Search

CN-115977736-B - Coal and gas outburst early warning method based on-site real-time data driving

CN115977736BCN 115977736 BCN115977736 BCN 115977736BCN-115977736-B

Abstract

The invention belongs to the technical field of coal mine safety production warning, and discloses a coal and gas outburst warning method based on-site real-time data driving. The method comprises the steps of continuously obtaining real-time data of gas and wind speed environment through a gas concentration sensor and a wind speed sensor which are arranged on a tunneling working face, calculating the gas emission quantity, carrying out deep analysis on the gas emission quantity, obtaining coal and gas outburst early warning indexes based on combination of dynamic change of the gas concentration and reverse identification based on gas emission quantity prediction evaluation indexes, establishing moving average line, deviation rate, discrete rate, fluctuation rate, root mean square error and average absolute percentage error early warning indexes, determining each index weight through a hierarchical analysis method, constructing a fuzzy comprehensive early warning model, calculating each index difference comprehensive index, and carrying out coal and gas outburst early warning. The method solves the problems that the data is difficult to acquire in real time and the on-site decision cannot be guided effectively in the existing method.

Inventors

  • WANG LI
  • HU HAIYONG
  • WEI LIKE

Assignees

  • 辽宁工程技术大学

Dates

Publication Date
20260505
Application Date
20211227

Claims (9)

  1. 1. The coal and gas outburst early warning method based on the on-site real-time data driving is characterized in that the on-site real-time data driving based coal and gas outburst early warning method continuously obtains real-time data of gas and wind speed environments through a gas concentration sensor and a wind speed sensor which are arranged on a tunneling working face, and calculates the gas emission quantity; Deep analysis is carried out on the gas emission quantity, coal and gas outburst early warning indexes combined based on dynamic change of gas concentration and reverse identification based on gas emission quantity prediction evaluation indexes are obtained, and moving average line, deviation rate, discrete rate, fluctuation rate, root mean square error and average absolute percentage error early warning indexes are established; Determining the weight of each index by an analytic hierarchy process, constructing a fuzzy comprehensive early warning model, calculating the differential comprehensive index of each index, and carrying out coal and gas outburst early warning; the coal and gas outburst early warning method based on-site real-time data driving specifically comprises the following steps: step one, continuously collecting and monitoring gas concentration and wind speed data of a return airway of a tunneling working face, and calculating gas emission quantity by combining the cross section area of the return airway; step two, carrying out five-point three-time smoothing treatment on the obtained gas emission quantity to remove interference data; step three, extracting a moving average line, a deviation rate, a discrete rate and a fluctuation rate of a gas emission quantity change value by adopting a statistical method as characteristic indexes; step four, data of a methane sensor and a wind speed sensor of a tunneling working face for 2 months in normal tunneling production of the coal mine are obtained according to a formula Calculating to obtain gas emission time sequence data, wherein Q s is roadway gas emission quantity, m 3 /min, C is roadway gas concentration,%, v s is roadway wind speed, m/S, S is the cross section area of a measuring point roadway, m 2 , inputting the calculated gas emission time sequence data into a bidirectional long and short time memory circulating neural network model for training and optimizing, so that model training errors are minimum, establishing a data model under normal production conditions, wherein the data model is a bidirectional long and short time memory circulating neural network model with the minimum errors, and parameters of the bidirectional long and short time memory circulating neural network model with the minimum training and establishing errors are set as follows: batch data at a time the amount batch_size=32; Learning rate learning u ratio=1e -4 ; Number of data executions epoch=20; data division ratio split data_ratio=0.833; The number of steps lag_num=16; the dimension input_dim_size=16 of the input value; The predicted output value dimension output_dim_size=1; The hidden layer dimension hidden_dim_size=64; LSTM layer number LSTM _layer_nums=1; Step five, collecting the calculated gas emission data in real time as a test data set, inputting the test data set into a data model in the step four to obtain a predicted value of the gas emission, calculating a loss value by utilizing the difference between the predicted value and a true value, and taking the root mean square error and the average absolute percentage error of the loss value as early warning indexes of coal and gas outburst; Step six, comparing the indexes in pairs according to the occurrence times of documents according to an analytic hierarchy process, wherein a ij is an importance comparison result of an element i and an element j, constructing a matrix by adopting a 1-9 scale method, namely a judgment matrix, calculating a maximum characteristic value lambda max and an index weight vector W, judging the consistency test of the matrix, determining the weight of each index, determining the membership function of each index through experimental data according to the principle of determining the membership function, establishing a fuzzy comprehensive judgment set by utilizing a fuzzy comprehensive judgment model, and representing the calculated judgment value as the differential comprehensive index of each index; And seventhly, carrying out data analysis on the gas emission quantity in the normal production period of the coal mine by utilizing the index differential comprehensive indexes, multiplying the maximum value of the calculated result by a safety coefficient to determine a differential threshold value, and judging the calculated index differential comprehensive index sequence by utilizing the threshold value to determine the outburst risk degree of the coal and the gas.
  2. 2. The method for early warning of coal and gas outburst based on-site real-time data driving according to claim 1, wherein in the second step, a five-point three-time smoothing method adopts a time series smoothing method, and the time series smoothing method utilizes a three-time least square polynomial to carry out smoothing, and the method specifically comprises the following steps: Has n+1 uniform sampling points Y 1 ,Y 2 ,…… ,Y n-1 ,Y n with sampling interval of h for transformation ,i=0,1,2......,n; The original node becomes μ 0 =0,μ 1 =1,μ 2 =2,…,μ n-1 =n-1,μ n =n; fitting data Y i by using m-degree polynomials, and setting a fitting polynomial as follows: , The undetermined coefficient a i in the fitting polynomial is determined by least squares method, the sum of variances is: , if F (a 0 ,a 1 ,a 2 ,…… ,a m ) is minimized, the minimum for a j (j=0, 1,2, third, m) partial derivative is calculated, there are , After finishing, the method comprises the following steps: , Substituting n=4, m=3 into the formula Solving for a j , j=0, 1,2,3, and substituting a j into the formula ; Is a five-point cubic smoothing formula, Is a smoothed value of Y i .
  3. 3. The method for early warning of coal and gas outburst based on-site real-time data driving according to claim 1, wherein in the third step, the method for extracting moving average line, deviation rate, discrete rate and fluctuation rate of the gas emission amount change value by adopting a statistical method as characteristic indexes specifically comprises: 1) Extracting a moving average line of a gas emission quantity change value to obtain a change trend of gas concentration time series data, wherein a gas concentration data average value calculation formula is as follows: In which, in the process, N is the number of records of the sequence in a certain time period, C i is the gas concentration of the ith record in the sequence; 2) Extracting the deviation rate of the gas emission quantity change value to obtain the quantity of the gas concentration moving average value of the gas concentration real-time sequence data which deviates from the time period, wherein the deviation rate calculation formula is as follows: wherein C t represents the gas concentration at time t; Mean value of n gas concentration data; 3) Extracting the discrete rate of the gas emission quantity change value to obtain the discrete degree of a gas concentration signal sequence, wherein the discrete rate calculation formula is as follows: Wherein, C t represents the gas concentration at the moment t, m is the sequence record number in a certain time period; 4) Extracting fluctuation rate of the gas emission quantity change value to obtain actual condition of gas concentration change, wherein the fluctuation rate extraction comprises amplitude change rate calculation and frequency change rate calculation, The amplitude change rate calculation formula R OSC is: Wherein: is the rate of change of amplitude over a period of n; , respectively a maximum value and a minimum value of the gas concentration in the time interval; The frequency change rate calculation formula R OFC is: Wherein: the gas concentration from the time t-n to the time t; indicating the number of calculated changes.
  4. 4. The on-site real-time data-driven coal and gas outburst prevention method according to claim 1, wherein in the fourth step, the bidirectional long-short-time memory cyclic neural network model comprises: Eight weights are cyclically used at each moment, namely the weight input to the forward and backward hidden layers (W 1 、W 4 ), the weight of the forward hidden layer (W 2 、W 3 ), the weight of the backward hidden layer (W 6 、W 7 ), the weight from the forward and backward hidden layers to the output layer (W 5 、W 8 ), and the forward hidden layer The calculation formulas of the backward hidden layer h t and the output layer o t are respectively as follows: 。
  5. 5. The on-site real-time data-driven coal and gas outburst early warning method according to claim 1, wherein the five root mean square error calculation formula is as follows: , the average absolute percentage error calculation formula is as follows: Wherein C i represents the true gas concentration; Indicating the predicted gas concentration and n indicating the number of samples.
  6. 6. The method for early warning of coal and gas outburst based on-site real-time data driving according to claim 1, wherein in the sixth step, determining each index weight specifically comprises constructing a judgment matrix A by adopting a 1-9 scale method, calculating a maximum eigenvalue lambda max and an index weight vector W, judging consistency check of the matrix, wherein the consistency check judgment matrix expression is: Wherein, C 1 is the consistency proportion, n is the number of comparison indexes, R 1 is the average random consistency index, and when C 1 is less than 0.1, the constructed judgment matrix meets the requirement.
  7. 7. The method for on-site real-time data-driven coal and gas outburst prevention according to claim 1, wherein in the sixth step, the membership function determination comprises: Moving average line by using real-time gas concentration moving average line and gas concentration average of previous period Dividing the ratio of (2); dividing a value interval by comparing the real-time deviation rate, the discrete rate and the average value of the calculation period; the fluctuation rate is compared with the average value of the calculation period by utilizing the real-time fluctuation rate to divide a value interval; dividing a value interval by utilizing the ratio of the estimated value of the gas concentration sequence predicted result before the real-time accident to the estimated average value of the normal production gas concentration sequence predicted result; 。
  8. 8. the method for early warning of coal and gas outburst based on-site real-time data driving according to claim 1, wherein in the sixth step, the establishing of the fuzzy comprehensive evaluation model comprises: a. establishing a set of factors The factor set U is expressed as Wherein element u i represents an influencing factor, wherein ; B. establishing an evaluation set The evaluation set V is expressed as Wherein element v j represents the result of the evaluation, wherein Taking v= { accident occurred, no accident occurred }, ; C. establishing a weight set Each element in the factor set U has different importance degrees in the judgment, different weight sets W are given to each element U i according to the importance degrees, and W is a fuzzy subset on the factor set U and expressed as Weight set is In the time-course of which the first and second contact surfaces, ; D. single factor fuzzy evaluation The object to be evaluated evaluates the ith factor u i in the factor set, and if the membership degree of the jth element v j in the evaluation set is r ij , the evaluation result is expressed as , Representing a single factor evaluation set; e. Fuzzy comprehensive judgment Constructing a multi-factor comprehensive judgment matrix R by a single-factor judgment set, And adopting a weighted average model, and obtaining a fuzzy comprehensive judgment set B according to multiplication of a fuzzy matrix, wherein the fuzzy comprehensive judgment set B is: ; Wherein b i is a fuzzy comprehensive judgment index, wherein, Representing the membership degree of the evaluation object to the ith element in the evaluation set V under the condition of comprehensively considering all influence factors.
  9. 9. The coal and gas outburst early warning method based on-site real-time data driving according to claim 1 is characterized in that in the seventh step, the value of a fuzzy comprehensive evaluation set B is used for representing the differential comprehensive index of each index, the root mean square error and the average absolute percentage error of the evaluation index are predicted through a model, coal and gas outburst evaluation and early warning are carried out by combining with the characteristic index of dynamic change of gas concentration, when the coal mine is produced normally, the differential comprehensive index of each index fluctuates in a normal range, orange early warning is carried out when the index reaches the maximum value in the normal range, and when the differential comprehensive index of each index approaches to 1, the probability of occurrence of coal and gas outburst is larger, and red early warning is carried out.

Description

Coal and gas outburst early warning method based on-site real-time data driving Technical Field The invention belongs to the technical field of coal mine safety production warning, and particularly relates to a coal and gas outburst warning method based on-site real-time data driving. Background At present, because the coal and gas outburst accident still continuously occurs, the coal and gas outburst early warning model mainly selects indexes such as gas content, gas pressure, gas emission quantity, drilling cutting method, electromagnetic radiation method and the like, and a good prediction effect can be obtained theoretically. The gas content and gas pressure in the main control factors in the coal and gas outburst early warning model cannot be obtained in real time, and the existing mechanism model is easily influenced by manual operation and underground geological environment by adopting a drilling cutting method, an electromagnetic radiation method and the like, so that the prediction result has certain hysteresis and larger error. These situations lead to that the model can obtain better prediction effect in theory, but in practical application, the prediction result is often limited by the real-time property of data acquisition, so that the on-site decision cannot be guided effectively. The gas concentration is analyzed and predicted singly by a statistical method or an artificial intelligence algorithm, so that the respective limitations exist, and the satisfactory effect cannot be achieved. Through the analysis, the problems and defects of the prior art are that most of the prior art tries to analyze the evolution rule before disaster based on the coal and gas outburst mechanism, and the risk degree of the data sequence can be acquired through the research of the evolution rule so as to realize early warning. The data needed to be collected for risk study and judgment based on the evolution rule are various, most of the data are not easy to acquire in real time, and meanwhile the data are easily influenced by manual operation and underground geological environment, so that a prediction result has certain hysteresis and poor existence accuracy, and the effectiveness of early warning is difficult to guarantee. The difficulty in solving the problems and the defects is that a large number of online sensors for acquiring relevant parameters of an evolution mechanism are required to be developed so as to ensure the accuracy and the effectiveness of the prediction of the technical route. The method has the advantages that the gas emission quantity can represent the difference of different stages of the coal and gas outburst whole process, the gas emission quantity change rule during normal coal mine tunneling implicitly highlights the gas emission quantity change rule of the inoculation stage, if abnormal changes against the rule can be found out, the coal mine tunneling working face can be determined to be in the outburst formation and development stage in the period, and the abnormal change characteristics of the gas emission quantity are excavated by combining statistical indexes with artificial intelligent model indexes, so that the coal and gas outburst risk degree is determined. The gas emission amount can be obtained by acquiring and calculating the existing gas concentration and wind speed on-line sensor in real time, is not interfered by manpower, and solves the problem of poor accuracy and effectiveness of the traditional method. Disclosure of Invention In order to overcome the problems in the related art, the embodiment of the invention provides a coal and gas outburst early warning method based on-site real-time data driving. In particular to a coal and gas outburst early warning method based on-site real-time data driving based on abnormal characteristics of gas emission amount time sequence data. The technical proposal is that a coal and gas outburst early warning method based on-site real-time data driving, The method comprises the steps of continuously obtaining real-time data of gas and wind speed environments through a gas concentration sensor and a wind speed sensor which are arranged on a tunneling working face, calculating the gas emission quantity, carrying out deep analysis on the gas emission quantity, obtaining coal and gas outburst early warning indexes based on combination of dynamic change of the gas concentration and reverse identification of gas emission quantity prediction evaluation indexes, establishing moving average line, deviation rate, discrete rate, fluctuation rate, root mean square error (Root Mean Square Error, RMSE) and average absolute percentage error (mean absolute percentage, MAPE) early warning indexes, determining various index weights through a hierarchical analysis method (AnalyticHierarchyProcess, AHP), constructing a fuzzy comprehensive early warning model, calculating various index differential comprehensive indexes, and carrying out coal and gas outbu