CN112668749A - Coal mine gas early warning method based on class mark weighting extreme learning machine - Google Patents

Coal mine gas early warning method based on class mark weighting extreme learning machine Download PDF

Info

Publication number
CN112668749A
CN112668749A CN202011325639.9A CN202011325639A CN112668749A CN 112668749 A CN112668749 A CN 112668749A CN 202011325639 A CN202011325639 A CN 202011325639A CN 112668749 A CN112668749 A CN 112668749A
Authority
CN
China
Prior art keywords
data
gas concentration
early warning
learning machine
extreme learning
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
Application number
CN202011325639.9A
Other languages
Chinese (zh)
Other versions
CN112668749B (en
Inventor
虎东成
王超
曹文敬
许冉
黄永正
赵青青
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Jiangsu Zhong Kuang An Hua Technology Development Co ltd
Original Assignee
Jiangsu Zhong Kuang An Hua Technology Development Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Jiangsu Zhong Kuang An Hua Technology Development Co ltd filed Critical Jiangsu Zhong Kuang An Hua Technology Development Co ltd
Priority to CN202011325639.9A priority Critical patent/CN112668749B/en
Publication of CN112668749A publication Critical patent/CN112668749A/en
Application granted granted Critical
Publication of CN112668749B publication Critical patent/CN112668749B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30—Computing systems specially adapted for manufacturing

Landscapes

  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a coal mine gas early warning method based on a class mark weighting extreme learning machine, which comprises the steps of obtaining historical gas concentration monitoring data, preprocessing the data, filling missing data, generating a sample set, setting a gas concentration early warning threshold value, determining the imbalance ratio of a sample by adopting an early warning threshold value mu, carrying out weighting operation on a class mark matrix of the sample according to the feedback value of the imbalance ratio, determining the optimal number of hidden layer nodes of the extreme learning machine by utilizing five-fold cross verification, taking a Sigmoid function as an activation function, taking the weighted class mark matrix as an expected output matrix of a model, training and generating an extreme learning machine model. The method and the device realize the prediction of whether the gas concentration of a single monitoring point exceeds the warning line in the future 3 hours, carry out early warning on potential production safety risks, consider the influence of class imbalance, and have the advantages of simple realization, high detection rate and low false alarm rate.

Description

Coal mine gas early warning method based on class mark weighting extreme learning machine
Technical Field
The invention relates to a coal mine gas early warning method, and belongs to the technical field of gas concentration prediction.
Background
In the energy industry of China, coal accounts for about 70% of disposable energy production and consumption structures of China, and in a long period in the future, coal still serves as a main energy source of China, but at present, the coal mine production safety situation is very severe, safety accidents still happen, and a plurality of coal mine monitoring systems of China lack a disaster prediction function. Mine gas is one of main factors influencing coal mine safety, and a series of safety accidents can be caused by the fact that the gas concentration is too high. Therefore, whether the gas concentration in the next hours exceeds an early warning line or not is accurately predicted, potential safety hazards are timely found, and the method has great significance for safety production of coal mines. However, in coal mine gas monitoring data, the early warning data is often far less than the non-early warning data, which causes the problem of category imbalance, and can greatly reduce the detection rate of the traditional prediction model and increase the false alarm rate, thereby affecting the actual use effect.
Disclosure of Invention
The purpose of the invention is as follows: in order to overcome the defects in the prior art, the invention provides a coal mine gas early warning method based on a class-standard weighting extreme learning machine, which can effectively solve the problem of class imbalance, accurately predict whether the gas concentration will exceed an early warning line within the next several hours according to the historical data of the gas, is mainly applied to coal mine safety early warning, and can improve the pre-control capability of a coal mine on gas explosion accidents.
The technical scheme is as follows: in order to achieve the purpose, the invention adopts the technical scheme that:
a coal mine gas early warning method based on a class mark weighting extreme learning machine comprises the following steps:
step 1, acquiring historical gas concentration monitoring data of a certain monitoring point under a mine. Usually, a monitoring point is arranged under a mine, and gas concentration data are extracted and transmitted in real time.
Step 2, data preprocessing and missing data filling: calculating the gas concentration mean value once every ten minutes for one monitoring point to obtain regular gas concentration monitoring data, performing linear filling processing on a null value by using the mean values calculated before and after, if the data are continuously missing for more than two hours, not performing filling, discarding the corresponding time data point, and continuously calculating a new data point at the appearance position of new detection data to obtain a regular gas concentration monitoring time sequence.
And 3, generating a sample set: and taking gas concentration data of 24 continuous hours as a characteristic X, taking gas concentration data of 3 subsequent hours as a prediction target Y, and generating a training sample set phi one by one according to a single-point sequential movement mode.
And 4, setting a gas concentration early warning threshold value mu according to specific requirements, and determining that the safety production risk occurs when the gas concentration exceeds the early warning threshold value mu.
And 5, class mark weighting, wherein the imbalance ratio of the sample is determined by adopting an early warning threshold value mu, and the weighting operation is carried out on the class mark matrix of the sample according to the feedback value of the imbalance ratio.
Step 51, scanning sample sets phi one by one to predict target Y1,Y2,...,YNThe number of the early warning threshold value mu and the number of the early warning threshold value mu are respectively recorded as N-And N+And calculating the ratio λ ═ N-/N+。
And step 52, performing preliminary assignment on the target output matrix T according to the scanning result in the step 51.
Figure BDA0002794187960000021
Wherein, if Yij≥μ，tij1, otherwise, tij-1, i denotes the position of the training sample, j denotes the predicted time point, ti j denotes the label; .
Step 53, scanning the initial target output matrix T and weighting it, if TijWhen it is 1, it is rewritten as
Figure BDA0002794187960000022
If tijIf-1, then no operation is performed, and the quasi-scale weighting matrix T is obtainedWAnd λ represents an unbalance ratio.
And 6, training a model, namely determining the number of the optimal hidden layer nodes of the extreme learning machine by utilizing five-fold cross validation, taking a Sigmoid function as an activation function, taking a weighted class mark matrix as an expected output matrix of the model, and training and generating the extreme learning machine model.
Step 61, setting the number L of hidden layer nodes in the extreme learning machine, and randomly generating a weight and offset sequence from an input layer to a hidden layer, wherein a is [ a ]1,a2,...,ai2,...,aL]And b ═ b1,b2,...,bi2,...,bL]Wherein a isi2∈[-1,1]，bi2∈[-1,1]Converting a training sample set phi into a hidden layer output matrix by using a Sigmoid activation function:
Figure BDA0002794187960000023
wherein, H (a)1,...,aL,b1,...,bL,X1,...,XS) The expression represents a hidden output matrix, G is a Sigmoid activation function, and the calculation mode is as follows:
Figure BDA0002794187960000024
wherein, G (a)i2,bi2,Xj2) Representing an activation function.
Step 62, combining the hidden layer output matrix H and the quasi-label weighting matrix TWSolving the following formula:
Hβ＝Tw
wherein the content of the first and second substances,
Figure BDA0002794187960000031
using the least squares method, the calculation results are as follows:
Figure BDA0002794187960000032
wherein, beta represents the weight matrix of the hidden layer under the ideal state,
Figure BDA0002794187960000033
representing the calculated hidden layer weight matrix,
Figure BDA0002794187960000034
represents the weight value of the Lth hidden node,
Figure BDA0002794187960000035
the generalized inverse Moore-Penrose is a Moore-Penrose generalized inverse of the hidden layer output matrix, and the result is a least-norm least square solution of beta.
And 7, data prediction: and predicting whether the future gas concentration exceeds a warning line by using the trained class-standard weighted extreme learning machine model, and determining whether to give an early warning according to a prediction result.
Acquiring gas concentration monitoring data of the last 24 hours in real time, recording as XX after normalization processing, and storing a ═ a [ [ a ═ a [ ]1,a2,...,aL]And b ═ b1,b2,...,bL]Converting X into H, combining the stored hidden layer weight matrix beta, obtaining a corresponding early warning prediction result Y in the next three hours by calculating H multiplied by beta, checking whether a value larger than 0 exists in Y one by one, and if yes, sending an alarm signal.
Preferably: in step 1, historical gas concentration monitoring data of at least 30 days is included for one monitoring point.
Preferably: the method for obtaining the regular gas concentration monitoring time series in the step 2 comprises the following steps:
step 21, reading historical gas concentration monitoring data, extracting all data within 10 minutes from the first data, and performing mean value calculation to generate regular monitoring gas concentration data x1And by analogy, calculating every 10 minutes to finally obtain regular gas concentration monitoring time sequence data x1,x2,...,xKAnd K is the total number of the regular monitoring data.
Step 22, monitoring the time-series data x if the gas concentration is detected1,x2,...,xKIn xtAnd xt′Continuous missing data exists between points, and meanwhile, t' -t is less than or equal to 12, the missing value is filled by adopting the following formula.
Figure BDA0002794187960000036
Wherein t, t' is belonged to {1, 2.., K }, and S is xtAnd xt′The number of data continuously missing between the points, i, indicates the missing value position of the current missing sequence. If xtAnd xt′And if the number of the continuously missing data among the points is more than 12, setting a breakpoint, and not compensating the missing data.
Preferably: the method for obtaining the sample set phi in the step 3 comprises the following steps: monitoring data time series x for regular gas concentration1, x2,...,xKFirst, the gas concentration data of 24 continuous hours is taken from the initial position as the characteristic X1＝[x1,x2,...,x144]And then 3 hours later as the corresponding prediction target Y1＝[x145,x146,...,x162]First sample is generated, and then sample 2 is generated by moving a single point, namely X2＝[x2,x3,...,x145]，Y2＝[x146,x147,...,x163]And by analogy, continuously generating samples until the data sequence is exhausted to obtain a sample set phi:
Figure BDA0002794187960000041
where N represents the number of samples generated.
Preferably: if the time sequence x of the data is monitored in the regular gas concentration in the step 31,x2,...,xKWhen a breakpoint is encountered, the extraction of the new sample resumes after the breakpoint.
Preferably: and (4) adopting iterative solution to the early warning threshold value mu in the step 4, wherein the initial value is set according to the coal mine safety production standard and the actual requirement of the coal mine safety production.
Preferably: the number L of hidden layer nodes set in the step 61 is obtained by a five-fold cross validation method, that is, the preset number L is {10,50,100,200}, a sample set Φ is randomly and evenly divided into five parts, four parts of the sample set Φ are used as a training set for training each preset value L one by one to mark a weighted limit learning machine model, the rest parts are validated, after each sample is used as a test set, the root mean square prediction precision of each value L is calculated, the value L with the maximum precision is selected as the optimal hidden layer node parameter value, and the optimal hidden layer node parameter value is fed back to the step 61, and the square root error can ensure that convergence is stable.
Preferably: the trained model in the step S7 requires that the latest 24-hour gas concentration monitoring data of the same monitoring point be read in real time when in use.
Compared with the prior art, the invention has the following beneficial effects:
1. the method adopts the class mark weighting extreme learning machine to predict whether the gas concentration of a single monitoring point in the future 3 hours exceeds the warning line or not, and carries out early warning on potential production safety risks, thereby considering the influence of class unbalance, and having the advantages of simple realization, high detection rate and low false alarm rate.
2. The benefits of the process of calculating the scaling-like weights and weighting the scaling-like weight output matrix and the process of solving for β in 35 are described.
The step 52 is already the process, and has the advantages that the problem of unbalanced class in the data is fully considered, the class label with the high-concentration gas value is endowed with a larger expected output value, the error range can be enlarged, and conversely, the class label with the low-concentration gas value is endowed with a smaller expected output value, so that the prediction effect is achieved, and the output of the two is neutralized.
Drawings
FIG. 1 is a flow chart of the present invention.
FIG. 2 is a detailed flow chart of the method of the present invention.
Fig. 3 is a flow chart of the method of the present invention running online.
Detailed Description
The present invention is further illustrated by the following description in conjunction with the accompanying drawings and the specific embodiments, it is to be understood that these examples are intended only to illustrate the present invention and not to limit the scope of the present invention, which is defined in the appended claims to the present application, and that modifications of various equivalent forms to the present invention by those skilled in the art will fall within the scope of the present invention after reading the present invention.
A coal mine gas early warning method based on a class mark weighting extreme learning machine is shown in figures 1-3 and comprises the following steps:
A. and C, extracting gas concentration monitoring historical data of a certain coal mine monitoring point acquired by a sensor, calculating the average value at equal intervals of 10 minutes to obtain normalized gas concentration monitoring data, analyzing whether the monitoring data is missing or not, and entering the step B if the monitoring data is missing.
B. And C, filling missing data by using a linear filling compensation method, if the number of the continuous missing data is less than or equal to 12, filling, if the number of the continuous missing data exceeds 12, not filling, setting a breakpoint, and then entering the step C.
C. Taking gas concentration data for 24 continuous hours from the head of a normalized data time sequence as a characteristic X, taking gas concentration data for 3 hours later as a prediction target Y, moving according to a single-point sequence to generate training samples one by one, and if a breakpoint is met in the normalized gas concentration monitoring data time sequence, restarting the extraction of a new sample after the breakpoint to finally obtain a data set:
Figure BDA0002794187960000051
where N represents the number of samples generated.
D. Setting a gas concentration early warning threshold value mu, counting the number of predicted concentration values smaller than and larger than mu in all sample predicted targets Y, and respectively recording the number of the predicted concentration values as N-And N+And calculating the ratio λ ═ N-/N+。
E. And generating an initial target output matrix T corresponding to the extreme learning machine according to the predicted targets Y of all the samples, if the corresponding value in Y is greater than the threshold value mu, assigning a value of 1 to the corresponding position in T, and otherwise assigning a value of-1.
F. The initial target output matrix T is scanned and weighted, and if 1 is the value it is rewritten to
Figure BDA0002794187960000052
If the value is-1, no operation is performed, and a class mark weighting matrix T is obtainedW。
G. And training the quasi-standard weighted extreme learning machine model.
The concrete steps of the step G are as follows:
and setting the optimal number L of hidden layer nodes in the extreme learning machine by a five-fold cross-validation method.
Randomly generating a weight and bias sequence from an input layer to a hidden layer, a ═ a1,a2,...,ai2,...,aL]And b ═ b1,b2,...,bi2,..., bL]Wherein a isi2∈[-1,1]，bi2∈[-1,1]Converting a training sample set phi into a hidden layer output matrix by using a Sigmoid activation function:
Figure BDA0002794187960000061
wherein H(a1,...,aL,b1,...,bL,X1,...,XS) And representing a hidden layer output matrix, wherein G is a Sigmoid activation function, and the calculation mode is as follows:
Figure BDA0002794187960000062
wherein, G (a)i2,bi2,Xj2) Representing an activation function.
Combining hidden layer output matrix H and quasi-scale weighting matrix TWSolving the following formula:
Hβ＝Tw (4)
wherein the content of the first and second substances,
Figure BDA0002794187960000063
using the least squares method, the calculation results are as follows:
Figure BDA0002794187960000064
wherein, beta represents the weight matrix of the hidden layer under the ideal state,
Figure BDA0002794187960000065
representing the calculated hidden layer weight matrix,
Figure BDA0002794187960000066
represents the weight value of the lth hidden node,
Figure BDA0002794187960000067
is Moore-Penrose generalized inverse of the hidden layer output matrix. By definition, the result can be inferred to be a least norm least squares solution for β.
H. Arranging the model into a system, acquiring gas concentration monitoring data of last 24 hours in real time, recording as X' after carrying out normalization processing, and storing a ═ a [ a ]1,a2,...,aL]And b ═ b1,b2,...,bL]Converting X 'into H', combining with the stored hidden layer weight matrix beta, obtaining a corresponding early warning prediction result Y 'in the future three hours by calculating H' X beta, checking whether values larger than 0 exist in Y 'one by one, if yes, sending an alarm signal, and if not, returning to H' for running again after 10 minutes.
The above description is only of the preferred embodiments of the present invention, and it should be noted that: it will be apparent to those skilled in the art that various modifications and adaptations can be made without departing from the principles of the invention and these are intended to be within the scope of the invention.

Claims (8)

1. A coal mine gas early warning method based on a class mark weighting extreme learning machine is characterized by comprising the following steps:
step 1, acquiring historical gas concentration monitoring data of a certain monitoring point under a mine;
step 2, data preprocessing and missing data filling: calculating the gas concentration mean value once every ten minutes for a monitoring point to obtain regular gas concentration monitoring data, performing linear filling processing on a null value by using the mean values calculated before and after, if the data are continuously missing for more than two hours, not performing filling, discarding the corresponding time data point, and continuously calculating a new data point at the appearance position of new detection data to obtain a regular gas concentration monitoring time sequence;
and 3, generating a sample set: taking gas concentration data of 24 continuous hours as a characteristic X, taking gas concentration data of 3 subsequent hours as a prediction target Y, and generating a training sample set phi one by one according to a single-point sequential movement mode;
step 4, setting a gas concentration early warning threshold value mu, setting the gas concentration early warning threshold value mu according to specific needs, and determining that the safety production risk occurs when the gas concentration exceeds the early warning threshold value mu;
step 5, class mark weighting, namely determining the unbalance ratio of the sample by adopting an early warning threshold value mu, and weighting the class mark matrix of the sample according to the feedback value of the unbalance ratio;
step 51, scanning sample sets phi one by one to predict target Y1,Y2,...,YNThe number of the early warning threshold value mu and the number of the early warning threshold value mu are respectively recorded as N-And N+And calculating the ratio λ ═ N-/N+；
Step 52, performing preliminary assignment on the target output matrix T according to the scanning result in the step 51;
Figure FDA0002794187950000011
wherein, if Yij≥μ，tij1, otherwise, tij-1, i denotes the position of the training sample, j denotes the predicted time point, tijA presentation label;
step 53, scanning the initial target output matrix T and weighting it, if TijWhen it is 1, it is rewritten as
Figure FDA0002794187950000012
If tijIf-1, then no operation is performed, and the quasi-scale weighting matrix T is obtainedWλ represents an unbalance ratio;
step 6, training a model, namely determining the number of optimal hidden layer nodes of the extreme learning machine by utilizing five-fold cross validation, taking a Sigmoid function as an activation function, taking a weighted quasi-standard matrix as an expected output matrix of the model, and training and generating the extreme learning machine model;
step 61, setting the number L of hidden layer nodes in the extreme learning machine, and randomly generating a weight and a bias sequence from an input layer to a hidden layer, wherein a is [ a ]1,a2,...,ai2,...,aL]And b ═ b1,b2,...,bi2,...,bL]Wherein a isi2∈[-1,1]，bi2∈[-1,1]Converting a training sample set phi into a hidden layer output matrix by using a Sigmoid activation function:
Figure FDA0002794187950000021
wherein, H (a)1,...,aL,b1,...,bL,X1,...,XS) And representing a hidden output matrix, wherein G is a Sigmoid activation function and is calculated as follows:
Figure FDA0002794187950000022
wherein, G (a)i2,bi2,Xj2) Representing an activation function;
step 62, combining the hidden layer output matrix H and the quasi-label weighting matrix TWSolving the following formula:
Hβ＝Tw
wherein the content of the first and second substances,
Figure FDA0002794187950000023
using the least squares method, the calculation results are as follows:
Figure FDA0002794187950000024
wherein, beta represents the weight matrix of the hidden layer under the ideal state,
Figure FDA0002794187950000025
representing the calculated hidden layer weight matrix,
Figure FDA0002794187950000026
represents the weight value of the lth hidden node,
Figure FDA0002794187950000027
Moore-Penrose generalized inverse of hidden layer output matrix, the result is the minimum of betaNorm least squares solution;
and 7, data prediction: and predicting whether the future gas concentration exceeds a warning line by using the trained class-standard weighted extreme learning machine model, and determining whether to give an early warning according to a prediction result.
2. The coal mine gas early warning method based on the class mark weighting extreme learning machine as claimed in claim 1, wherein: in step 1, historical gas concentration monitoring data of at least 30 days is included for one monitoring point.
3. The coal mine gas early warning method based on the class mark weighting extreme learning machine as claimed in claim 2, wherein: the method for obtaining the regular gas concentration monitoring time series in the step 2 comprises the following steps:
step 21, reading historical gas concentration monitoring data, extracting all data within 10 minutes from the first data, and performing mean value calculation to generate regular monitoring gas concentration data x1And by analogy, calculating every 10 minutes to finally obtain regular gas concentration monitoring time sequence data x1,x2,...,xKWherein K is the total number of the regular monitoring data;
step 22, monitoring the time-series data x if the gas concentration is detected1,x2,...,xKIn xtAnd xt′Continuous missing data exist among the points, and meanwhile, t' -t is less than or equal to 12, the missing values are filled by adopting the following formula;
Figure FDA0002794187950000031
wherein t, t' is belonged to {1, 2.., K }, and S is xtAnd xt′The number of data continuously missing between points, i represents the missing value position of the current missing sequence; if xtAnd xt′And if the number of the continuously missing data among the points is more than 12, setting a breakpoint, and not compensating the missing data.
4. The coal mine gas early warning method based on the class mark weighting extreme learning machine as claimed in claim 3, wherein: the method for obtaining the sample set phi in the step 3 comprises the following steps: monitoring data time series x for regular gas concentration1,x2,...,xKFirst, the gas concentration data of 24 continuous hours is taken from the initial position as the characteristic X1＝[x1,x2,...,x144]And then 3 hours later as the corresponding prediction target Y1＝[x145,x146,...,x162]First sample is generated, and then sample 2 is generated by moving a single point, namely X2＝[x2,x3,...,x145]，Y2＝[x146,x147,...,x163]And by analogy, continuously generating samples until the data sequence is exhausted to obtain a sample set phi:
Figure FDA0002794187950000032
where N represents the number of samples generated.
5. The coal mine gas early warning method based on the class mark weighting extreme learning machine as claimed in claim 4, wherein: if the time sequence x of the data is monitored in the regular gas concentration in the step 31,x2,...,xKWhen a breakpoint is encountered, the extraction of the new sample resumes after the breakpoint.
6. The coal mine gas early warning method based on the class mark weighting extreme learning machine as claimed in claim 5, wherein: and (4) adopting iterative solution to the early warning threshold value mu in the step 4, wherein the initial value is set according to the coal mine safety production standard and the actual requirement of the coal mine safety production.
7. The coal mine gas early warning method based on the class mark weighting extreme learning machine as claimed in claim 6, wherein: the number L of hidden layer nodes set in the step 61 is obtained by a five-fold cross validation method, that is, preset L ═ {10,50,100,200}, by randomly and equally dividing the sample set Φ into five parts, using four parts of the sample set Φ as a training set one by one to train and mark a weighted limit learning machine model for each preset L value, and carrying out validation on the remaining one, after each sample is used as a test set, calculating the root mean square prediction accuracy of each L parameter, and selecting the L value with the maximum accuracy as the optimal hidden layer node parameter value.
8. The coal mine gas early warning method based on the class mark weighting extreme learning machine as claimed in claim 7, wherein: the model trained in the step S7 requires that the latest 24-hour gas concentration monitoring data of the same monitoring point be read in real time when in use.
CN202011325639.9A 2020-11-24 2020-11-24 Coal mine gas early warning method based on class mark weighting extreme learning machine Active CN112668749B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011325639.9A CN112668749B (en) 2020-11-24 2020-11-24 Coal mine gas early warning method based on class mark weighting extreme learning machine

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011325639.9A CN112668749B (en) 2020-11-24 2020-11-24 Coal mine gas early warning method based on class mark weighting extreme learning machine

Publications (2)

Publication Number Publication Date
CN112668749A true CN112668749A (en) 2021-04-16
CN112668749B CN112668749B (en) 2023-07-07

Family

ID=75403516

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011325639.9A Active CN112668749B (en) 2020-11-24 2020-11-24 Coal mine gas early warning method based on class mark weighting extreme learning machine

Country Status (1)

Country Link
CN (1) CN112668749B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113605980A (en) * 2021-07-23 2021-11-05 上海德衡数据科技有限公司 Intelligent mine safety early warning method and system based on Internet of things
CN114648217A (en) * 2022-03-17 2022-06-21 重庆邮电大学 Coal mine safety risk early warning method based on incremental extreme learning machine

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103678886A (en) * 2013-11-26 2014-03-26 北京空间飞行器总体设计部 Satellite Bayesian Network health determination method based on ground test data
CN105975611A (en) * 2016-05-18 2016-09-28 天津大学 Self-adaptive combined downsampling reinforcing learning machine
CN108875783A (en) * 2018-05-09 2018-11-23 西安工程大学 A kind of extreme learning machine Diagnosis Method of Transformer Faults towards unbalanced dataset
CN110084291A (en) * 2019-04-12 2019-08-02 湖北工业大学 A kind of students ' behavior analysis method and device based on the study of the big data limit
US10510003B1 (en) * 2019-02-14 2019-12-17 Capital One Services, Llc Stochastic gradient boosting for deep neural networks
CN111102005A (en) * 2019-09-16 2020-05-05 中煤科工集团重庆研究院有限公司 Mine environment gas concentration online prediction method
CN111160461A (en) * 2019-12-30 2020-05-15 浙江省机电设计研究院有限公司 Fuzzy clustering-based weighted online extreme learning machine big data classification method

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103678886A (en) * 2013-11-26 2014-03-26 北京空间飞行器总体设计部 Satellite Bayesian Network health determination method based on ground test data
CN105975611A (en) * 2016-05-18 2016-09-28 天津大学 Self-adaptive combined downsampling reinforcing learning machine
CN108875783A (en) * 2018-05-09 2018-11-23 西安工程大学 A kind of extreme learning machine Diagnosis Method of Transformer Faults towards unbalanced dataset
US10510003B1 (en) * 2019-02-14 2019-12-17 Capital One Services, Llc Stochastic gradient boosting for deep neural networks
CN110084291A (en) * 2019-04-12 2019-08-02 湖北工业大学 A kind of students ' behavior analysis method and device based on the study of the big data limit
CN111102005A (en) * 2019-09-16 2020-05-05 中煤科工集团重庆研究院有限公司 Mine environment gas concentration online prediction method
CN111160461A (en) * 2019-12-30 2020-05-15 浙江省机电设计研究院有限公司 Fuzzy clustering-based weighted online extreme learning machine big data classification method

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113605980A (en) * 2021-07-23 2021-11-05 上海德衡数据科技有限公司 Intelligent mine safety early warning method and system based on Internet of things
CN113605980B (en) * 2021-07-23 2024-03-29 上海德衡数据科技有限公司 Intelligent early warning mine safety method and system based on Internet of things
CN114648217A (en) * 2022-03-17 2022-06-21 重庆邮电大学 Coal mine safety risk early warning method based on incremental extreme learning machine

Also Published As

Publication number Publication date
CN112668749B (en) 2023-07-07

Similar Documents

Publication Publication Date Title
CN111830408B (en) Motor fault diagnosis system and method based on edge calculation and deep learning
US20200150622A1 (en) Method for detecting abnormity in unsupervised industrial system based on deep transfer learning
CN104933841B (en) A kind of fire prediction method based on self organizing neural network
CN112987675B (en) Method, device, computer equipment and medium for anomaly detection
CN106897178B (en) Slow disk detection method and system based on extreme learning machine
CN109992921B (en) On-line soft measurement method and system for thermal efficiency of boiler of coal-fired power plant
US11093314B2 (en) Time-sequential data diagnosis device, additional learning method, and recording medium
CN113723010B (en) Bridge damage early warning method based on LSTM temperature-displacement correlation model
CN108470095B (en) TBM (tunnel boring machine) propulsion prediction method based on data-driven radial basis function model
CN112668749A (en) Coal mine gas early warning method based on class mark weighting extreme learning machine
CN108921359B (en) Distributed gas concentration prediction method and device
CN112766429B (en) Method, device, computer equipment and medium for anomaly detection
CN112308306A (en) Multi-mode input coal and gas outburst risk prediction method
CN110490229A (en) A kind of electric energy meter calibration error diagnostics method based on spark and clustering algorithm
CN112069787A (en) Log parameter anomaly detection method based on word embedding
CN110011990B (en) Intelligent analysis method for intranet security threats
CN102013148A (en) Multi-information fusion fire hazard detection method
CN112799130A (en) Seismic wave amplitude prediction method based on deep learning
CN112149750A (en) Water supply network pipe burst identification data driving method
CN114002332A (en) Structural damage monitoring and early warning method and structural integrity digital twinning system
CN114138855B (en) Multivariable rainfall type landslide hazard monitoring method and monitoring system
CN111400850B (en) Equipment fault analysis method, device, equipment and storage medium
CN113157561A (en) Defect prediction method for numerical control system software module
CN115700363B (en) Method and system for diagnosing faults of rolling bearing of coal mining machine, electronic equipment and storage medium
CN116502119A (en) Sensor fault detection method and device of structural health monitoring system

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