Gas concentration prediction method for coal mining face based on multi-factor generalized linear regression

Abstract

The coal mining face gas concentration prediction method based on a multi-factor generalized linear regression belongs to the technical field of coal mine gas detection, solves the problems that a traditional safety monitoring system can only passively monitor gas concentration, cannot early warning analysis, and previous big data prediction can only analyze single sensor historical data, and data accuracy is poor; the present application is based on coal seam thickness, gas extraction amount, daily output, wind speed, T0 methane sensor gas concentration and the T2 methane sensor gas concentration to be predicted, establishes a multi-factor generalized linear regression model, can process a large amount of historical data, can effectively mine the corresponding linear relationship from each related factor affecting the T2 methane sensor concentration, is beneficial to accurately predicting the methane concentration, so as to achieve the effect of predicting the methane concentration of the coal mining face return airway, early understanding the gas concentration change trend, and taking control measures in advance.

Description

Technical Field

[0001] This invention belongs to the field of coal mine gas detection technology, and relates to a method for predicting gas concentration in coal mining faces based on multi-factor generalized linear regression. Background Technology

[0002] Gas disasters are a major safety issue in the coal mining industry. During normal production, coal mines strictly control the gas concentration at the working face to below 1%. However, when encountering structural problems, insufficient ventilation, or increased regional gas content, the gas concentration at the working face can easily exceed the limit. Anomalies in gas concentrations are often preceding gas disasters in coal mines, and accurately determining gas concentrations is fundamental for predicting gas outbursts and designing ventilation systems.

[0003] The multi-source characteristics of gas in coal mining faces and the migration characteristics of gas mixtures make gas concentration both regular and complex, presenting a typical nonlinear time series prediction problem. First, mine gas concentration is typical time series data, with the gas concentration at a single monitoring point showing a temporal correlation with the historical gas concentration at that point. Second, gas within roadways is also influenced by multiple factors such as coal seam thickness, gas extraction rate, roadway ventilation rate, and production output, exhibiting the complexity of multi-factor influence.

[0004] Currently, coal mines can only monitor and warn of gas through the deployment of safety monitoring systems and real-time online monitoring using methane sensors. This is a passive monitoring method that can only monitor the concentration after gas has erupted. It cannot provide early warning or analysis of the gas concentration at a specific location underground. In addition, some methods that estimate future gas concentration trends based on the statistical analysis of historical data from sensors are limited in scope and accuracy, and cannot provide a comprehensive analysis of various factors that can affect gas concentration changes during production.

[0005] The gas concentration in the return airway of a coal mining face is relatively stable and can basically reflect the gas situation of the entire roadway. Therefore, this study aims to find out how to proactively predict the gas concentration in the return airway of a coal mining face, and to proactively issue early warnings and take control measures based on the prediction results. Summary of the Invention

[0006] This invention addresses the problems of traditional safety monitoring systems, which can only passively monitor gas concentration and cannot provide early warning analysis, as well as the poor accuracy of previous big data predictions, which could only analyze historical data from a single sensor.

[0007] The present invention solves the above-mentioned technical problems through the following technical solutions:

[0008] A method for predicting gas concentration in coal mining faces based on multi-factor generalized linear regression includes the following steps:

[0009] S1. Export data on various factors affecting the gas concentration of the T2 methane sensor from the MySQL database to establish a basic dataset; based on the basic dataset, organize the data in the basic dataset according to time correlation to form a dataset of gas concentration influencing factors.

[0010] S2. Perform feature standardization on the data set of factors affecting gas concentration to obtain a standardized data set for predicting the gas concentration of the T2 methane sensor at the next data acquisition time. Then, divide the standardized data set into a training set and a test set using a random sampling method.

[0011] S3, based on coal seam thickness D, gas extraction rate C, daily output Q, wind speed S, and T0 methane sensor gas concentration. And the T2 methane sensor gas concentration that needs to be predicted Establish a multi-factor generalized linear regression model;

[0012] The calculation formula for the multi-factor generalized linear regression model is as follows:

[0013]

[0014] The above formula can be simplified using vectors as follows:

[0015]

[0016] in, ω D ω represents the offset weight of the effect of coal seam thickness on the gas concentration of the T2 methane sensor. C ω represents the offset weight of the effect of gas extraction rate on the gas concentration of the T2 methane sensor. Q ω represents the offset weight of the effect of daily output on the gas concentration of the T2 methane sensor. s denoted as the offset weight of the effect of wind speed on the gas concentration of the T2 methane sensor, and b is the offset correction constant;

[0017] S4. Process the training set obtained in step S2 using the Kemin normal distribution method, and then train the offset weights of each influencing factor in the multi-factor generalized linear regression model using the least squares method.

[0018] The formula for calculating the least squares method is as follows:

[0019]

[0020] Where y is the original T2 methane sensor gas concentration in the training set. The methane concentration of the T2 sensor is predicted by a multifactor generalized linear regression model.

[0021] Furthermore, the data on various factors affecting the gas concentration of the T2 methane sensor exported from the MySQL database in step S1 includes: historical data of the T0 methane sensor, coal face production data, working face gas extraction data, working face coal seam thickness data, and return airway wind speed data.

[0022] Furthermore, the method for performing feature standardization processing on the data set of factors affecting gas concentration in step S2 to obtain a standardized data set for predicting the gas concentration of the T2 methane sensor at the next data acquisition time is as follows:

[0023] (1) Format the data tuples of the data set of gas concentration influencing factors into a vector form: (T0_MAX,EXTRACT_COUT,COAL_PRODUCTION,COAL_THINKNESS,WIN_SPEED_AVG,NEXT_T2);

[0024] (2) Standardize each column of the data tuples in the data set of gas concentration influencing factors to obtain the standardized dataset.

[0025] Furthermore, the method for standardizing each column of the data tuples in the dataset of factors affecting gas concentration is as follows:

[0026] 1) Calculate the average value for each column of data in the data tuple using the following formula: Where μ is the average value, x i For each sample data point, N represents the value of a single sample data point, where N is the total number of samples in each column of the data tuple.

[0027] 2) Calculate the standard deviation using the following formula: Where σ is the standard deviation;

[0028] 3) Perform z-score standardization on individual samples of each column in the data tuple, using the following formula:

[0029] The advantages of this invention are:

[0030] This invention establishes a multi-factor generalized linear regression model based on coal seam thickness, gas extraction volume, daily output, wind speed, T0 methane sensor gas concentration, and the T2 methane sensor gas concentration to be predicted. This model can process large amounts of historical data and effectively extract the corresponding linear relationships from various factors affecting the T2 methane sensor concentration. This is beneficial for accurately predicting methane concentration, thereby achieving the effects of predicting methane concentration in the return airway of the coal mining face, understanding the trend of gas concentration changes in advance, and taking proactive control measures. Attached Figure Description

[0031] Figure 1 This is a flowchart of the gas concentration prediction method for coal mining faces based on multi-factor generalized linear regression of the present invention.

[0032] Figure 2 This is a schematic diagram of the loss result of model training for the gas concentration prediction method for coal mining faces based on multi-factor generalized linear regression of the present invention.

[0033] Figure 3 This is a comparison chart of the prediction results of the gas concentration prediction method for coal mining faces based on multi-factor generalized linear regression of the present invention. Detailed Implementation

[0034] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0035] The technical solution of the present invention will be further described below with reference to the accompanying drawings and specific embodiments:

[0036] Example 1

[0037] like Figure 1 As shown, the method for predicting gas concentration in coal mining faces based on multi-factor generalized linear regression includes the following steps:

[0038] 1. Data Analysis

[0039] Data on various factors affecting the gas concentration of the T2 methane sensor were exported from a MySQL database, including: historical data of the T0 methane sensor, coal face production data, coal face gas extraction data, coal seam thickness data, and return airway wind speed data, to establish a basic dataset. The T0 methane sensor is installed at the upper corner of the return airway of the coal face; the T2 methane sensor is installed in the return airway of the coal face, within 10 to 15 meters of the first return airway intersection point in the direction of return air.

[0040] Based on the basic dataset, the data in the basic dataset are organized into a dataset of gas concentration influencing factors based on time correlation. The data tuple format in the dataset of gas concentration influencing factors is shown in Table 1 below.

[0041] Table 1 Data Correspondence Names

[0042] Serial Number Column Name Data Name 1 T0_MAX Historical data of T0 methane sensor 2 EXTRACT_COUNT Cumulative extraction volume 3 COAL_PRODUCTION Coal production at the working face 4 COAL_THINKNESS Coal seam thickness 5 WIN_SPEED_AVG Return air velocity at the working face

[0043] 2. Feature standardization and dataset partitioning

[0044] The dataset of gas concentration influencing factors established through data analysis in step 1 still needs further processing to obtain a standardized dataset for predicting the gas concentration of the T2 methane sensor.

[0045] First, the methane concentration influencing factors dataset in step 1 provides the influencing factors affecting the methane concentration of the T2 methane sensor. Its main purpose is to predict the methane concentration at location T2 at the next data acquisition time. Therefore, the data tuples of the methane concentration influencing factors dataset are formatted as vector form: (T0_MAX,EXTRACT_COUT,COAL_PRODUCTION,COAL_THINKNESS,WIN_SPEED_AVG,NEXT_T2).

[0046] Secondly, to ensure data availability, each column of the data tuple is standardized:

[0047] (1) Calculate the average value for each column of data in the data tuple using the following formula: Where μ is the average value, x i For each sample data point, N represents the value of a single sample data point, where N is the total number of samples in each column of the data tuple.

[0048] (2) Calculate the standard deviation using the following formula: Where σ is the standard deviation;

[0049] (3) Perform z-score standardization on individual samples of each column of data in the data tuple, using the following formula:

[0050]

[0051] Finally, using the standardized dataset obtained above, a random sampling method is employed to divide the standardized dataset into a training set and a test set for subsequent model training, prediction evaluation, and prediction calibration. In this embodiment, 70% of the data in the standardized dataset is used as the training set for model training, and 30% of the data is used as the test set for model testing and evaluation.

[0052] 3. Construct a multi-factor generalized linear regression model

[0053] First, the first criterion for linear regression is that the dependent variable Y must be a "quantitative variable," i.e., a numerical variable. The T2 methane sensor gas concentration to be predicted in this invention meets the first criterion for linear regression. Second, the coal seam thickness determines the total amount of gas; daily output affects the approximate daily gas emission. The main influences on the gas concentration in the working face are the extraction rate and ventilation rate. In ventilation, the gas follows the airflow through the working face and return airway to the return airway in the mining area. Therefore, the coal seam thickness, gas extraction rate, daily output, wind speed, and the values ​​of the T0 and T2 methane sensors have a univariate regression or higher-order regression relationship. Therefore, this invention uses a generalized linear regression algorithm for data fitting, and the data also follows a Gaussian distribution.

[0054] This invention is based on coal seam thickness D, gas extraction rate C, daily output Q, wind speed S, and T0 methane sensor gas concentration. Predicting the gas concentration of the T2 methane sensor A multi-factor generalized linear regression model is established, and the calculation formula for the multi-factor generalized linear regression model is as follows:

[0055]

[0056] Where, ω S ω represents the offset weight of the effect of coal seam thickness on the gas concentration of the T2 methane sensor. C ω represents the offset weight of the effect of gas extraction rate on the gas concentration of the T2 methane sensor. Q ω represents the offset weight of the effect of daily output on the gas concentration of the T2 methane sensor. s denoted as the offset weight of the effect of wind speed on the gas concentration of the T2 methane sensor, and b is the offset correction constant.

[0057] If we represent the factors affecting the gas concentration of the T2 methane sensor and their corresponding weights using a vector representation, the above formula can be simplified to:

[0058]

[0059] in,

[0060] This invention employs a generalized linear regression model, which boasts high computation speed and the ability to process large amounts of historical data. It overcomes the limitations of traditional safety monitoring systems that passively monitor methane concentration. This invention predicts methane concentration in the return airway of coal mining faces, enabling early understanding of methane concentration trends and allowing for proactive measures. Multi-factor generalized linear regression more effectively uncovers linear correlations among various factors influencing the T2 methane sensor concentration, leading to more accurate methane concentration predictions. This invention addresses the limitation of previous big data prediction methods that only analyzed historical data from a single sensor. It incorporates factors such as coal seam thickness, methane extraction volume, wind speed, and production output for comprehensive prediction.

[0061] 4. Model Training

[0062] (1) The training set obtained in step 2 is processed using the Kemin normal distribution method. The purpose of the Kemin normal distribution method is to make the training set conform to a normal distribution and remove useless data, which is more conducive to the convergence of model training.

[0063] (2) To achieve a good fit between the multi-factor generalized linear regression model and the measured values ​​of the T2 methane sensor, the weights of the influencing factors in the multi-factor generalized linear regression model are trained using the least squares method. The formula is as follows:

[0064]

[0065] Where y is the original T2 methane sensor gas concentration in the training set. The methane concentration of the T2 sensor is predicted by a multifactor generalized linear regression model.

[0066] After 12 rounds of model training, the training loss results are as follows: Figure 2 As shown, from Figure 2 As can be seen, the more training is done, the less loss occurs, meaning the deviation between the predicted and the true values ​​becomes smaller. With each iteration, the multi-factor generalized linear regression model learns more and obtains more features, resulting in a smaller error.

[0067] 5. Comparison of Prediction Results

[0068] The test set was used for prediction evaluation, and the data in Table 2 below was obtained. In Table 2, mae is the mean absolute error, which is the absolute difference between the predicted value and the actual value (the average of the differences), mse is the mean variance, and rmse is the root mean square error.

[0069] Table 2 Error Analysis Table

[0070] Serial Number Evaluation parameters numerical values 1 mae 0.0037652504378184798 2 rmse 0.04720791635997296 3 mse 0.002228587367050202

[0071] As can be seen from the error analysis table in Table 2, the error between the predicted value and the actual value is within the acceptable error range, indicating that the model training process of the present invention is good and the model is usable.

[0072] like Figure 3 The image shows a comparison of the prediction results of the method of the present invention, where the curve "predict" represents the predicted value curve, and the curve "T2_max" represents the actual value curve collected by the T2 methane sensor; from Figure 3 As can be seen from the results, the predicted results of this invention are basically consistent with the changing trends of the actual values, achieving a relatively accurate prediction effect.

[0073] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting gas concentration in a coal mining face based on multi-factor generalized linear regression, characterized in that, The method comprises the following steps: S1, exporting data of various factors affecting the gas concentration of the T2 methane sensor from a MySQL database to establish a basic data set; based on the basic data set, the data in the basic data set is sorted to form a gas concentration influencing factor data set based on time correlation; S2, performing feature standardization processing on the gas concentration influencing factor data set to obtain a standardized data set for predicting the T2 methane sensor gas concentration at the next data collection time, and dividing the standardized data set into a training set and a test set by using a random sampling method; S3, based on the thickness of coal seam D, gas extraction amount C, daily output Q, wind speed S, T0 methane sensor gas concentration and the T2 methane sensor gas concentration to be predicted A multi-factor generalized linear regression model is established; The calculation formula of the multi-factor generalized linear regression model is: The above formula is simplified in a vector manner as: wherein, ω D is the offset weight of the coal seam thickness on the T2 methane sensor gas concentration, ω C is the offset weight of the gas extraction amount on the T2 methane sensor gas concentration, ω Q is the offset weight of the daily output on the T2 methane sensor gas concentration, ω s is the offset weight of the wind speed on the T2 methane sensor gas concentration, b is the offset correction constant; S4, processing the training set obtained in step S2 using the Kaiming normal distribution method, and training the offset weight of each influencing factor in the multi-factor generalized linear regression model by using the least square method; The calculation formula of the least square method is: where y is the original T2 methane sensor gas concentration in the training set, is the predicted T2 methane sensor gas concentration by the multi-factor generalized linear regression model.

2. The method for predicting gas concentration in a coal mining face based on multi-factor generalized linear regression according to claim 1, characterized in that, The data of various factors affecting the gas concentration of the T2 methane sensor exported from the MySQL database in step S1 includes: T0 methane sensor historical data, coal mining face production data, working face gas extraction amount data, working face coal seam thickness data, and return airway wind speed data.

3. The method for predicting gas concentration in a coal mining face based on multi-factor generalized linear regression according to claim 1, characterized in that, The method for performing feature standardization processing on the gas concentration influencing factor data set to obtain a standardized data set for predicting the T2 methane sensor gas concentration at the next data collection time in step S2 is as follows: (1) format the data tuple of the gas concentration influencing factor data set into a vector form: (T0_MAX, EXTRACT_COUT, COAL_PRODUCTION, COAL_THINKNESS, WIN_SPEED_AVG, NEXT_T2); (2) perform standardization processing on each column of the data tuple of the gas concentration influencing factor data set to obtain a standardized data set.

4. The method for predicting gas concentration in a coal mining face based on multi-factor generalized linear regression according to claim 3, characterized in that, The method for performing standardization processing on each column of the data tuple of the gas concentration influencing factor data set is as follows: 1) Average value solving for each column data of data tuples, formula is: Wherein, μ is average value, x i is the value of single sample data, N is the total number of sample data of each column data of data tuples; 2) Perform a standard deviation solution, formula is: wherein σ is the standard deviation; 3) z-score normalization of individual sample data for each column of data in the data tuple, formula:

Images