Coal and gas outburst prediction method and system based on grey correlation analysis

Through a method based on gray correlation analysis, combined with the physical properties of the mining area and artificial disturbance factors, coal and gas outburst prediction information is generated, which solves the problem of prediction inaccuracy caused by artificial disturbances in the prior art, and improves prediction accuracy and safety.

CN119066393AActive Publication Date: 2024-12-03XIAN XIKE MEASUREMENT & CONTROL EQUIP CO LTD

Patent Information

Application Number
CN202411195926.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-29
Publication Date
2024-12-03
Estimated Expiration
2044-08-29

AI Technical Summary

Technical Problem

The existing coal and gas outburst prediction methods ignore artificial disturbance factors, resulting in inaccurate prediction results and ineffective identification and prevention of gas outburst risks.

Method used

Using a method based on gray correlation analysis, combining physical attributes and artificial perturbation factors in the mining area, the coal and gas outburst prediction information is generated through data stratification processing, feature engineering and machine learning models to improve the prediction accuracy.

Benefits of technology

By comprehensively analyzing physical properties and artificial disturbance data, the prediction accuracy of coal and gas outbursts is improved, helping the mining area to identify and prevent the risks of gas outbursts in advance, and ensuring the safety of underground operations.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119066393A_ABST
    Figure CN119066393A_ABST
Patent Text Reader

Abstract

The invention discloses a coal and gas outburst prediction method and system based on grey correlation analysis, and relates to the field of coal and gas outburst prediction, and the method comprises the steps: obtaining to-be-processed data of a target region, carrying out the layering processing of the to-be-processed data into a first data layer used for representing the physical attributes of the target region, the second data layer is used for representing human activities in the target area; performing grey correlation analysis based on the first data layer and the second data layer, and determining at least one kind of data in each data layer as key data; performing feature engineering based on the key data of each data layer to generate target features; and based on the target features, predicting the target area by using a machine learning model, and generating coal and gas outburst prediction information. According to the method, complex human activity factors are processed by means of grey correlation analysis, high-correlation influence factors are found in combination with physical attribute data of a mining area, feature engineering is carried out on the high-correlation influence factors, and the prediction accuracy of a prediction model is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of coal and gas outburst prediction, and in particular to a coal and gas outburst prediction method and system based on grey correlation analysis. Background Art

[0002] With the great changes in mining methods and mining intensity, there are more and more outbursts in low-gas mines. The direct cause of outbursts does not depend on the gas in the coal seam, but on the concentration of ground stress, whether it can destroy the roof rock that plays the main bearing role and generate a strong impact force. Under the current conditions of large-scale mining or rapid excavation by tunneling machines, it is inevitable that a strong ground stress concentration phenomenon will be formed. The stress concentration formed by this mining plays an increasingly important role in outburst accidents. The reason why artificial disturbance induces outbursts in the working face is mainly because before the accident, the stress load on the coal rock that plays the main supporting role in the working face has reached a critical state, that is, the stress equilibrium state before collapse. Since the stress load has greatly exceeded the original stress load and has not reached the compressive strength of the coal rock, the main bearing rock of the roof still remains stable, the soft rock has long been broken, and a small slip has occurred, resulting in pressure conduction. The permeability of the compressed coal seam is further reduced than that of the original coal seam, resulting in a state where the gas outburst volume of the coal seam is reduced, which makes the operators have the illusion that the working face is still very safe at this time. It is the stress superposition and movement caused by artificial disturbance that breaks this balance, causing the concentrated stress to exceed the strength of the coal rock, which plays a major supporting role. The hard rock mass is destroyed instantly, and the powerful energy generated by the concentrated stress is suddenly released. The impact force on the coal rock in a large area crushes the coal mass, causing the adsorbed gas to suddenly desorb, the free gas to gather and expand, generate driving energy, and cause the outburst. However, the existing technology ignores the factor of artificial disturbance when performing data analysis.

[0003] For example, the Chinese patent with the authorization announcement number CN110397472B discloses a method, device and system for predicting coal and gas outbursts, which includes: obtaining at least one basic parameter of the briquette and the data value corresponding to the basic parameter; inputting the data value corresponding to the basic parameter into the prediction outburst model to determine multiple prediction parameters and the outburst results corresponding to the prediction parameters; obtaining the outburst warning index according to the prediction parameters and the outburst results according to the grey correlation analysis; obtaining the outburst warning level according to the outburst warning index and the warning level corresponding to the outburst warning index. This improves the rationality, accuracy and reliability of coal and gas outburst prediction, reduces the incidence of casualties, and ensures the safety of life and property underground.

[0004] The above methods have the problems raised by this background technology. It is one-sided to predict coal and gas outbursts based solely on the physical properties of the mining area. It is also necessary to conduct joint monitoring and analysis in combination with actual artificial disturbance factors. Summary of the invention

[0005] In view of the shortcomings of the prior art, this application discloses a coal and gas outburst prediction method and system based on grey correlation analysis, which is used to determine highly correlated data from complex artificial disturbance factors by means of grey correlation analysis, and to conduct a comprehensive analysis based on the physical properties of the mining area itself, and to perform feature engineering based on the analysis results as input features of the coal and gas outburst prediction model. The specific scheme is as follows:

[0006] In the first aspect, the present application provides a coal and gas outburst prediction method based on grey correlation analysis, including:

[0007] Acquire data to be processed of a target area, and hierarchically process the data to be processed into a first data layer for characterizing physical properties of the target area and a second data layer for characterizing human activities in the target area;

[0008] Performing grey relational analysis based on the first data layer and the second data layer, and determining at least one data in each of the data layers as key data;

[0009] Based on the key data of each of the data layers, feature engineering is performed to generate target features; based on the target features, a machine learning model is used to predict the target area to generate coal and gas outburst prediction information.

[0010] As an optional implementation, physical property data of the target area is obtained from a real-time monitoring system of the target area, including: gas concentration, gas emission rate, ground stress, ore layer thickness, ore layer slope, geological structure and mine depth;

[0011] Acquiring human activity data of the target area from the operation record system of the target area, including: mining method, working face footage speed, production shift, ventilation system type, and mining equipment power;

[0012] The human activity data is classified and coded.

[0013] As an optional implementation, the classifying and encoding the human activity data includes: classifying the human activity data into one of the following: categorical data, ordered data, and continuous data;

[0014] One-hot encoding is used for the categorical data in the human activity data to convert the categorical data into a plurality of binary features, wherein each feature represents a category;

[0015] Label encoding is used for the ordered data in the human activity data.

[0016] As an optional implementation, for the categorical data after one-hot encoding, the original binary feature value is converted to the range of 0 to 1 by using Min-Max normalization processing, and is converted to the range of -1 to 1 by using the conversion formula y=2x-1;

[0017] Where x represents the data point value after Min-Max standardization, and the y value obtained after conversion ranges from -1 to 1;

[0018] For the ordered data and continuous data after label encoding, Z-Score standardization and Min-Max standardization are used in turn to convert the data into the range of -1 to 1.

[0019] As an optional implementation, based on the information to be processed in the target area, the frequency of occurrence of gas outbursts is counted to form a gas outburst frequency sequence X0={x0(1), x0(2), ..., x0(n)}, where x0(n) represents the number of gas outbursts at the nth data point;

[0020] Using the gas outburst frequency sequence X0 as the parent sequence, performing grey correlation analysis on each data sequence in the first data layer and the second data layer, respectively, and calculating the correlation coefficient between each data sequence and the parent sequence;

[0021] Based on the calculated correlation coefficient, data sequences with correlation greater than a target correlation threshold are determined, and data points in these data sequences are used as key numbers.

[0022] As an optional implementation manner, the performing grey relational analysis based on the first data layer and the second data layer and determining at least one data in each data layer as key data further comprises:

[0023] Based on the correlation analysis between the historical data of gas outburst and the human activity data in the data to be processed in the target area, determining a lag period T for each type of human activity data;

[0024] Based on the lag period T of each type of human activity data, adjusting each type of human activity data to generate a lag data sequence;

[0025] For each human activity data series Y j ={y j (1),y j (2),...,y j (m)}, generating the lagged data series Y j,T ={y j (T+1),y j (T+2),...,y j (T+m)};

[0026] Among them, y j (m) is the sequence Y j The value at the mth data point, y j (T+m) is the lagged data sequence Y j,T The value at the T+mth data point; j represents the human activity data category index, j = {1, 2, 3, ...}.

[0027] As an optional implementation, a time series analysis is performed on the historical data of gas outburst and the data of human activities to determine the correlation function between the gas outburst event and the data of human activities;

[0028] Based on the correlation function, determining the location of a significant correlation peak;

[0029] Based on the time lag corresponding to the position of the significant correlation peak, the lag period T is determined.

[0030] As an optional implementation, the performing feature engineering based on the key data of each of the data layers to generate target features includes:

[0031] Based on the key data of each of the data layers, extract statistical features including: maximum value, minimum value, average value, standard deviation, and rate of change;

[0032] Extract temporal features, including moving average within the lag period, cumulative value within the time window, and peak frequency.

[0033] As an optional implementation, performing feature engineering based on the key data of each of the data layers to generate target features further includes:

[0034] Based on the statistical features and the time features, a combined feature is generated by utilizing linear combination and nonlinear combination.

[0035] In the second aspect, the present application also provides a coal and gas outburst prediction system based on grey correlation analysis, including:

[0036] A collection unit, used for acquiring the data to be processed of the target area, and hierarchically processing the data to be processed into a first data layer for characterizing the physical properties of the target area, and a second data layer for characterizing the human activities in the target area;

[0037] an analysis unit, configured to perform grey relational analysis based on the first data layer and the second data layer, and determine at least one data in each of the data layers as key data;

[0038] The prediction unit is used to perform feature engineering based on the key data of each of the data layers to generate target features; based on the target features, the target area is predicted using a machine learning model to generate coal and gas outburst prediction information.

[0039] In a third aspect, an optional implementation of the present application further provides a computer device, a processor, and a memory, wherein the memory stores machine-readable instructions executable by the processor, and the processor is used to execute the machine-readable instructions stored in the memory, and when the machine-readable instructions are executed by the processor, the machine-readable instructions perform the steps in the above-mentioned first aspect, or any possible implementation of the first aspect.

[0040] In a fourth aspect, an optional implementation of the present application further provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed, the steps of the above-mentioned first aspect, or any possible implementation of the first aspect are executed.

[0041] Compared with the prior art, the invention has the following beneficial effects: the invention not only considers the physical attribute data of the mining area, but also comprehensively considers the influence of human activity data. Through the comprehensive analysis of these two types of data, the risk of coal and gas outburst can be more accurately reflected. This comprehensive analysis method improves the accuracy of the prediction results, helps the mining area to identify and prevent the risk of gas outburst in advance, and ensures the safety of underground operations.

[0042] In view of the various and complex types of human activities, the present invention adaptively classifies and codes them to conduct a unified grey correlation analysis. In addition, considering that some human activities have a certain delay in affecting the mining area, the present invention also optimizes the data sequence based on this.

[0043] In addition, based on the results of grey correlation analysis, the present invention performs feature engineering based on high correlation data to generate target features for prediction models to predict coal and gas outbursts. Such target features that take human activities into consideration can improve the prediction accuracy of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 A flow chart of a method for predicting coal and gas outburst based on grey correlation analysis provided in an embodiment of the present disclosure;

[0045] Figure 2 A schematic diagram of the structure of a coal and gas outburst prediction system based on grey correlation analysis provided in an embodiment of the present disclosure;

[0046] Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0047] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0048] See also Figure 1 , Figure 1 The flowchart of the method for predicting coal and gas outburst based on grey correlation analysis is shown in FIG. 1 , wherein:

[0049] S101, obtaining data to be processed in a target area, and layering the data to be processed into a first data layer for characterizing physical properties of the target area and a second data layer for characterizing human activities in the target area;

[0050] S102, performing grey relational analysis based on the first data layer and the second data layer, and determining at least one data in each of the data layers as key data;

[0051] S103. Based on the key data of each of the data layers, feature engineering is performed to generate target features; based on the target features, a machine learning model is used to predict the target area to generate coal and gas outburst prediction information.

[0052] The prediction of coal and gas outburst is crucial in mine safety production. Traditional methods often rely on a single data source and it is difficult to fully reflect the various factors that affect gas outburst. By introducing the grey correlation analysis method, the relationship between different factors and gas outburst can be more comprehensively evaluated, thereby improving the accuracy of the prediction. Grey correlation analysis is a systematic analysis method for processing incomplete information, which is suitable for situations with small amounts of data and incomplete information. By analyzing multiple data layers, key data that is highly correlated with gas outburst events can be determined, and then feature engineering and machine learning models can be constructed to improve the prediction effect.

[0053] In a specific implementation, the disclosed embodiment obtains the data to be processed of the target area, which includes physical attribute data and human activity data, and layers the data to be processed into a first data layer for characterizing the physical attributes of the target area and a second data layer for characterizing the human activities of the target area.

[0054] Grey correlation analysis is performed on the first data layer and the second data layer respectively. At least one of the data in each data layer is determined to be key data through grey correlation analysis. Feature engineering is performed based on the key data of each data layer to generate target features. Based on the target features, the target area is predicted using a machine learning model to generate coal and gas outburst prediction information.

[0055] Regarding S101 above:

[0056] As an optional implementation, the physical property data of the target area are obtained from the real-time monitoring system of the target area, including but not limited to: gas concentration, gas emission rate, ground stress, ore layer thickness, ore layer slope, geological structure and mine depth; the human activity data of the target area are obtained from the operation recording system of the target area, including but not limited to: mining method, working face advancement speed, production shift, ventilation system type, and mining equipment power; the human activity data are classified and encoded.

[0057] In the disclosed embodiment, in order to effectively predict coal and gas outburst events, comprehensive and accurate data needs to be obtained. Physical attribute data and human activity data of the target area are two key factors affecting gas outburst. By obtaining these data from the real-time monitoring system and the operation record system, and classifying and coding the data, a data basis can be provided for subsequent grey correlation analysis and feature engineering.

[0058] In specific implementations, physical property data can be obtained from the real-time monitoring system of the mining area. For example, gas concentration data can be recorded once a minute, ore layer thickness and slope data can be measured once a day, and ground stress and geological structure data can be obtained through regular geological surveys.

[0059] It should be noted that in the subsequent grey correlation analysis, each data point in each sequence for correlation analysis needs to be consistent in time, that is, for each data with different collection frequencies, the data with the highest frequency needs to be interpolated as a benchmark.

[0060] For example, if the data series with the highest collection frequency is gas concentration and gas emission rate, and the frequency is recorded once per minute (real-time recording may also be selected based on actual conditions), then other data series should also be interpolated based on this. For example, the ground stress data recorded once an hour is converted into data per minute through linear interpolation. For example, the ore layer thickness recorded once an hour is converted into data per minute through linear interpolation.

[0061] In addition, the collection of various types of data is not limited to a preset fixed period. For example, the collection of physical attribute data can be associated with human activity data. In response to the conduct of a certain target human activity, the collection of certain physical attribute data can also start. For example, in response to the start of the tunneling activity of the tunnel boring machine, the collection of ground stress data also starts. When the tunneling activity ends, the collection of ground stress data can end, or it can be delayed for a certain period of time. This is because the changes in some physical attributes have a certain lag, and the delayed end of the collection can capture this lag. Similarly, the collected data is kept consistent with the collection frequency of the highest frequency data in the subsequent grey correlation analysis through interpolation.

[0062] Human activity data is obtained from the operation recording system. For example, the mining method is recorded as the specific mining method of each working face (such as longwall mining or shortwall mining), the working face footage rate is recorded as the daily footage, the production shift is recorded as the daily shift schedule, the ventilation system type is recorded as the ventilation method used for each working face, and the mining equipment power is recorded as the power usage of each equipment.

[0063] As an optional implementation, classifying and encoding human activity data includes: classifying the human activity data into one of the following: categorical data, ordered data, and continuous data; using one-hot encoding for the categorical data in the human activity data, and converting the categorical data into multiple binary features, where each feature represents a category; and using label encoding for the ordered data in the human activity data.

[0064] The research results show that no matter how large the ground stress is, if it is in a static and constant state, it will not trigger an outburst; the "essence of an outburst accident" is that the ground stress has a strong impact on the coal body, causing the coal body structure to be crushed and causing a sudden change in the coal body permeability (suddenly changing from low to high). At this time, the coal seam gas in the "adsorbed state" suddenly desorbs rapidly and turns into "free gas", desorbs rapidly in a very short time, and the "free gas" quickly gathers and expands in the coal body, generating expansion pressure; only then is the "expansion energy" that drives the outburst formed; when the expansion pressure is greater than the binding force of a coal wall at a certain point, the high-pressure gas will erupt with the crushed coal body, forming an outburst. This "dynamic dangerous state" is the object that should be captured in the outburst warning. The outburst properties of the coal seam are in a changing process, especially the size and position of the ground stress are always changing due to the influence of the underground operation process and scale. The change of ground stress will change the gas properties of the coal seam - the coal seam gas will desorb or adsorb under the action of pressure.

[0065] Therefore, in addition to the physical property data of the mining area, the embodiment of the present disclosure also collects various types of human activity data, and reflects the impact of underground operation activities on coal and gas outbursts through comprehensive analysis.

[0066] In specific implementations, these human activity data are obtained through the operation recording system, such as the specific mining method of each working face, the daily footage speed of the working face, the daily shift schedule, the type of ventilation system of each working face, and the daily power usage of the mining equipment.

[0067] It is understandable that the human activities in coal and gas mining projects are diverse. For example, the human activity data may also include the usage of the tunnel boring machine, the speed of the tunnel boring machine, etc. These underground operation activity related information can be used as data analyzed in the embodiments of the present disclosure.

[0068] In specific implementation, human activity data are different from physical attribute data. Human activity data may be discrete data or even non-numerical data. For a wide variety of human activity data, it is necessary to use data classification and coding methods to process them in preparation for subsequent grey correlation analysis.

[0069] The disclosed embodiment divides human activity data into three types: categorical data, ordered data, and continuous data. Among them, categorical data is represented as classification or category, and cannot be sorted or calculated, such as mining method and ventilation system type; ordered data represents categories with a clear order, which can be sorted, but cannot be directly calculated numerically, such as production shifts; continuous data is represented as numerical data, which can be calculated numerically and sorted, which is similar to the above-mentioned physical attribute data, such as working face footage speed and mining equipment power.

[0070] In the specific implementation, for categorical data, one-hot encoding is used to convert the categorical data into multiple binary features, where each feature represents a category. For example, mining methods include longwall mining, shortwall mining, and pillar mining, which are encoded as [1,0,0], [0,1,0], and [0,0,1] respectively.

[0071] For ordinal data, label encoding is used to convert the ordinal data into numerical labels. For example, the production shifts include day shift, night shift and holiday shift, which are coded as 1, 2 and 3 respectively.

[0072] It should be noted that the reason why production shifts are used as an example of ordered data is that different production shifts may be assigned different mining tasks, which can be used in the subsequent grey correlation analysis to determine which shifts' mining behaviors have a greater impact on the target observation data. In addition, ordered data also include: safety level, working face construction progress, equipment maintenance status, and mine ventilation conditions.

[0073] In this way, by obtaining detailed human activity data from the operation record system and classifying and coding these data, the comprehensiveness and accuracy of the data can be ensured. The human activity data after classification and coding can better meet the needs of grey correlation analysis and improve the accuracy and reliability of the analysis results. This method provides a data basis for subsequent feature engineering and machine learning models, which helps to improve the accuracy and reliability of gas outburst prediction.

[0074] As an optional implementation, for categorical data after one-hot encoding, Min-Max standardization is used to convert the original binary feature value into the range of 0 to 1, and converted into the range of -1 to 1 through the conversion formula y=2x-1, where x represents the data point value after Min-Max standardization, and the y value range after conversion is -1 to 1; for ordered data and continuous data after label encoding, Z-Score standardization and Min-Max standardization are used in sequence to convert the data into the range of -1 to 1.

[0075] In specific implementation, in order to process different types of data in grey relational analysis and machine learning models, the data must be standardized. This process helps to eliminate the dimensional differences between different data features so that they can be compared and analyzed within the same numerical range. Specifically:

[0076] Categorical data: After one-hot encoding, it needs to be further standardized to meet the needs of subsequent analysis. Min-Max normalization and linear transformation can adjust it to a uniform range.

[0077] Ordered and continuous data: After standardization, these data help eliminate the scale differences of the data and improve the convergence speed and prediction accuracy of the model. Z-Score standardization makes the data have a standard normal distribution with a mean of 0 and a standard deviation of 1, and then Min-Max standardization is performed to adjust it to a specific range.

[0078] In the specific implementation, the binary feature values ​​(0 and 1) after one-hot encoding are subjected to Min-Max normalization. Since the original value is 0 or 1, the Min-Max normalization result is kept in the range of 0 to 1. The conversion formula y=2x-1 is applied to convert the standardized data from the range of 0 to 1 to the range of -1 to 1. Among them, x represents the data point value after Min-Max normalization, and the y value range obtained after conversion is -1 to 1. For ordered data and continuous data after label encoding, the data is first Z-Score standardized to have a standard normal distribution with a mean of 0 and a standard deviation of 1. Then the standardized data is subjected to Min-Max normalization to adjust its range to -1 to 1.

[0079] For example, the mining method data obtained is one-hot encoded to form binary features such as [1,0,0] and [0,1,0]. These binary features are Min-Max standardized, and the results are kept in the range of 0 to 1. The formula y=2x-1 is applied to convert it to the range of -1 to 1. For example, the encoded data [1,0,0] is converted to [1,-1,-1].

[0080] The production shift data is encoded with labels to form values ​​such as [1, 2, 3]. These values ​​are Z-Score standardized. For example, data with a mean of 2 and a standard deviation of 1 are standardized to [-1, 0, 1]. Then Min-Max standardization is performed to adjust the range to -1 to 1.

[0081] For example, the face footage speed data is recorded as [2.5, 3.0, 3.5] m / day. These data are Z-Score standardized, for example, the mean is 3, the standard deviation is 0.5, and the standardized data is [-1, 0, 1]. Min-Max standardization is performed to adjust the range to -1 to 1.

[0082] In this way, through the above standardization processing steps, all data features are adjusted to a uniform numerical range, which contributes to the effectiveness of grey relational analysis and the accuracy of the machine learning model.

[0083] Regarding S102 above:

[0084] As an optional implementation, based on the information to be processed in the target area, the frequency of occurrence of gas outbursts is counted to form a gas outburst frequency sequence X0 = {x0(1), x0(2), ..., x0(n)}, wherein x0(n) represents the number of gas outbursts of the nth data point; the gas outburst frequency sequence X0 is used as a parent sequence, and grey correlation analysis is performed on each data sequence in the first data layer and the second data layer respectively, and the correlation coefficient between each data sequence and the parent sequence is calculated; based on the calculated correlation coefficient, several data sequences whose correlation is greater than the target correlation threshold are determined, and the data points in these data sequences are used as key numbers.

[0085] In the specific implementation, in order to predict coal and gas outburst events, the historical frequency of gas outburst can be counted and used to form a gas outburst frequency sequence. Through grey correlation analysis, the degree of correlation between different data sequences and gas outburst events can be determined. Specifically, the gas outburst frequency sequence is used as the parent sequence, which is compared with the data sequences in each data layer to calculate the correlation coefficient. Based on the correlation coefficient, key data that are highly correlated with gas outbursts can be identified, which will be used in subsequent feature engineering and prediction models.

[0086] In the specific implementation, the occurrence data of gas outburst events are obtained from the historical records, and the historical data are divided into multiple time periods (such as monthly, weekly, etc.), and the number of gas outbursts in each time period is counted. The number of gas outbursts in each time period is recorded as a sequence.

[0087] Use the gas outburst frequency sequence X0 as the mother sequence. At the same time, prepare the data sequences in the first data layer and the second data layer. These data sequences are X i ={x i (1),x i (2),...,x i (n)}, where X i represents the i-th data sequence, x i (n) represents the data sequence X i The value at the nth data point. Calculate each data sequence X i The correlation coefficient between X0 and the parent sequence X0. The calculation of the correlation coefficient is based on the following formula:

[0088]

[0089] Among them, Δ(x0(k),x i (k)) represents the parent sequence x0(k) and the comparison data sequence x i (k) the difference at data point k, k = 1, 2, 3, ..., n; For each comparison sequence x i Perform the maximum value operation on all data points k of , find the maximum difference between the sequence and the parent sequence x0 at all data points, and then perform the maximum value operation on all comparison sequences x i Perform the maximum value operation on these maximum difference values ​​and find the largest difference value in all sequences. For each comparison sequence x i Perform the minimum operation on all data points k of , find the minimum difference between the sequence and the parent sequence x0 at all data points, and then perform the minimum operation on all comparison sequences x i The minimum difference values ​​of the sequences are calculated to find the smallest difference value in all sequences. ρ is the resolution coefficient, which is usually a small positive number, such as 0.5.

[0090] Set a target correlation threshold γ threshold , based on the calculated correlation coefficient, filter out the correlation coefficients greater than γ threshold The data points in the selected data sequences are used as key data for subsequent feature engineering and prediction models.

[0091] For example, the data is divided into months, and the number of gas outbursts in each month is counted to form a gas outburst frequency sequence. For example, the gas outburst frequency sequence of a mining area is:

[0092] X0={3,5,2,4,6,7,2,3,5,8,4,6}

[0093] Among them, the sequence has 12 data points, each data point corresponds to the number of gas outbursts in each month.

[0094] Prepare data sequences in the first data layer (such as gas concentration, ore layer thickness) and the second data layer (such as mining method, footage speed). Calculate the correlation coefficient between each data sequence and the gas outburst frequency sequence. For example, the correlation coefficient between the gas concentration sequence X1 and the gas outburst frequency sequence X0 is γ1, and the correlation coefficient between the ore layer thickness sequence X2 is γ2, etc.

[0095] Set the target correlation threshold γ threshold =0.8, filter out data sequences with correlation coefficient greater than 0.8.

[0096] In this way, by counting the frequency of gas outbursts and forming a frequency series, a basis can be provided for subsequent grey correlation analysis. Grey correlation analysis can accurately evaluate the degree of correlation between different data sequences and gas outburst events, and identify key data that are highly correlated with gas outbursts. After feature engineering processing, these key data can significantly improve the accuracy and reliability of the prediction model, helping mining areas to identify and prevent the risk of gas outbursts in advance.

[0097] As an optional implementation, the grey correlation analysis is performed based on the first data layer and the second data layer to determine that at least one of the several types of data in each of the data layers is key data, further comprising: determining a lag period T for each type of human activity data based on a correlation analysis between the historical data of gas outbursts and the human activity data in the data to be processed in the target area; adjusting each type of human activity data based on the lag period T of each type of human activity data to generate a lagged data sequence;

[0098] For each human activity data series Y j ={y j (1),y j (2),...,y j (m)}, generating the lagged data series Y j,T ={y j (T+1),y j (T+2),...,y j (T+m)};

[0099] Among them, y j(m) is the sequence Y j The value at the mth data point, y j (T+m) is the lagged data sequence Y j,T The value at the T+mth data point; j represents the human activity data category index, j = {1, 2, 3, ...}.

[0100] The lagged data series Y j,T With the unlagged data series Y j Merge to form the new second data layer.

[0101] In specific implementation, the occurrence of gas outburst events is not only related to the current mine environment and operating conditions, but also affected by historical operating data. Different operating data may have an impact on gas outbursts within different lag periods. By introducing the time lag factor, the impact of human activity data on gas outbursts can be more comprehensively analyzed, thereby improving the accuracy and reliability of grey correlation analysis.

[0102] In order to determine the lag period T, it is necessary to obtain the historical data of gas outbursts and human activity data in the target area. By calculating the cross-correlation function (CCF) between gas outburst events and various types of human activity data, the correlation under different lag periods can be analyzed. The correlation peak is identified to determine the optimal lag period T for each type of human activity data. Based on the determined lag period T, each type of human activity data is adjusted to generate a lagged data series.

[0103] In addition, some types of human activity data have a smaller lag period, for example, within a time range of several seconds or even smaller. For such human activity data, the specific value of the lag period T can be set to 0, that is, there is no lag period.

[0104] Exemplarily, historical data on gas outbursts and data on human activities are obtained, such as mining methods, face footage speed, production shifts, ventilation system types, and mining equipment power. By calculating the cross-correlation function between gas outburst events and mining methods, it is found that the lag period T is 2 hours; by calculating the cross-correlation function between gas outburst events and face footage speed, it is found that the lag period T is 1 hour. Based on these lag periods, the mining method data Y1 = {y1(1), y1(2), ..., y1(m)} is lagged and adjusted to generate a lagged data sequence Y 1,2 ={y1(3) y1(4) y1(m+)}; perform hysteresis adjustment on the working face advance speed data Y2 ={y2(1), y2(2), ..., y2(m)} to generate the hysteresis data sequence Y 2,1 ={y2(2) y2(3)y2(m+1)}.

[0105] It should be noted that the lag adjustment of the data series may result in missing data at some data points. For example, under the premise that the data points are counted in hours, the data series A has no lag period, the data series B has a lag period of 2 hours, and the data series C has a lag period of 1 hour. After the B and C series are lagged, they cannot be directly combined with the A series for grey correlation analysis, because the data points of the A series start from the first hour, and the B and C series after the lag processing start from the third hour and the second hour respectively, which will cause time misalignment. Similarly, if the data points of the A series end from the 24th hour, the B and C series after the lag processing end from the 26th hour and the 25th hour respectively. At this time, the data points common to the three series can be selected for analysis, that is, from the third hour to the 24th hour, so as to retain the correlation of each data series under the premise of introducing time lag.

[0106] In this way, by introducing the time lag factor, the impact of human activity data on gas outburst can be analyzed more comprehensively. This method not only considers the impact of current data, but also the lag effect of historical data, thereby improving the accuracy and reliability of grey correlation analysis.

[0107] As an optional implementation, a time series analysis is performed on historical data of gas outbursts and human activity data to determine the correlation function between gas outburst events and human activity data; based on the correlation function, the position of a significant correlation peak is determined; based on the time lag corresponding to the position of the significant correlation peak, the lag period T is determined.

[0108] In the specific implementation, historical data of gas outbursts and related human activity data are collected to ensure that the time range and time unit of the two data are consistent, for example, both are in days. These data include the number of daily gas outburst events and records of human activities such as mining methods and footage speed. Use the correlation function tool in statistical software or programming language to calculate the cross-correlation function (CCF) between gas outburst events and various types of human activity data. During the analysis process, the data correlation under different lag periods (from negative to positive, such as from -30 days to +30 days) is considered. By calculating the CCF under different lag periods, the temporal correlation between gas outburst events and human activity data can be understood. Then, the peaks with the highest correlation are identified from the CCF results, which represent the most significant correlation between the data. The lag period T corresponding to these peaks is recorded as a basis for further analysis.

[0109] To ensure the reliability of the correlation peaks, perform a statistical significance test on the peaks in the CCF results to confirm whether these correlations are significantly higher than the random background level. Use p-values ​​or other statistical criteria to evaluate the significance of the correlation peaks. Finally, select the lag period T corresponding to the correlation peak with the highest significance as the optimal lag period.

[0110] For example, the relationship between the change in gas concentration and the mining progress is analyzed. The daily gas concentration and the corresponding daily mining footage rate data for the past year were collected, standardized, and the CCF between the two was calculated. The analysis found that the correlation between gas concentration and mining progress reached a significant peak at a lag of 3 days, and the p value of the significance test was less than 0.05. Therefore, 3 days was determined to be the optimal lag period for the impact of mining progress on gas concentration changes.

[0111] In this way, the temporal correlation between gas outburst and human activities can be understood more accurately, and the accuracy of grey correlation analysis can be improved.

[0112] Regarding S103 above:

[0113] As an optional implementation, the feature engineering is performed based on the key data of each of the data layers to generate target features, including: based on the key data of each of the data layers, statistical features are extracted, including: maximum value, minimum value, average value, standard deviation, and rate of change; time features are extracted, including: moving average value within the lag period, cumulative value within the time window, and peak frequency.

[0114] In the specific implementation, after the key data are determined through grey correlation analysis, feature engineering is carried out to make more in-depth use of these data to predict gas outburst events. The purpose of feature engineering is to extract more informative features from the raw data, which should be able to better represent the essential attributes of the data and have a positive impact on the performance of the prediction model.

[0115] Confirm the key data obtained from the grey correlation analysis, which are highly correlated with the parent sequence (gas outburst frequency sequence). These key data may include various physical and operational parameters of the mine, such as gas concentration, ventilation speed, mining speed, etc.

[0116] Statistical features are extracted for each key data, including maximum, minimum, mean, standard deviation, and rate of change. These statistics depict the basic distribution and fluctuation of the data, providing the model with a basic understanding of data volatility and trends. For example, the maximum and minimum values ​​capture extreme behaviors, the mean and standard deviation describe the central trend and fluctuation range, and the rate of change reflects the speed of change of time series data.

[0117] Secondly, the extraction of temporal features focuses on the time series dynamics of the data, including the moving average within the lag period, the cumulative value within the time window, and the peak frequency. The moving average helps smooth short-term fluctuations and highlight long-term trends, making the model resistant to temporary anomalies. The cumulative value and peak frequency within the time window provide insights into periodic and sudden events, which is very important for predicting gas outbursts with periodic patterns or sudden peaks.

[0118] For example, in a study on gas outburst prediction in a mining area, gas concentration and mining speed were identified as key data highly correlated with gas outburst frequency. Detailed statistics and time feature extraction were performed on these data, including calculation of monthly average gas concentration, monthly rate of change of gas concentration, and weekly peak frequency. These features were used to train a prediction model for predicting gas outburst events.

[0119] As an optional implementation, the feature engineering based on the key data of each of the data layers to generate target features also includes: generating combined features using linear combinations and nonlinear combinations based on the statistical features and the time features.

[0120] In the specific implementation, after extracting statistical features and time features of key data, it is further proposed to use these features to generate combined features through linear and nonlinear combination, in order to further improve the complexity and accuracy of model prediction. Combined features can reveal the relationship between data features and provide a more comprehensive data expression, thereby enhancing the predictive ability of machine learning models. New features are generated using linear combination methods, such as weighted summation. For example, the moving average and the rate of change can be weighted to construct a new feature that reflects the speed and stability of trend changes. In addition, a linear regression model can be used to generate combined features using statistical features and time features as input. These features can help the model better understand the linear relationship between data.

[0121] For nonlinear combination features, interaction terms between features can be created. For example, by using grey correlation analysis to determine that the gas concentration change rate in the first data layer and the mining speed in the second data layer are key data, an interaction term between the gas concentration change rate and the mining speed can be created.

[0122] For the target features processed by grey correlation analysis and feature engineering, a variety of prediction models can be considered for risk assessment of gas outburst or other types of events. For example, decision trees, random forests, neural networks, and deep learning methods can be used.

[0123] Taking neural networks as an example, a typical network includes an input layer, several hidden layers, and an output layer. The number of neurons in the input layer should match the number of features. The hidden layer can be experimented with different numbers of neurons and layers to find the best network structure. Common activation functions include ReLU or Sigmoid. The design of the output layer depends on the specific task (such as classification or regression).

[0124] In this way, the model using the target features generated by the embodiment of the present disclosure can make full use of the key data obtained through grey correlation analysis, thereby providing accurate support in predicting gas outburst risks. This method not only improves the accuracy of prediction, but also enhances the generalization ability of the model, enabling it to perform well on new and unseen data.

[0125] Based on the same inventive concept, the embodiments of the present disclosure also provide a coal and gas outburst prediction system based on grey correlation analysis corresponding to the coal and gas outburst prediction method based on grey correlation analysis. Since the principle of the system in the embodiments of the present disclosure to solve the problem is similar to the above-mentioned coal and gas outburst prediction method based on grey correlation analysis in the embodiments of the present disclosure, the implementation of the system can refer to the implementation of the method, and the repeated parts will not be repeated.

[0126] Reference Figure 2 FIG. 1 is a schematic diagram of the structure of a coal and gas outburst prediction system based on grey correlation analysis provided by an embodiment of the present disclosure. The system includes:

[0127] The acquisition unit 10 is used to obtain the data to be processed in the target area, and to hierarchically process the data to be processed into a first data layer for characterizing the physical properties of the target area and a second data layer for characterizing the human activities in the target area;

[0128] An analysis unit 20, configured to perform grey relational analysis based on the first data layer and the second data layer, and determine at least one data in each of the data layers as key data;

[0129] The prediction unit 30 is used to perform feature engineering based on the key data of each of the data layers to generate target features; based on the target features, the target area is predicted using a machine learning model to generate coal and gas outburst prediction information.

[0130] The present disclosure also provides an electronic device, such as Figure 3 FIG. 1 is a schematic diagram of the structure of an electronic device provided in an embodiment of the present disclosure, including:

[0131] A processor 31 and a memory 32; the memory 32 stores machine-readable instructions executable by the processor 31, and the processor 31 is used to execute the machine-readable instructions stored in the memory 32. When the machine-readable instructions are executed by the processor 31, the processor 31 performs the following steps:

[0132] Acquire data to be processed of a target area, and hierarchically process the data to be processed into a first data layer for characterizing physical properties of the target area and a second data layer for characterizing human activities in the target area;

[0133] Performing grey relational analysis based on the first data layer and the second data layer, and determining at least one data in each of the data layers as key data;

[0134] Based on the key data of each of the data layers, feature engineering is performed to generate target features; based on the target features, a machine learning model is used to predict the target area to generate coal and gas outburst prediction information.

[0135] The above-mentioned memory 32 includes internal memory 321 and external memory 322; the memory 321 here is also called internal memory, which is used to temporarily store the calculation data in the processor 31, as well as the data exchanged with the external memory 322 such as the hard disk. The processor 31 exchanges data with the external memory 322 through the internal memory 321.

[0136] The specific execution process of the above instructions can refer to the steps of the coal and gas outburst prediction method based on grey correlation analysis described in the embodiment of the present disclosure, which will not be repeated here.

[0137] The embodiment of the present disclosure also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the coal and gas outburst prediction method based on grey correlation analysis described in the above method embodiment are executed. The storage medium can be a volatile or non-volatile computer-readable storage medium.

[0138] The embodiments of the present disclosure also provide a computer program product, which carries a program code. The instructions included in the program code can be used to execute the steps of the coal and gas outburst prediction method based on grey correlation analysis described in the above method embodiment. Please refer to the above method embodiment for details, which will not be repeated here.

[0139] The computer program product may be implemented in hardware, software or a combination thereof. In one optional embodiment, the computer program product is implemented as a computer storage medium. In another optional embodiment, the computer program product is implemented as a software product, such as a software development kit (SDK).

[0140] It should be understood that determining B based on A does not mean determining B only based on A. B can also be determined based on A and / or other information.

[0141] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in the present invention can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0142] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0143] The preferred embodiments of the present invention disclosed above are only used to help explain the present application. The preferred embodiments do not describe all the details in detail, nor do they limit the present application to specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present application, so that those skilled in the art can understand and use the present application well. The present application is limited only by the claims and their full scope and equivalents.

Claims

1. A coal and gas outburst prediction method based on grey correlation analysis, characterized in that: include: Acquire data to be processed of a target area, and hierarchically process the data to be processed into a first data layer for characterizing physical properties of the target area and a second data layer for characterizing human activities in the target area; Performing grey relational analysis based on the first data layer and the second data layer, and determining at least one data in each of the data layers as key data; Based on the key data of each of the data layers, feature engineering is performed to generate target features; Based on the target features, a machine learning model is used to predict the target area and generate coal and gas outburst prediction information.

2. The method according to claim 1, characterized in that Acquiring physical property data of the target area from a real-time monitoring system of the target area, including: gas concentration, gas emission rate, ground stress, ore layer thickness, ore layer slope, geological structure and mine depth; Acquiring human activity data of the target area from the operation record system of the target area, including: mining method, working face footage speed, production shift, ventilation system type, and mining equipment power; The human activity data is classified and coded.

3. The method according to claim 2, characterized in that The classifying and coding the human activity data includes: classifying the human activity data into one of the following: categorical data, ordered data, and continuous data; One-hot encoding is used for the categorical data in the human activity data to convert the categorical data into a plurality of binary features, wherein each feature represents a category; Label encoding is used for the ordered data in the human activity data.

4. The method according to claim 3, characterized in that For the categorical data after one-hot encoding, the original binary feature value is converted to the range of 0 to 1 using Min-Max normalization, and then converted to the range of -1 to 1 using the conversion formula y=2x-1; Where x represents the data point value after Min-Max normalization, and the y value obtained after conversion ranges from -1 to 1; For the ordered data and continuous data after label encoding, Z-Score standardization and Min-Max standardization are used in turn to convert the data into the range of -1 to 1.

5. The method according to claim 4, characterized in that Based on the information to be processed in the target area, the frequency of occurrence of gas outburst is counted to form a gas outburst frequency sequence X0={x0(1), x0(2), ..., x0(n)}, where x0(n) represents the number of gas outbursts at the nth data point; Using the gas outburst frequency sequence X0 as the parent sequence, performing grey correlation analysis on each data sequence in the first data layer and the second data layer, respectively, and calculating the correlation coefficient between each data sequence and the parent sequence; Based on the calculated correlation coefficient, data sequences with correlation greater than a target correlation threshold are determined, and data points in these data sequences are used as key numbers.

6. The method according to claim 5, characterized in that The performing grey relational analysis based on the first data layer and the second data layer to determine at least one data in each data layer as key data further comprises: Based on the correlation analysis between the historical data of gas outburst and the human activity data in the data to be processed in the target area, determining a lag period T for each type of human activity data; Based on the lag period T of each type of human activity data, adjusting each type of human activity data to generate a lag data sequence; For each human activity data sequence Yj = {yj(1), yj(2), ..., yj(m)}, generate a lagged data sequence Yj, T ={yj(T+1),yj(T+2),...,yj(T+m)}; Among them, y j (m) is the sequence Y j The value of the mth data point, y j (T+m) is the lagged data sequence Y j,T The value at the T+mth data point; j represents the human activity data category index, j = {1, 2, 3, ...}.

7. The method according to claim 6, characterized in that Conduct time series analysis on historical data of gas outburst and human activity data to determine the correlation function between gas outburst events and human activity data; Based on the correlation function, determining the location of a significant correlation peak; Based on the time lag corresponding to the position of the significant correlation peak, the lag period T is determined.

8. The method according to claim 7, characterized in that The feature engineering based on the key data of each data layer to generate target features includes: Based on the key data of each of the data layers, extract statistical features including: maximum value, minimum value, average value, standard deviation, and rate of change; Extract temporal features, including moving average within the lag period, cumulative value within the time window, and peak frequency.

9. The method according to claim 8, characterized in that The performing feature engineering based on the key data of each of the data layers to generate target features also includes: Based on the statistical features and the time features, a combined feature is generated by utilizing linear combination and nonlinear combination.

10. The coal and gas outburst prediction system based on grey correlation analysis is characterized by: Executing the coal and gas outburst prediction method based on grey correlation analysis as described in any one of claims 1 to 9 comprises: A collection unit, used for acquiring the data to be processed of the target area, and hierarchically processing the data to be processed into a first data layer for characterizing the physical properties of the target area, and a second data layer for characterizing the human activities in the target area; an analysis unit, configured to perform grey relational analysis based on the first data layer and the second data layer, and determine at least one data in each of the data layers as key data; The prediction unit is used to perform feature engineering based on the key data of each of the data layers to generate target features; based on the target features, the target area is predicted using a machine learning model to generate coal and gas outburst prediction information.

Citation Information

Patent Citations

  • Mine gas gushing amount prediction method based on MR / S analysis

    CN108364102A

  • Coal rock gas hybrid dynamic disaster prediction method based on deep learning

    CN116467572A

  • Monitoring and forewarning method for coal-rock dynamic disasters based on electromagnetic radiation and earth sound

    US20200033496A1

Cited By

  • Method and system for automatically searching and checking coal and gas outburst early warning criterion

    CN119066404A

  • Method and system for automatically finding and verifying coal and gas outburst early warning criteria

    CN119066404B