-CPSO-BP algorithms based on Matlab programming realizations, chooses that " four will
Element " sample data, is predicted the outcome, and is demonstrated and is better than PSO- through the sorted CPSO-BP Prediction Accuracies of LVQ
BP neural network, GA-BP neutral nets and BP neural network.
(5) based on predicated error evaluation index, 4 kinds of error amounts of forecast model are calculated, wherein sorted based on LVQ
The various error amounts of CPSO-BP forecast models are respectively：MRE is that 6.36%, MAE is 0.0852 × 10-15m2, RMSE is 0.1741
×10-15m2, FMA is 95.26%, shows that the model has preferably accuracy and convergence, and predict the outcome more approaching to reality
Value.
(6) many combinational algorithm precision of predictions are substantially better than single algorithm, also superior to double combinational algorithms, demonstrate LVQ-
The many combinational algorithms of CPSO-BP are feasible, effective in terms of coal gas Permeability Prediction precision raising.
The above, specific embodiment only of the invention, but protection scope of the present invention is not limited thereto, and it is any
The change or replacement expected without creative work, should all be included within the scope of the present invention.Therefore, it is of the invention
Protection domain should be determined by the scope of protection defined in the claims.
Claims (10)
1. the coal gas Permeability Prediction method of LVQ-CPSO-BP algorithms is based on, it is characterised in that be made up of following steps：
Step one, coal seam buried depth are divided and selected：Coal seam buried depth is studied：Coal seam buried depth is divided into shallow embedding layer and non-shallow embedding
Layer, and shallow embedding layer is general in more than 150m, in gas weathered zone, the present invention is not discussed；Rather than shallow embedding layer refer to 150m with
Under, most as deep as 1500m or so is arrived, most of coal seam is between 200m-700m, therefore the present invention ignores shallow embedding layer, will be non-shallow
Buried regions is self-defined to be divided into two-layer, if critical value be 700m, with it is shallow be middle buried regions, be buried layer with depth；
Step 2, major influence factors selection：Selection effective stress, temperature, gas pressure, compression strength as main influence because
Element；
There is flex point in relation between effective stress and coal gas permeability, and when effective stress is less than flex point, coal gas ooze
Saturating rate change is very fast, and during more than flex point, coal gas permeability variation is gentle, and 2MPa is set into effective stress flex point；
Step 3, on the basis of coal seam buried depth non-shallow embedding layer macroscopic view is divided, using LVQ algorithms to effective stress sample data by turning
Point carries out microscopic classification；
The coal gas Permeability Prediction submodel of step 4, foundation based on CPSO-BP algorithms；
Step 5, the training by LVQ-CPSO-BP forecasting systems, obtain coal gas Permeability Prediction result.
2. method according to claim 1, it is characterised in that in order to evaluate the confidence level of set up forecast model and accurate
Degree, quantifies to error and precision, and coal gas infiltration is carried out using formula (1)-formula (4) predicated error and precision evaluation index
Rate predicts the outcome qualitative assessment：
1) the reliability average relative error of predicated error：
2) the size mean absolute error of predicted value deviation true value error：
3) sample data dispersion degree root-mean-square error：
4) forecast model accuracy：
Wherein：XiIt is actual value, X 'iIt is and predicted value that M is sample size.
3. method according to claim 1, it is characterised in that in the step 3, uses the LVQ to be with effective stress flex point
Boundary is classified as coal gas permeability variation high order smooth pattern and precipitous type two types, and effective stress is (1- with sample data
5) average value of MPaTemperature averagesGas pressure average valueCompression strength average valueAs the input of LVQ,
By after the training of LVQ Classification and Identifications, obtaining two class classification samples data.
4. method according to claim 3, it is characterised in that in the step 3, the study of coal gas Permeability Prediction
Sample vector represents that sample set is with X：
[X]=[x1,x2,x3,x4]T=[σ, T, P, σm]T (5)
Wherein σ is effective stress units MPa, T is temperature unit DEG C, P is gas pressure units MPa, σmIt is compression strength unit
MPa；
Input layer input set is：
The sample of competition layer is expressed as with Y：
Y=(y1,y2,…ym)T,yi∈ { 0,1 }, i=1,2 ..., m (7)
Weight matrix W between input layer and competition layer1：
Competition layer is to the weight matrix W between output layer2：
The desired value output of network is expressed as with d：
D=(d1,d2,…dl)T (10)
The output of output layer represents that output collection is with O：
[O]=[o1,o2]T=[1,0]T (11)
Wherein, o1It is precipitous type classification, o2It is high order smooth pattern classification.
5. method according to claim 4, it is characterised in that screened in the step 3, by influence factor, with effective
Stress, temperature, gas pressure, the corresponding data of compression strength and permeability is learnt for sample；
Tetra- kinds of BP, GA-BP, PSO-BP, CPSO-BP has been write using Matlab2015 Neural Network Toolbox combination sample datas
Combination Permeability Prediction model code, setting BP neural network iterations 1500 times, learning rate 0.01, lowest mean square root
Error 10-5；The particle of CPSO optimization neural networks takes W=0.7298, c1=c2=1.4962, the upper dividing value of particle is 5, lower bound
It is worth for -5.
6. method according to claim 5, it is characterised in that in the step 4, sample data is divided into two in LVQ big
After class, predicted using CPSO-BP, CPSO algorithms map chaos system using Logistic, iterative formula is：
zi+1=μ zi(1-zi), i=0,1 ..., and μ ∈ (2,4] (12)
In formula：μ is control variables；
When μ=4,0≤z0When≤1, Logistic is completely in chaos state, chaotic motion improved properties mark during using μ=4
Quasi particle group's algorithm, i.e.,
zi+1=4zi(1-zi), i=0,1 ... (13)
Population initial phase produces excellent initial population as primary group by the use of formula (13), random to mitigate
Influence of the initialization population to algorithm optimization performance.
7. method according to claim 6, it is characterised in that in the step 4, by the number of plies of BP neural network structure
3 layers are set to, if R, S1, S2 are respectively input layer, hidden layer, output layer neuron number, input layer therein is LVQ
Sorted sample data, hidden layer is S function, and most preferably implicit nodes are 3, and output layer is coal gas permeability, BP
Shown in mapping relations such as formula (14) between the dimension N of weights and threshold value and population individuality the position vector P of neutral net：
N=S1 × R+S1 × S2+S1+S2 (14)
8. method according to claim 7, it is characterised in that instructed in the step 5, by sample data input LVQ
Practice, netinit is carried out first：The smaller random number of setting is used as initial weight W1And W2；It is determined that initial learning efficiency η (0)=
0.9 and frequency of training tm=100 and classification species number K=2, it is four factors, 40 groups of average vectors of sample vector in table 1 to be input into,
Triumph neuron is found by the way that desired value is minimum, with neuron of winning less than the effective stress average value specified of the invention (about
Being divided into 2.5MPa) is correctly classified, weights when otherwise adjustment is classified, renewal learning efficiency, until t=tm, classified
As a result.
9. method according to claim 8, it is characterised in that in the step 5, with effective stress, temperature, gas pressure
Power, the average value of compression strength are used as input sample, it is stipulated that the critical value left side and the right with effective stress average value as flex point
Output type be followed successively by 1,0, obtain Classification and Identification result；
Sample data input LVQ is carried out into sorted two groups of sample datas input CPSO-BP to be trained, BP is initialized first
Neutral net 3-tier architecture, input layer is 4：Effective stress, temperature, gas pressure, compression strength；Hidden layer neuron
It is 5, output layer neuron is 1：Coal gas permeability；By chaos intialization Population Size, dimension D and particle S speed and
, from optimizing, the optimal solution that will be searched out is assigned to the weights and threshold value of BP networks for position and mean square error population fitness function,
Continue the learning training prediction of BP networks.
10. method according to claim 9, it is characterised in that the present invention has carried out 40 groups of samples using four kinds of forecast models
The learning training of notebook data, forecast model one predicts that forecast model two is GA-BP neural network predictions, in advance for BP neural network
Survey model three is PSO-BP neural network predictions, and forecast model four is CPSO-BP；Input data in forecast model is LVQ
Sorted data, wherein Class1 training sample are 26 groups；The training sample of type 0 is 14 groups, using LVQ models to there is effect
Power flex point is divided has precision higher, and Classification and Identification accuracy reaches as high as 96.15%.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710121485.3A CN106869990B (en) | 2017-03-02 | 2017-03-02 | Coal gas Permeability Prediction method based on LVQ-CPSO-BP algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710121485.3A CN106869990B (en) | 2017-03-02 | 2017-03-02 | Coal gas Permeability Prediction method based on LVQ-CPSO-BP algorithm |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106869990A true CN106869990A (en) | 2017-06-20 |
| CN106869990B CN106869990B (en) | 2019-03-19 |
Family
ID=59169287
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710121485.3A Expired - Fee Related CN106869990B (en) | 2017-03-02 | 2017-03-02 | Coal gas Permeability Prediction method based on LVQ-CPSO-BP algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106869990B (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109120961A (en) * | 2018-07-20 | 2019-01-01 | 南京邮电大学 | The prediction technique of the QoE of IPTV unbalanced dataset based on PNN-PSO algorithm |
| CN109191423A (en) * | 2018-07-18 | 2019-01-11 | 中国矿业大学 | A kind of porous media Permeability Prediction method based on machine image intelligence learning |
| CN109657361A (en) * | 2018-12-21 | 2019-04-19 | 成都四方伟业软件股份有限公司 | Gas content prediction method, terminal device and computer readable storage medium |
| CN110633504A (en) * | 2019-08-21 | 2019-12-31 | 中联煤层气有限责任公司 | Prediction method for coal bed gas permeability |
| CN111398130A (en) * | 2020-04-30 | 2020-07-10 | 太原理工大学 | Method for analyzing permeability of lump coal from multi-dimensional data source, measuring device and method |
| CN111461457A (en) * | 2020-04-17 | 2020-07-28 | 福州大学 | Foundation pit displacement prediction method based on particle swarm optimization BP neural network |
| CN111521537A (en) * | 2020-04-30 | 2020-08-11 | 太原理工大学 | Multidimensional data measuring device for coal block drilling process |
| CN111929285A (en) * | 2020-08-11 | 2020-11-13 | 安徽理工大学 | Spectral identification method for alcohol doped in laser-induced fluorescent red wine |
| CN113570165A (en) * | 2021-09-03 | 2021-10-29 | 中国矿业大学 | Coal reservoir permeability intelligent prediction method based on particle swarm optimization |
| CN116930042A (en) * | 2023-09-19 | 2023-10-24 | 常州建昊建筑鉴定检测有限公司 | Building waterproof material performance detection equipment and method |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040060351A1 (en) * | 2002-09-30 | 2004-04-01 | Gunter William Daniel | Process for predicting porosity and permeability of a coal bed |
| CN104316448A (en) * | 2014-11-03 | 2015-01-28 | 中国石油大学(华东) | Prediction method for dynamic change of high-rank coal rock gas-phase permeability |
| CN104500050A (en) * | 2014-10-31 | 2015-04-08 | 中国石油大学(华东) | Fractured reservoir permeability tensor and anisotropy quantitative prediction method |
| CN104612635A (en) * | 2014-12-10 | 2015-05-13 | 中煤科工集团重庆研究院有限公司 | Standard-reaching pre-judgment method for coal seam group gas combined extraction |
| CN104794550A (en) * | 2015-05-13 | 2015-07-22 | 山东科技大学 | WT-KPCA-SVR coupling model based gas emission quantity prediction method |
| CN106202748A (en) * | 2016-07-14 | 2016-12-07 | 中国海洋石油总公司 | A kind of based on permeability with the predictive methodology of coalbed methane recovery efficiency of well spacing |
-
2017
- 2017-03-02 CN CN201710121485.3A patent/CN106869990B/en not_active Expired - Fee Related
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040060351A1 (en) * | 2002-09-30 | 2004-04-01 | Gunter William Daniel | Process for predicting porosity and permeability of a coal bed |
| CN104500050A (en) * | 2014-10-31 | 2015-04-08 | 中国石油大学(华东) | Fractured reservoir permeability tensor and anisotropy quantitative prediction method |
| CN104316448A (en) * | 2014-11-03 | 2015-01-28 | 中国石油大学(华东) | Prediction method for dynamic change of high-rank coal rock gas-phase permeability |
| CN104612635A (en) * | 2014-12-10 | 2015-05-13 | 中煤科工集团重庆研究院有限公司 | Standard-reaching pre-judgment method for coal seam group gas combined extraction |
| CN104794550A (en) * | 2015-05-13 | 2015-07-22 | 山东科技大学 | WT-KPCA-SVR coupling model based gas emission quantity prediction method |
| CN106202748A (en) * | 2016-07-14 | 2016-12-07 | 中国海洋石油总公司 | A kind of based on permeability with the predictive methodology of coalbed methane recovery efficiency of well spacing |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109191423A (en) * | 2018-07-18 | 2019-01-11 | 中国矿业大学 | A kind of porous media Permeability Prediction method based on machine image intelligence learning |
| CN109120961B (en) * | 2018-07-20 | 2020-11-03 | 南京邮电大学 | QoE prediction method of IPTV unbalanced data set based on PNN-PSO algorithm |
| CN109120961A (en) * | 2018-07-20 | 2019-01-01 | 南京邮电大学 | The prediction technique of the QoE of IPTV unbalanced dataset based on PNN-PSO algorithm |
| CN109657361A (en) * | 2018-12-21 | 2019-04-19 | 成都四方伟业软件股份有限公司 | Gas content prediction method, terminal device and computer readable storage medium |
| CN109657361B (en) * | 2018-12-21 | 2022-09-13 | 成都四方伟业软件股份有限公司 | Gas content prediction method, terminal device and computer readable storage medium |
| CN110633504A (en) * | 2019-08-21 | 2019-12-31 | 中联煤层气有限责任公司 | Prediction method for coal bed gas permeability |
| CN111461457A (en) * | 2020-04-17 | 2020-07-28 | 福州大学 | Foundation pit displacement prediction method based on particle swarm optimization BP neural network |
| CN111521537A (en) * | 2020-04-30 | 2020-08-11 | 太原理工大学 | Multidimensional data measuring device for coal block drilling process |
| CN111398130A (en) * | 2020-04-30 | 2020-07-10 | 太原理工大学 | Method for analyzing permeability of lump coal from multi-dimensional data source, measuring device and method |
| CN111929285A (en) * | 2020-08-11 | 2020-11-13 | 安徽理工大学 | Spectral identification method for alcohol doped in laser-induced fluorescent red wine |
| CN113570165A (en) * | 2021-09-03 | 2021-10-29 | 中国矿业大学 | Coal reservoir permeability intelligent prediction method based on particle swarm optimization |
| CN113570165B (en) * | 2021-09-03 | 2024-03-15 | 中国矿业大学 | Intelligent prediction method for permeability of coal reservoir based on particle swarm optimization |
| CN116930042A (en) * | 2023-09-19 | 2023-10-24 | 常州建昊建筑鉴定检测有限公司 | Building waterproof material performance detection equipment and method |
| CN116930042B (en) * | 2023-09-19 | 2023-12-01 | 常州建昊建筑鉴定检测有限公司 | Building waterproof material performance detection equipment and method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106869990B (en) | 2019-03-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106869990A (en) | Coal gas Permeability Prediction method based on LVQ CPSO BP algorithms | |
| Tahmasebi et al. | Data mining and machine learning for identifying sweet spots in shale reservoirs | |
| Elkatatny et al. | New insights into porosity determination using artificial intelligence techniques for carbonate reservoirs | |
| Bhattacharjya et al. | Optimal management of coastal aquifers using linked simulation optimization approach | |
| Aïfa et al. | Neuro-fuzzy system to predict permeability and porosity from well log data: A case study of Hassi R׳ Mel gas field, Algeria | |
| CN106295199B (en) | Automatic history matching method and system based on autocoder and multiple-objection optimization | |
| CN110363344A (en) | Probability integral parameter prediction method based on MIV-GP algorithm optimization BP neural network | |
| Karkevandi-Talkhooncheh et al. | Application of hybrid adaptive neuro-fuzzy inference system in well placement optimization | |
| CN105913450A (en) | Tire rubber carbon black dispersity evaluation method and system based on neural network image processing | |
| CN103617147A (en) | Method for identifying mine water-inrush source | |
| Behnia et al. | Modeling of shear wave velocity in limestone by soft computing methods | |
| CN105913332A (en) | Method for determining tight gas reservoir development index and device thereof | |
| CN105044775A (en) | Seismic fluid inversion identification method and apparatus | |
| CN104680025A (en) | Oil pumping unit parameter optimization method on basis of genetic algorithm extreme learning machine | |
| Shokir et al. | Permeability estimation from well log responses | |
| Verma et al. | Quantification of sand fraction from seismic attributes using Neuro-Fuzzy approach | |
| Chatterjee et al. | Ore grade estimation of a limestone deposit in India using an artificial neural network | |
| Sfidari et al. | Electrofacies clustering and a hybrid intelligent based method for porosity and permeability prediction in the South Pars Gas Field, Persian Gulf | |
| Djebbas et al. | Integrating hydraulic flow unit concept and adaptive neuro-fuzzy inference system to accurately estimate permeability in heterogeneous reservoirs: Case study Sif Fatima oilfield, southern Algeria | |
| CN106503793A (en) | A kind of neural network short-term wind speed forecasting method based on improvement difference algorithm | |
| Vakili-Ghahani et al. | Control-relevant upscaling | |
| Luo et al. | Optimization Method of Hydraulic Fracturing Design for Horizontal Shale Gas Well Based on Artificial Neural Network and Genetic Algorithm | |
| Akhlaghi et al. | Optimizing the location of the gas injection well during gas assisted gravity drainage in a fractured carbonate reservoir using artificial intelligence | |
| Khoukhi et al. | A data-driven genetic neuro-fuzzy system to PVT properties prediction | |
| CN103049632A (en) | Non-supervision dimension reducing method aiming at nonlinear data |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190319 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |