Upper corner gas outburst early warning method based on contrast-region uniformity map

Through contrast-region uniformity graph analysis, the gas immersion graph is generated using the space-time Kriging interpolation method and the Matron model, which solves the problem of early warning difficulty of gas outbursts in corner corners, improves early warning accuracy and safety, and reduces costs .

CN118911769APending Publication Date: 2024-11-08SHANDONG UNIV OF SCI & TECH
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202410952909.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-16
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

During coal mining, it is difficult to accurately predict the early warning of corner gas outbursts. The existing methods have problems such as overfitting, relying on professional knowledge to set parameters, and complex data-driven methods to calculate, resulting in high warning accuracy and cost.

Method used

Using the early warning method based on the contrast-region uniformity graph, a gas data point interpolation model is constructed through the spatiotemporal Kriging interpolation method, and a gas sampling point is fitted using the spatiotemporal variogram and the Matron model to generate a comprehensive gas contrast ratio of the surface corner- Regional uniformity chart, evaluate gas concentration and set risk levels, and initiate prevention measures.

Benefits of technology

It realizes timely identification and early warning of gas outburst risks, improves coal mining safety, has high warning accuracy of risk level, and reduces early warning costs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118911769A_ABST
    Figure CN118911769A_ABST
Patent Text Reader

Abstract

The invention discloses an upper corner gas outburst early-warning method based on a contrast-region uniformity graph, and the method comprises the following steps: S1, constructing an upper corner gas data point interpolation model of a fully mechanized caving face by using the gas concentration data of a known gas sampling point of the fully mechanized caving face and the spatial position information of the gas sampling point through a space-time Kriging interpolation method, acquiring the position information of an unknown gas sampling point through an upper corner gas data point interpolation model of the fully mechanized caving face; s2, representing space-time relevance among the gas sampling points according to the space-time variation function, and estimating space-time information of an unknown gas sampling point; s3, fitting all the gas sampling points by using a Matter model to obtain a gas contrast-region uniformity graph of the corner of the fully mechanized caving face; and S4, determining the gas concentration of the corner of the fully-mechanized caving face through the gas contrast-region uniformity graph of the corner of the fully-mechanized caving face, evaluating the gas concentration of the corner of the fully-mechanized caving face, determining the risk level of the gas concentration, and starting corresponding gas outburst prevention measures.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the field of coal mining, and in particular to an upper corner gas outburst early warning method based on a contrast-region uniformity diagram. Background Art

[0002] The fully mechanized caving face coal mining technology is a coal mining method with high frequency and good benefits. With the increase in people's demand for coal use, the depth of coal mining has gradually deepened, and the mining environment has become more complex, which makes the fully mechanized caving face face safety production face more challenges. Among them, the most notable is the problem of corner gas outburst. Corner gas outburst will not only reduce the oxygen content of the working environment and suffocate people, but also may cause secondary disasters such as fire and coal dust explosion, which will bring great safety hazards to coal mine production. Therefore, the research on the early warning of corner gas outburst on the fully mechanized caving face is of great significance to ensure the safe production of coal mines and improve the efficiency of coal mining. In the process of fully mechanized caving face construction, due to the complex and changeable factors such as the occurrence state of coal seams, gas occurrence laws, and geological structures, the uneven and unpredictable distribution of gas increases the difficulty and uncertainty of early warning, making it impossible to timely warn of gas outbursts in some corner areas.

[0003] In order to effectively prevent and respond to gas outburst disasters, domestic and foreign scholars have conducted a lot of research and practice on gas outburst prediction. Yan Xin et al. published a study on coal and gas outburst warning in mines based on image recognition. They obtained the current gas concentration data and gas outburst hazard images through the VGG16 network, and used evidence theory to judge whether there is a gas outburst risk and risk level in the current corner. This method can effectively improve the warning speed, but the VGG16 network involves a large number of parameters. When the training data is small, the model is prone to overfitting, which affects the accuracy of the warning; Heng Xianwei et al. published a method for early warning of microseismic dynamic response and multi-parameter indicators of outburst coal seams. They used microseismic monitoring technology and combined spatiotemporal analysis to construct a gas outburst warning model, and divided its hazard level through joint analysis of multiple indicator parameters. The risk level of gas outburst is determined by analyzing the correlation between microseismic indicators and conventional prediction indicators. However, the setting and adjustment of the model parameters need to be based on practical experience and professional knowledge. If the practical experience is insufficient or the professional knowledge and skills are poor, the warning results will be affected. Sun Jiping et al. published a study on the perception and alarm method of coal mine rock burst and coal and gas outburst. According to the principle of gas outburst, the risk thresholds of each level are set, and infrared thermal imagers, laser radars, methane sensors and other equipment are used to monitor the changes in the concentration of even-corner gas. By analyzing the monitoring data, the risk level of gas is determined, and corresponding defense measures are initiated accordingly. This method can achieve efficient early warning according to the distribution of gas concentration in time and space, but due to the large number of equipment used in the process of gas concentration collection, the early warning cost is relatively high. He Qiao et al. published a study on the coal mine gas anomaly early warning model based on mathematical statistics. They first studied the characteristics of gas concentration changes and used data-driven design to design a corner gas outburst early warning system. The system can quantitatively assess the risk based on the current gas concentration, fluctuation amplitude index and other parameters, and timely and effectively determine the warning level. However, the data-driven method involves a large amount of data and high computational complexity. If the data is incomplete or has quality problems, it will directly affect the accuracy of the gas outburst warning. Therefore, it is necessary to study an upper corner gas outburst warning method based on contrast-regional uniformity map to achieve timely identification and warning of gas outburst risks. Summary of the invention

[0004] The purpose of the present invention is to address the above-mentioned problems and to provide an upper corner gas outburst early warning method based on a contrast-regional uniformity diagram that effectively improves coal mining safety.

[0005] In order to achieve the above object, the technical solution of the present invention is:

[0006] An upper corner gas outburst early warning method based on contrast-region uniformity map comprises the following steps:

[0007] S1. Using the gas concentration data of known gas sampling points in the fully mechanized caving face and the spatial location information of the gas sampling points, a spatiotemporal Kriging interpolation method is used to construct an interpolation model for gas data points in the upper corners of the fully mechanized caving face, and the location information of unknown gas sampling points is obtained through the interpolation model for gas data points in the upper corners of the fully mechanized caving face;

[0008] S2, characterize the spatiotemporal correlation between gas sampling points according to the spatiotemporal variogram, and estimate the spatiotemporal information of unknown gas sampling points;

[0009] S3. Use the Mattelon model to fit all gas sampling points to obtain the gas contrast-regional uniformity diagram of the fully mechanized caving face corners;

[0010] S4. Clarify the gas concentration in the corners of the fully-mechanized caving surface through the gas contrast in the corners of the fully-mechanized caving surface - regional uniformity diagram, evaluate the gas concentration in the corners of the fully-mechanized caving surface, determine its risk level, and initiate corresponding gas outburst prevention measures.

[0011] Furthermore, in step S1, the formula of the gas data point interpolation model at the upper corner of the fully mechanized caving surface is:

[0012]

[0013] Where: x a * is the predicted value of the gas concentration at gas sampling point a, n is the total number of gas sampling points; x b is the observed value of gas content at a known gas sampling point, where b is the index, representing the bth gas sampling point in the data set; w b For x b The weight is the weight coefficient associated with the bth gas sampling point.

[0014] Furthermore, the weight w b The calculation process includes the following steps:

[0015] S11. Calculate the unbiased estimation constraint condition and the optimization objective function, the formula is:

[0016] E(x a * -x a )=0;

[0017] Q = Var(x a * -x a );

[0018] Where: E is the expected value; x a is the actual value of gas concentration at gas sampling point a, x a *is the predicted value of gas concentration at gas sampling point a; Var is the variance, indicating x a * and x a The degree of dispersion between

[0019] S12. Use the semivariance function to evaluate the degree of connection between variables. The calculation formula is:

[0020]

[0021] Where: b is a gas sampling point at a certain position on the fully mechanized caving surface, j is any gas sampling point on the fully mechanized caving surface except b; x b is the observed value of gas content at a known gas sampling point; x j is the observed value of gas content at any gas sampling point on the fully mechanized caving face except b; r bc is the semivariance function, reflecting the similarity at different distances, r bc The lower the value, the higher the similarity, indicating that there is a stronger correlation between the gas content observations at the two measurement points, and vice versa;

[0022] S13, combining the formula in step S11 and the formula in step S12 to obtain the weight w b The calculation formula is:

[0023]

[0024] Where: r is the degree of correlation between the gas concentration value of the gas sampling point represented by each subscript and the gas concentration value of different b points; w i is the weight of the actual value of gas concentration at gas sampling point i, r ij is the semivariance between two gas sampling points i and j, r i0 It is the weighted sum of squared differences, which is used in Kriging interpolation to measure the smoothness of the interpolation surface relative to the original data, helping to ensure that the interpolation result does not deviate too far from the known data points, thereby maintaining the rationality of the interpolation result.

[0025] Furthermore, in step S2, the temporal and spatial information calculation formula of the unknown gas sampling point is:

[0026]

[0027] Where: (x0, t0) is the time-space point to be predicted or estimated, x0 is the spatial position of the gas sampling point, and t0 is the temporal coordinate of the gas sampling point; X(x0, t0) is the estimated value of the time-space point (x0, t0), that is, the target point to be predicted or interpolated; w nis the weight coefficient, which indicates the contribution of each known gas sampling point to the target point estimate, where n represents the number of known gas sampling points involved in the weighting; X(x b ,t b ) is the value of the known gas sampling point, where x b and t b Respectively represent the spatial coordinates and time coordinates of the known gas sampling points;

[0028] Solving the above formula using the Lagrange factor μ yields:

[0029]

[0030] Where: γ is the spatiotemporal variation function.

[0031] Furthermore, the calculation process of the spatiotemporal variation function includes the following steps:

[0032] S21. Calculate the spatiotemporal empirical variation function and the spatiotemporal covariance function, and the calculation formula is:

[0033]

[0034] C(h s ,h t )=Cov(X(x+h s ,t+h t ),X(x,t));

[0035] Where: γ(h s ,h t ) is the spatial lag h s (representing the distance between two data points in space) and the time lag h t The semivariogram function under (representing the time gap between two data points) is used to measure the degree of change between data points under a specific spatial and temporal lag; n1(h s ,h t ) represents the number of pairs of data points used to calculate the semivariance when there is spatial lag and time lag, that is, how many pairs of data points can be used to calculate the semivariance; X(x b ,t b ) is the spatial position x b and time t b The observed value at X(x b +h s ,t b +h t ) is the spatial position x b Plus the spatial lag h s and time t b Add the time lag h tThe position observations are used to quantify the degree of change of the observations at given spatial and temporal lags, thus providing key spatial and temporal correlation information for estimating the values ​​of unknown points.

[0036] S22. Calculate the spatiotemporal variation function, and its calculation formula is:

[0037]

[0038] Where: 2 is the variance of X(x,t).

[0039] Furthermore, in step S3, the fitting formula for fitting all gas sampling points using the Mattelon model is:

[0040]

[0041] Where: C0 is the nugget value; C0+C is the base value; C is the block arch height; e is the range;

[0042] When h=0, the variation of the gas sampling points is only determined by the nugget effect; when 0<h≤e, the correlation of gas concentrations between gas sampling points begins to weaken with the increase of spatial distance; when h>e, the correlation of gas concentrations between gas sampling points reaches a stable value.

[0043] Furthermore, in step S4, when evaluating the gas concentration in the corners of the fully-mechanized caving surface, the influencing factors of the gas outburst in the corners are first found; then, the early warning evaluation index is calculated according to the weight of the influencing factors in each gas sampling point and the early warning evaluation value of the influencing factors, and the risk level of the gas outburst warning in the corners of the fully-mechanized caving surface is obtained by grading the early warning evaluation index.

[0044] Furthermore, the early warning evaluation indicators include corner gas indicators and coal quality indicators.

[0045] Compared with the prior art, the present invention has the following advantages and positive effects:

[0046] The present invention proposes a method for early warning of gas outburst in the upper corner of a fully-mechanized caving surface based on a contrast-regional uniformity map. The method firstly constructs an interpolation model of gas data points in the upper corner of the fully-mechanized caving surface by using an empty Kriging interpolation method, and obtains information of gas sampling points by using known data of the fully-mechanized caving surface; then, the spatiotemporal correlation between the gas sampling points is characterized according to the spatiotemporal variation function and the spatiotemporal covariance function, and the spatiotemporal information of the unknown gas sampling points is estimated; then, the Mattelon model is used to fit all sampling points to obtain a gas contrast-regional uniformity map of the fully-mechanized caving surface; finally, the gas concentration in the corner of the fully-mechanized caving surface is clarified by the contrast-regional uniformity map, and its risk level is set according to historical data analysis and expert experience, and its corresponding risk level is calculated by using a weighted average method, and a corresponding early warning plan is initiated to realize timely identification and early warning of gas outburst risks; the method can accurately reflect the temporal and spatial variation characteristics of the gas concentration data of the fully-mechanized caving surface, and its risk level early warning accuracy is high, which improves the safety of coal mining operations in the fully-mechanized caving surface, and has excellent market promotion value. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0048] Figure 1 This is the gas early warning analysis diagram of the upper corner of the fully mechanized caving face;

[0049] Figure 2 This is a schematic diagram of the prediction results of corner gas concentration. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work, any modifications, equivalent substitutions, improvements, etc., should be included in the protection scope of the present invention.

[0051] The present invention discloses a high-efficiency and high-precision gas outburst early warning method for the upper corner of a fully-mechanized caving surface, namely, contrast-regional uniformity map analysis. The method generates a gas concentration data contrast-regional uniformity image of the upper corner of a fully-mechanized caving surface through a spatiotemporal Kriging interpolation method and a Matlon model, sets its risk level based on historical data analysis and expert experience, calculates its corresponding risk level through a weighted average method, activates a corresponding early warning plan, and realizes timely identification and early warning of gas outburst risks.

[0052] 1. Contrast of gas in the upper corner of fully mechanized caving face - generation of regional uniformity map

[0053] 1. Gas regionalization variable analysis based on spatiotemporal kriging interpolation method

[0054] The contrast-regional uniformity map can clearly show the distribution and variation characteristics of the gas concentration in the corners of the fully-mechanized caving surface. By comparing the gas concentration in different areas, it can be judged whether there is a gas outburst problem and the risk level. At present, the methods for generating contrast-regional uniformity maps of gas in the corners of fully-mechanized caving surfaces include grayscale histograms, image processing, and spatiotemporal distribution. Among them, the spatiotemporal distribution method can intuitively display the spatial distribution and variation of the gas concentration in the corners through color, brightness or other visual elements, and then quickly identify areas with higher gas concentrations and the contrast and uniformity differences between different areas. [8] For this purpose, a contrast-area uniformity map of the corner gas is generated using this method.

[0055] Among the time-space distribution methods, Kriging interpolation is a commonly used technique that can effectively use known gas concentration data and their spatial relationships to estimate unknown gas concentration data. This method can more accurately and intuitively understand the state of corner gas outbursts. The estimation process of the sampling point is as follows:

[0056]

[0057] Where: x a * is the predicted value of the gas concentration at gas sampling point a, n is the total number of gas sampling points; x b is the observed value of gas content at a known gas sampling point, where b is the index, representing the bth gas sampling point in the data set; w b For x b The weight is the weight coefficient associated with the bth gas sampling point.

[0058] Among them, w b According to the unbiased estimation constraint formula (2) and the optimization objective function formula (3), we have:

[0059] E(x a* -x a )=0; (2)

[0060] Q = Var(x a * -x a ); (3)

[0061] Where: E is the expected value; x a is the actual value of gas concentration at gas sampling point a, x a * is the predicted value of gas concentration at gas sampling point a; Var is the variance, indicating x a * and x a The degree of dispersion between

[0062] In order to select the appropriate data sampling density to obtain a better spatial data representation of corner gas, the semivariogram function is used to evaluate the degree of connection between variables, that is:

[0063]

[0064] Where: b is a gas sampling point at a certain position on the fully mechanized caving surface, j is any gas sampling point on the fully mechanized caving surface except b; x b is the observed value of gas content at a known gas sampling point; x j is the observed value of gas content at any gas sampling point on the fully mechanized caving face except b; r bc is the semivariance function, reflecting the similarity at different distances, r bc The lower the value, the higher the similarity, indicating that there is a stronger correlation between the gas content observations at the two measurement points, and vice versa;

[0065] Arranging formula (2)-formula (4) can yield w b The mathematical expression of is:

[0066]

[0067] Where: r is the degree of correlation between the gas concentration value of the gas sampling point represented by each subscript and the gas concentration value of different b points; w i is the weight of the actual value of gas concentration at gas sampling point i, r ij is the semivariance between two gas sampling points i and j, r i0 It is the weighted sum of squared differences, which is used in Kriging interpolation to measure the smoothness of the interpolation surface relative to the original data, helping to ensure that the interpolation result does not deviate too far from the known data points, thereby maintaining the rationality of the interpolation result.

[0068] The Kriging interpolation method mainly focuses on the local spatial correlation of the data, but does not fully consider the overall spatial correlation. When there are fewer corner gas data points, the generated contrast-regional uniformity map will have a large estimation error. In order to address this problem, the time dimension is expanded [9] That is, when interpolating unknown gas data points in the fully mechanized caving face, the data distribution and correlation in space and time are considered at the same time to improve the estimation accuracy of gas data points. The expanded mathematical expression is:

[0069]

[0070] Where: (x0, t0) is the time-space point to be predicted or estimated, x0 is the spatial position of the gas sampling point, and t0 is the temporal coordinate of the gas sampling point; X(x0, t0) is the estimated value of the time-space point (x0, t0), that is, the target point to be predicted or interpolated; w n is the weight coefficient, which indicates the contribution of each known gas sampling point to the target point estimate, where n represents the number of known gas sampling points involved in the weighting; X(x b ,t b ) is the value of the known gas sampling point, where x b and t b Respectively represent the spatial coordinates and time coordinates of the known gas sampling points;

[0071] Formula (6) can be solved using the Lagrangian factor μ. The calculation process is:

[0072]

[0073] Where: γ is the spatiotemporal variation function.

[0074] Among them, γ can characterize the variation structure and spatiotemporal continuity of spatiotemporal variables. In spatiotemporal Kriging interpolation, γ can provide a mathematical model of the relationship between data points. This model is based on a key assumption, namely, spatiotemporal stationarity. Under the assumption of spatiotemporal stationarity, it is believed that the statistical characteristics of spatiotemporal variables are constant, and these characteristics are only related to the relative spatiotemporal distance between gas sampling points and the spatiotemporal covariance function. Based on this assumption, there is a mapping relationship between the spatiotemporal variation function and the spatiotemporal covariance function, that is, each spatiotemporal variation function corresponds to a specific spatiotemporal covariance function, and vice versa. Among them, the spatiotemporal empirical variation function and the spatiotemporal covariance function are shown in formulas (8) and (9).

[0075]

[0076] C(h s ,h t )=Cov(X(x+h s ,t+ht ),X(x,t)) ; (9)

[0077] Where: γ(h s ,h t ) is the spatial lag h s (representing the distance between two data points in space) and the time lag h t The semivariogram function under (representing the time gap between two data points) is used to measure the degree of change between data points under a specific spatial and temporal lag; n1(h s ,h t ) represents the number of pairs of data points used to calculate the semivariance when there is spatial lag and time lag, that is, how many pairs of data points can be used to calculate the semivariance; X(x b ,t b ) is the spatial position x b and time t b The observed value at X(x b +h s ,t b +h t ) is the spatial position x b Plus the spatial lag h s and time t b Add the time lag h t The position observations are used to quantify the degree of change of the observations at given spatial and temporal lags, thus providing key spatial and temporal correlation information for estimating the values ​​of unknown points.

[0078] According to formula (9), C(h s ,h t ) is only related to the distance and has nothing to do with the time and space position. The expected invariance and cosquare matrix of X(x,t) Under the premise of positive definiteness, the calculation process of the space-time theoretical variation function is:

[0079]

[0080] Where: 2 is the variance of X(x,t);

[0081] The positive definite condition is a certain e b ∈R(b=1,2,...,n) needs to meet the conditions of formula (11).

[0082]

[0083] Where: e is the variable range.

[0084] The variation range refers to the range in which the gas regionalization variable maintains a certain degree of correlation in time and space, that is, the relative distance before the variation function reaches the base value.

[0085] 2. Contrast-regional uniformity map generation based on spatiotemporal theoretical variogram

[0086] In order to generate the contrast-regional uniformity map of gas outburst in the upper corner of the fully mechanized caving face, the Mattelon model is used to fit the gas sampling data points to obtain the concentration and distribution law of gas outburst. The fitting mathematical expression is:

[0087]

[0088] In the formula: C0 is the nugget value; C0+C is the base value; C is the block arch height.

[0089] Among them, the nugget value refers to the size of spatial variation that may exist within a very short distance. When the nugget value is large, it may mean that there is a gas protrusion problem; the base value is used to measure the magnitude of the change in regionalized variables; the block arch height indicates the degree of gas protrusion.

[0090] According to formula (12), there are three possible situations for the fitting of gas sampling data points: when h = 0: the value of Y(h) is C0, which means that at a distance of almost zero (i.e., the same position or a very close position), the variation of gas data points is only determined by the nugget effect (microscopic variation or random error); when 0 < h ≤ e: within this range, as the distance h increases, Y(h) gradually increases, but the growth rate gradually slows down, which means that as the spatial distance increases, the correlation between gas concentrations begins to weaken, but the weakening speed gradually slows down; when h > e: when the distance h reaches or exceeds the range e, Y(h) reaches a stable value, i.e., the base value C0+C. This means that after exceeding the range, the correlation between data points is very weak or non-existent, and the variation reaches a stable level. Under this condition, the gas variation is determined by C0+C.

[0091] Through the above process, the sampling data points of various situations of corner gas outburst can be fitted to generate a smooth interpolation surface, that is, the contrast-regional uniformity map of corner gas outburst on the fully-mechanized caving surface. The concentration and change characteristics of corner gas outburst can be understood based on the information in this map.

[0092] 2. Early warning of gas outburst in the upper corner of the fully mechanized caving face

[0093] The establishment of the gas outburst warning model in the upper corner of the fully mechanized caving face is divided into three parts, namely, the indicator data set, the warning result evaluation set and the warning evaluation standard.

[0094] 1. Indicator Dataset

[0095] The index set is a combination of factors for evaluating corner gas outburst, represented by O = {O1, O2, ..., O m} represents, m represents the number of factors affecting gas outburst warning. The outburst warning indicator database is constructed using gas concentration indicators and geological structure indicators (including geological information such as coal seam thickness, dip, and strike).

[0096] 2. Highlight the early warning result evaluation set

[0097] The early warning result judgment set is generated according to each outburst prediction result, and is represented by L. In the early warning of gas outburst in the upper corner of the fully mechanized caving face, there are three situations, namely danger, threat, and normal, which are represented by the numbers 3, 2, and 1 respectively, so the early warning result evaluation set is L = {1, 2, 3}.

[0098] 3. Corner gas outburst warning assessment standard

[0099] In formulating the evaluation criteria, we should first consider the individual early warning indicator factors, assuming that the impact factor O k When judging, construct O k The early warning evaluation set P. Since P may exist in three situations, P can be expressed as P = Y(h) {P k,1 ,P k,2 ,...,P k,3}. The final evaluation index is calculated using the single factor weight set W and the weighted average method, and then:

[0100] K=W×P; (13)

[0101] Since the corner gas index and coal quality index are key factors in evaluating the safety status of the fully mechanized caving face in a mine, the changes of these two indicators will directly affect the stability of the working face. For this reason, these two indicators are set as the influencing factors of the evaluation, and the outstanding warning results are divided into two categories, namely, state warning and trend warning (state warning mainly focuses on the safety status of the current working face, while trend warning focuses on predicting the safety trend of the working face in the future). Then, according to the specific values ​​of gas index A and coal quality index B, the warning results are divided into different levels. These levels are based on a large amount of historical data analysis and expert experience. That is, the risk warning levels corresponding to gas indexes (0, 0.7), (0.7, 1.0), and (1.0, +∞) are normal, threatening, and dangerous, respectively, and the risk warning levels corresponding to coal quality indexes (0, 0.6), (0.6, 0.8), and (0.8, +∞) are normal, threatening, and dangerous, respectively. Each level corresponds to a different safety situation, and corresponding countermeasures have been formulated. For details, see Table 1.

[0102] Table 1 Corner gas outburst warning assessment standards

[0103]

[0104]

[0105] 3. Experiment

[0106] The complexity of the geological structure of the coal seam in the No. 4 mine of a coal mining enterprise mainly includes faults, folds, etc. Once the external force reaches a certain intensity, it will break and the corner gas outburst problem often occurs. Therefore, the No. 4 mine is taken as the research object. This mine mainly produces bright coal and dark coal. The coal seam thickness is about 0.97m, the inclination angle range is [5°, 50°], the distance between adjacent mining coal seams is about 5.94m, and the average gas concentration is 6.637m 3 / t. The experiment used FW-2 high negative pressure gas sampling device, CFJD5 wind speed and direction sensor, GWK42 (GWK40) temperature sensor to collect key data such as gas concentration, wind speed, temperature, etc. of the fully mechanized caving face.

[0107] In order to verify the effect of contrast-regional uniformity map analysis in the early warning of gas outburst in the corner of the fully mechanized caving face, an event was randomly selected from the historical corner gas outburst events of No. 4 Mine as the research object. The proposed method was used to generate a contrast-regional uniformity map and give an early warning. The results are shown in Figure 2. Figure 1 shown.

[0108] according to Figure 1 The results show that from the contrast dimension, there are obvious differences in gas concentration in different areas. Specifically, the contrast index of the upper corner area of ​​the fully-mechanized caving surface is significantly higher than that of other areas, which clearly indicates that there is gas accumulation in this area. From the perspective of regional uniformity, the gas concentration inside the fully-mechanized caving surface is unevenly distributed, with high concentrations in some areas and low concentrations in some areas, showing obvious non-uniformity. This non-uniform distribution further suggests that there is a risk of local gas accumulation in this area. Based on the above analysis, it is confirmed that there is a gas outburst problem in the upper corner of the fully-mechanized caving surface, which verifies that the proposed method can achieve gas outburst warning.

[0109] The quality of the prediction results of corner gas concentration directly determines the effect of the warning of gas outburst in the corner of the fully mechanized caving face. Therefore, this experiment verifies the performance of the contrast-regional uniformity map in the warning of gas concentration in the corner of the fully mechanized caving face from the perspective of corner gas concentration prediction.

[0110] Contrast-regional uniformity map analysis, VGG16 network and machine learning methods were used to predict the gas concentration in the corner of a fully mechanized caving face in the No. 4 mine during the period of 14:00-14:14 on March 16, 2024. The results are as follows: Figure 2 shown.

[0111] according to Figure 2The analysis results show that the contrast-regional uniformity map shows that the predicted results of the gas concentration in the corners of the fully mechanized caving surface are basically consistent with the actual concentration. This is because it uses the spatiotemporal Kriging interpolation method for prediction, which fully considers the spatial correlation between the data points of the corner gas. By calculating the spatial weight and contribution rate between the interpolation point and the surrounding data points, an accurate and smooth interpolation result is obtained. At the same time, the spatiotemporal Kriging interpolation method also considers the time factor, combines the correlation between space and time, and more comprehensively reflects the change law of gas concentration. Therefore, the prediction results are closer to the actual situation and improve the prediction accuracy. In contrast, the prediction method of corner gas concentration using the VGG16 network and machine learning method is more complex, and it is easy to lose some sampling data point information during the prediction process, resulting in a slightly worse prediction result than the spatiotemporal Kriging interpolation method.

[0112] Ten historical gas cases in the upper corners of fully mechanized caving faces in No. 4 Mine were randomly selected, and contrast-regional uniformity map analysis, VGG16 network and machine learning methods were used to give prominent warnings to these cases. The results are shown in Table 2.

[0113] Table 2 Highlight warning results

[0114]

[0115] According to the results in Table 2, the contrast-regional uniformity diagram analysis shows that each corner gas outburst warning result is consistent with the actual situation. This is because the final evaluation index is calculated by the weighted average method. This method can comprehensively consider the impact of different factors on the corner gas warning level, effectively avoiding the one-sidedness and error that may be caused by a single factor evaluation. The accuracy of the warning level evaluation result is higher than that of the two compared methods.

[0116] IV. Conclusion

[0117] In the process of fully-mechanized top coal caving in coal mines, gas outburst in the upper corner is a serious safety hazard. In order to achieve effective early warning of gas outburst, the present invention proposes an early warning method for gas outburst in the upper corner of the fully-mechanized caving surface under contrast-regional uniformity map analysis, using spatiotemporal Kriging interpolation method, spatiotemporal variation function, and spatiotemporal covariance function to calculate and obtain the spatiotemporal information and concentration value of each data point, using the Mattelon model to fit these points, generating a gas contrast-regional uniformity map of the fully-mechanized caving surface corner, using the weight function and weighted average method to construct a corner gas outburst early warning model, and starting the corresponding prevention strategy according to the concentration of corner gas outburst, accurately reflecting the temporal and spatial variation characteristics of the gas concentration data of the fully-mechanized caving surface, and the risk level early warning accuracy is high.

Claims

1. A method for early warning of upper corner gas outburst based on contrast-regional uniformity diagram, characterized by: The following steps are involved: S1. Using the gas concentration data of known gas sampling points in the fully mechanized caving face and the spatial location information of the gas sampling points, a spatiotemporal Kriging interpolation method is used to construct an interpolation model for gas data points in the upper corners of the fully mechanized caving face, and the location information of unknown gas sampling points is obtained through the interpolation model for gas data points in the upper corners of the fully mechanized caving face; S2, characterize the spatiotemporal correlation between gas sampling points according to the spatiotemporal variogram, and estimate the spatiotemporal information of unknown gas sampling points; S3. Use the Mattelon model to fit all gas sampling points to obtain the gas contrast-regional uniformity diagram of the fully mechanized caving face corners; S4. Clarify the gas concentration in the corners of the fully-mechanized caving surface through the gas contrast in the corners of the fully-mechanized caving surface - regional uniformity diagram, evaluate the gas concentration in the corners of the fully-mechanized caving surface, determine its risk level, and initiate corresponding gas outburst prevention measures.

2. The upper corner gas outburst early warning method based on contrast-regional uniformity map according to claim 1, characterized in that: In step S1, the formula of the gas data point interpolation model at the upper corner of the fully mechanized caving surface is: Where: x a * is the predicted value of the gas concentration at gas sampling point a, n is the total number of gas sampling points; x b is the observed value of gas content at a known gas sampling point, where b is the index, representing the bth gas sampling point in the data set; w b For x b The weight is the weight coefficient associated with the bth gas sampling point.

3. The upper corner gas outburst early warning method based on contrast-regional uniformity map according to claim 2 is characterized in that: The weight w b The calculation process includes the following steps: S11. Calculate the unbiased estimation constraint condition and the optimization objective function, the formula is: E(x a * -x a )=0; Q=Var(x a * -x a ); Where: E is the expected value; x a is the actual value of gas concentration at gas sampling point a, x a * is the predicted value of gas concentration at gas sampling point a; Var is the variance, indicating x a and x a * The degree of dispersion between S12. Use the semivariance function to evaluate the degree of connection between variables. The calculation formula is: Where: b is a gas sampling point at a certain position on the fully mechanized caving surface, j is any gas sampling point on the fully mechanized caving surface except b; x b is the observed value of gas content at a known gas sampling point; x j is the observed value of gas content at any gas sampling point on the fully mechanized caving face except b; r bc is the semivariogram function; S13, combining the formula in step S11 and the formula in step S12 to obtain the weight w b The calculation formula is: Where: r is the degree of correlation between the gas concentration value of the gas sampling point represented by each subscript and the gas concentration value of different b points; w i is the weight of the actual value of gas concentration at gas sampling point i, r i j is the semivariance between two gas sampling points i and j, r i0 is the weighted sum of squared differences.

4. The upper corner gas outburst early warning method based on contrast-regional uniformity map according to claim 3 is characterized by: In step S2, the temporal and spatial information calculation formula of the unknown gas sampling point is: Where: (x0, t0) is the time-space point to be predicted or estimated, x0 is the spatial position of the gas sampling point, t0 is the time coordinate of the gas sampling point; X(x0, t0) is the estimated value of the time-space point (x0, t0); w n is the weight coefficient; X(x b ,t b ) is the value of the known gas sampling point; Solving the above formula using the Lagrange factor μ yields: Where: γ is the spatiotemporal variation function.

5. The upper corner gas outburst early warning method based on contrast-regional uniformity diagram according to claim 4 is characterized in that: The calculation process of the spatiotemporal variogram includes the following steps: S21. Calculate the spatiotemporal empirical variation function and the spatiotemporal covariance function, and the calculation formula is: C(h s ,h t )=Cov(X(x+h s ,t+h t ),X(x,t)); Where: γ(h s ,h t ) is the spatial lag h s and time lag h t The semivariogram function under s ,h t ) represents the number used to calculate the semivariance when the spatial lag and time lag are used; X(x b ,t b ) is the spatial position x b and time t b The observed value at X(x b +h s ,t b +h t ) is the spatial position x b Plus the spatial lag h s and time t b Add the time lag h t The position observation value of S22. Calculate the spatiotemporal variation function, and its calculation formula is: Where: 2 is the variance of X(x,t).

6. The upper corner gas outburst early warning method based on contrast-regional uniformity map according to claim 5, characterized in that: In step S3, the fitting formula for fitting all gas sampling points using the Mattelon model is: Where: C0 is the nugget value; C0+C is the base value; C is the block arch height; e is the range; When h=0, the variation of the gas sampling points is only determined by the nugget effect; when 0<h≤e, the correlation of gas concentrations between gas sampling points begins to weaken with the increase of spatial distance; when h>e, the correlation of gas concentrations between gas sampling points reaches a stable value.

7. The upper corner gas outburst early warning method based on contrast-regional uniformity map according to claim 6, characterized in that: In step S4, when evaluating the gas concentration in the corners of the fully-mechanized caving surface, the influencing factors of the gas outburst in the corners are first found; then, the early warning evaluation index is calculated according to the weight of the influencing factors in each gas sampling point and the early warning evaluation value of the influencing factors, and the risk level of the gas outburst early warning in the corners of the fully-mechanized caving surface is obtained by grading the early warning evaluation index.

8. The upper corner gas outburst early warning method based on contrast-regional uniformity map according to claim 7 is characterized in that: The early warning evaluation indicators include corner gas indicators and coal quality indicators.

Citation Information

Cited By

  • Constructional coal seam gas content non-uniform distribution processing method based on space statistics

    CN121010460A