Mine gas monitoring and early warning method and system

Abstract

The invention relates to the field of mining safety, in particular to a mine gas monitoring and early warning method and system, and the method comprises the steps: obtaining mine data related to mine gas explosion; performing data preprocessing on the mine data to generate preprocessed data; introducing a sparrow algorithm, optimizing the sparrow algorithm, and constructing a parameter optimization model; constructing a mine gas monitoring and early warning model, and optimizing hyper-parameters of the mine gas monitoring and early warning model by using the parameter optimization model; and training and testing the mine gas monitoring and early warning model by using the preprocessed data, and performing mine gas monitoring and early warning by using the mine gas monitoring and early warning model. According to the invention, the problem that the mine gas explosion risk is difficult to quickly and accurately judge from real-time monitoring data due to the fact that mine gas monitoring and early warning excessively depends on manual analysis and judgment in related problems is solved. The method can quickly and accurately judge whether the current mine has the risk of gas explosion or not.

Description

Technical Field

[0001] The invention relates to the field of mining safety, and in particular to a mine gas monitoring and early warning method and system. Background Art

[0002] As the depth and intensity of coal mining increase, the amount of gas outflow increases. Coal mining operations destroy coal and rock mass, causing gas release. At the same time, the underground working environment is complex, with a variety of fire sources, such as blasting sparks, electric sparks generated by electrical equipment failure, and friction and impact sparks. If the ventilation system is unreasonable, gas is easy to accumulate. When the gas concentration reaches 5%-16%, it will cause an explosion when it encounters a fire source. Gas explosions not only cause casualties, but also destroy equipment, damage tunnels, and cause secondary disasters such as coal dust explosions, seriously affecting coal mine safety production and sustainable development.

[0003] In the existing technology, mine gas monitoring mainly relies on gas sensors and monitoring instruments. These devices can usually only provide gas concentration data for local areas and cannot fully reflect the gas distribution in the entire mine. In addition, traditional gas monitoring methods rely heavily on the experience and judgment of operators, which makes the accuracy and reliability of monitoring results affected by human factors and increases the risk of operational errors. Secondly, existing monitoring systems often lack effective data processing capabilities, and it is difficult to extract useful information from large amounts of complex data for accurate prediction of gas explosion risks. This limits the system's ability to conduct in-depth analysis and prediction of gas concentration trends. Therefore, more scientific, efficient, and accurate mine gas monitoring and early warning are of great significance. Summary of the invention

[0004] In view of the defects in the prior art, the present invention provides a mine gas monitoring and early warning method and system, which solves the problem that the mine gas monitoring and early warning in the related art relies too much on human analysis and judgment, so that the comprehensive judgment of mine gas explosion conditions is not fast and accurate enough.

[0005] In order to achieve the above-mentioned purpose, one aspect of the present invention provides a mine gas monitoring and early warning method, which includes: acquiring mine data related to mine gas explosions; performing data preprocessing on the mine data to generate preprocessed data; introducing a sparrow algorithm, optimizing the sparrow algorithm, and constructing a parameter optimization model; constructing a mine gas monitoring and early warning model, and using the parameter optimization model to optimize the hyperparameters of the mine gas monitoring and early warning model; using the preprocessed data to train and test the mine gas monitoring and early warning model, and using the mine gas monitoring and early warning model to perform mine gas monitoring and early warning.

[0006] After optimizing the hyperparameters of the width network neural network by the sparrow algorithm, the training speed can be accelerated, and a mine gas monitoring and early warning model with more rapid and accurate gas prediction can be obtained, thereby improving the accuracy of mine gas explosion judgment. At the same time, the method provided by the present invention can realize automated judgment and improve the speed of mine gas explosion judgment.

[0007] Optionally, the preprocessing of the mine data to generate preprocessed data includes: removing abnormal values ​​and interpolating missing values ​​from the mine data to generate multidimensional target data; and extracting features from the multidimensional target data to generate the preprocessed data.

[0008] By removing outliers and interpolating missing values ​​from mine data, the mine data has a higher credibility. At the same time, the processed data is processed using wavelet packet transform to extract characteristic data of small fluctuations in gas concentration, thereby improving the accuracy of model training.

[0009] Optionally, the introduction of the Sparrow algorithm is performed to optimize the Sparrow algorithm to construct a parameter optimization model, including: constructing a discoverer location update optimization model based on the Sparrow algorithm; constructing a joiner location update optimization model based on the Sparrow algorithm; constructing a sentinel location update optimization model based on the Sparrow algorithm; and constructing a parameter optimization model according to the discoverer location update optimization model, the joiner location update optimization model and the sentinel location update optimization model.

[0010] By optimizing the sparrow algorithm, compared with the traditional sparrow algorithm, the flexibility of data search is increased, which can effectively avoid falling into the local optimal solution and improve the accuracy of the output results.

[0011] Optionally, constructing a finder location update optimization model based on the sparrow algorithm includes: setting a first optimization condition and a second optimization condition according to a finder location update method of the sparrow algorithm; optimizing the finder location update method to obtain a finder location update optimization algorithm; constructing a first optimization model based on the first optimization condition and the finder location update optimization algorithm; constructing a second optimization model based on the second optimization condition and the finder location update optimization algorithm; and constructing the finder location update optimization model by combining the first optimization model and the second optimization model.

[0012] The present invention sets different conditions to perform more refined control on the search process, so that the algorithm adopts a more radical search strategy, thereby obtaining a wider solution space.

[0013] Optionally, the finder location update optimization model satisfies the following formula: ,in, For the A sparrow in the The iteration The value of the dimension, is the number of sparrows, is the current iteration number, represents the currently found global optimal position, is the maximum number of iterations, is the number of sparrows in the population, is the average value of all sparrow positions during iteration, is a random number between 0 and 1, is a random variable between 0 and 1, is the step size calculation parameter, is a random variable between 0 and 1, represents the standard deviation of the adjusted distribution, is a random variable between 0 and 1, Indicates A sparrow in the The random solution in the search range is searched at the iteration. represents the spiral search angle parameter, For the A sparrow in the The first iteration The value of the dimension, is a normally distributed random number, is the variable matrix to be optimized, is the warning value, is the safety threshold.

[0014] The present invention further increases the optimization conditions, realizes the refined control over the data search process, obtains a wider solution space, and can further avoid falling into the local optimal solution.

[0015] Optionally, the joiner location update optimization model satisfies the following formula: , in, For the A sparrow in the The iteration The value of the dimension, is a normally distributed random number, is the current iteration number, is the number of sparrows, For the dimension, is the position of the sparrow in the worst condition, For the A sparrow in the The iteration The value of the dimension, is the total number of the population, represents the search space, For the The best position at iteration .

[0016] The present invention increases the flexibility of search and reduces the possibility of falling into local optimum by introducing random variables.

[0017] Optionally, the alerter position update optimization model satisfies the following formula: , in, Indicates A sparrow in The value of the dimension, is the linear reduction factor, is the initial value of the linear reduction factor, is the current iteration number, is the final value of the linear reduction factor, is the maximum number of iterations, is the population size, is the number of sparrows, is the upper boundary of the dimensional search space, is the lower boundary of the dimensional search space, is the interaction force function between individuals, Expressed as A sparrow in The value of the dimension, Indicates The position of a sparrow, Expressed as A sparrow in The value of the dimension, Indicates The position of a sparrow, Expressed as The sparrow and The distance between sparrows, is the best individual in the current population.

[0018] The present invention increases the accuracy of the output result by further getting rid of the local optimum trap.

[0019] Optionally, the using the preprocessed data to train and test the mine gas monitoring and early warning model includes: dividing the preprocessed data into training data and test data; training the mine gas monitoring and early warning model using the training data; and testing the trained mine gas monitoring and early warning model.

[0020] The present invention divides the preprocessed data into training data and test data, which can avoid interference with the training model and the test model, and realizes mine gas monitoring and early warning through a scientific and effective method, thereby improving the accuracy and practicality of mine gas monitoring and early warning.

[0021] Optionally, the testing of the trained mine gas monitoring and early warning model includes: obtaining a mine gas monitoring and early warning comparison model; obtaining the true value of gas concentration and the true average value of gas concentration according to the test data; inputting the test data into the mine gas monitoring and early warning model and the mine gas monitoring and early warning comparison model, and outputting a predicted gas concentration; substituting the predicted gas concentration value and the true gas concentration value into a mean absolute error expression to obtain a mean absolute error data set; substituting the predicted gas concentration value and the true gas concentration value into a root mean square error expression to obtain a root mean square error data set; substituting the predicted gas concentration value, the true gas concentration value and the true average value of gas concentration into a determination coefficient expression to obtain a determination coefficient data set; and obtaining the performance test result of the mine gas monitoring and early warning model by analyzing the absolute error data set, the root mean square error data set and the determination coefficient data set.

[0022] By cross-validating and evaluating the performance of the mine gas monitoring and early warning model, an analysis and comparison can be made with the performance of other models, thereby ensuring the prediction effect of the mine gas monitoring and early warning model and improving the practicality and actual application capability of the mine gas monitoring and early warning model.

[0023] Another aspect of the present invention provides a mine gas monitoring and early warning system, which includes: a processor, an input device, an output device and a memory, wherein the processor, the input device, the output device and the memory are interconnected, wherein the memory is used to store a computer program, and the computer program includes program instructions, and the processor is configured to call the program instructions to execute a mine gas monitoring and early warning method described in the previous aspect of the present invention.

[0024] A mine gas monitoring and early warning system of the present invention has a compact structure, stable performance, high integration and simple composition, and can stably execute a mine gas monitoring and early warning method provided in the previous aspect of the present invention, further improving the overall applicability and practical application capability of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 The present invention is a flowchart of a mine gas monitoring and early warning method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0026] The specific embodiments of the present invention will be described in detail below. It should be noted that the embodiments described herein are only for illustration and are not intended to limit the present invention. In the following description, a large number of specific details are set forth in order to provide a thorough understanding of the present invention. However, it is obvious to those of ordinary skill in the art that these specific details do not need to be adopted to implement the present invention. In other examples, in order to avoid confusing the present invention, known circuits, software or methods are not specifically described.

[0027] Throughout the specification, references to "one embodiment," "an embodiment," "an example," or "an example" mean that a particular feature, structure, or characteristic described in conjunction with the embodiment or example is included in at least one embodiment of the present invention. Therefore, the phrases "in one embodiment," "in an embodiment," "an example," or "an example" appearing in various places throughout the specification do not necessarily all refer to the same embodiment or example. In addition, particular features, structures, or characteristics may be combined in one or more embodiments or examples in any suitable combination and / or subcombination. In addition, it should be understood by those of ordinary skill in the art that the figures provided herein are for illustrative purposes and that the figures are not necessarily drawn to scale.

[0028] In one embodiment of the present invention, a mine gas monitoring and early warning method is provided. It should be noted that, if Figure 1 As shown, the steps of the flowchart representing the method can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be performed in an order different from that herein.

[0029] See also Figure 1 , which solves the problem that the comprehensive judgment of mine gas explosion conditions in the related technology is not fast and accurate enough. The present invention obtains data related to mine gas explosion. First, the data must be processed to ensure the integrity and accuracy of the data, and then a monitoring model is constructed. The processed data is used to train and test the monitoring model to achieve the role of real-time monitoring and early warning of mine gas concentration, such as Figure 1 The figure is a flow chart of a mine gas monitoring and early warning method according to an embodiment of the present invention, the method comprising the following steps: Step S1, obtaining mine data related to mine gas explosion.

[0030] Wherein, obtaining the mine data related to the mine gas explosion specifically includes the following sub-steps: Step S101, obtaining the gas concentration data of the mine.

[0031] In this embodiment, the instrument used to obtain the gas concentration data of the mine includes a gas sensor installed in the mine. These sensors are generally placed in key locations such as the mining face, return air lane, and goaf where gas accumulation is likely to occur underground. For example, in the upper corner of the coal mining face, gas is prone to accumulate due to poor ventilation and other reasons. The gas sensor installed here can detect the gas concentration in real time. There are many types of sensors, the most common of which are catalytic combustion gas sensors and infrared absorption gas sensors. The catalytic combustion gas sensor uses the combustion of gas under the action of a catalyst to increase the temperature of the component, and then detects the gas concentration through changes in electrical signals; the infrared absorption gas sensor measures the concentration based on the absorption characteristics of gas to infrared light of a specific wavelength.

[0032] Gas concentration data needs to be obtained in real time, because the gas concentration in the mine may change in a short period of time due to mining activities, changes in the ventilation system, etc. For example, after a blasting operation, gas may surge out instantly, and real-time monitoring data can reflect this change in time so that emergency measures can be taken.

[0033] At the same time, in order to fully grasp the changing trend of gas concentration, data collection is continuous. Through long-term continuous data recording, the law of gas emission can be analyzed, such as the increase in gas emission at a certain time of the day or when mining a certain coal seam.

[0034] In addition, since even a small change in gas concentration can cause a safety accident, data accuracy is crucial. The accuracy of the sensor usually needs to meet certain standards, and regular calibration and maintenance are required to ensure that the collected data truly reflects the gas concentration in the well.

[0035] Step S102, obtaining ventilation condition data of the mine.

[0036] Ventilation condition data is one of the key indicators for measuring mine ventilation conditions. Good ventilation conditions can effectively dilute and discharge the gas in the mine, ensuring that the air quality underground meets safety standards.

[0037] The basic conditions for mine ventilation include: supplying enough fresh air underground to meet the oxygen needs of personnel; diluting harmful gases and dust underground to ensure safe production; regulating the underground climate to create a good working environment. In order to achieve these conditions, the mine ventilation system needs to have complete ventilation facilities and wind flow control facilities to ensure that a sufficient amount of fresh air is continuously supplied to each mining face underground and that the polluted air is discharged from the mine.

[0038] Mine ventilation methods are generally divided into four basic types according to the position relationship between the air intake shaft and the return air shaft: central, diagonal, regional and mixed. Central ventilation means that the air intake shaft and the return air shaft are both located in the center of the mine; diagonal ventilation means that the air intake shaft and the return air shaft are located at the two ends of the diagonal of the mine; regional ventilation means that the mine is divided into several areas, each area has its own air intake shaft and return air shaft; mixed ventilation is a combination of the above methods.

[0039] In this embodiment, the equipment used to obtain the ventilation condition data of the mine includes wind speed sensors. These sensors are usually installed in appropriate locations of ventilation tunnels, such as the air inlet and outlet of the main fan, each branch tunnel, etc. Common wind speed sensors include impeller wind speed sensors and ultrasonic wind speed sensors. The impeller wind speed sensor uses wind flow to drive the impeller to rotate, calculates the wind speed by measuring the speed of the impeller, and then obtains the ventilation volume. The ultrasonic wind speed sensor measures the wind speed based on the change in the propagation speed of ultrasonic waves in the wind flow.

[0040] Ventilation volume data can be used to evaluate the operating efficiency of the mine ventilation system. If the ventilation volume is lower than the design requirements, it may lead to safety hazards such as gas accumulation. Based on the ventilation volume data, ventilation managers can adjust the operating frequency or blade angle of the fan in time to ensure that the ventilation volume meets the needs of safe production.

[0041] Step S103, obtaining temperature data of the mine.

[0042] The temperature of the mine has a significant impact on the activity of gas. The main component of gas is methane, and the volume fraction of the explosion limit of gas is 5% to 15%. The increase in temperature will intensify the movement of gas molecules and enhance their activity. When the temperature rises to a certain level, the lower limit of gas explosion will decrease, that is, gas may explode at a lower concentration. For example, under normal circumstances, the gas concentration needs to reach 5% before it will explode, but in a high temperature environment, it may explode when the concentration drops to 4%.

[0043] At the same time, the temperature rise may be caused by underground electrical equipment failure, frictional heat, etc., which may also produce fire sources. If the gas concentration is within the explosion limit at this time, it will cause a gas explosion. Therefore, real-time mine temperature data monitoring can help predict the risk of gas explosion. Once the temperature rises abnormally and the gas concentration is close to the explosion limit, immediate measures need to be taken, such as stopping operations, cutting off power, and strengthening ventilation.

[0044] In this embodiment, the equipment used to obtain the temperature data of the mine includes temperature sensors. These temperature sensors are distributed in various key locations underground, such as mining working faces, transportation tunnels, return air tunnels, etc. There are various types of sensors, and the most common ones are thermocouple temperature sensors and thermal resistor temperature sensors. Thermocouple temperature sensors are based on the Seebeck effect, that is, in a loop composed of two different metal materials, when the temperatures at both ends are different, a thermoelectric potential is generated, and the temperature is determined by measuring the thermoelectric potential; thermal resistor temperature sensors use the property that the resistance of metals or semiconductors changes with temperature to measure temperature. They can collect temperature information underground in real time and accurately.

[0045] It should be noted that in some cases, manual measurement is also used as an auxiliary means. Workers can use portable thermometers to measure the temperature in specific areas, such as when inspecting newly developed lanes or areas where abnormal temperatures are suspected. This method can serve as a supplement to sensor data to ensure the integrity of temperature data.

[0046] Step S104, obtaining historical gas explosion event data of the mine.

[0047] In an optional embodiment, the sources for obtaining historical gas explosion event data of mines include: the State Administration of Mine Safety Supervision, emergency management departments at all levels, China National Knowledge Infrastructure and authoritative media.

[0048] Step S2, performing data preprocessing on the mine data to generate preprocessed data.

[0049] The step of preprocessing the mine data to generate preprocessed data specifically includes the following sub-steps: Step S201, removing abnormal values ​​and interpolating missing values ​​from the mine data to generate multi-dimensional target data.

[0050] In an optional embodiment, outliers caused by sensor failure, measurement error or extreme events can be eliminated by removing outliers, thereby improving the reliability of data and the accuracy of analysis. The outlier removal algorithm is used to remove outliers from mine data. Outliers refer to data that are significantly deviated from other data points in the data set. The classification of outlier removal algorithms includes methods based on statistical distribution, methods based on distance, and methods based on model prediction.

[0051] In this embodiment, using According to the principle of eliminating outliers, the mine data approximately obeys the normal distribution. According to the properties of the normal distribution, about 99.7% of the data are in In the range, represents the mean, represents the standard deviation. The outlier range is determined as less than and greater than data.

[0052] Mean Satisfies the following formula: , In the formula, Represents data, Indicates the amount of data, Indicates the position of the data in the data sequence.

[0053] Standard Deviation Satisfies the following formula: , In the formula, Represents data, Indicates the amount of data, Indicates the position of the data in the data sequence.

[0054] In some other embodiments, the box-and-whisker plot (IQR) method can also be used to eliminate outliers. The principle of the box-and-whisker plot is to determine outliers by the quartiles of the data. The first quartile Represents the lower 25% quantile and the third quartile of the data Indicates the upper 25% quantile and interquartile range of the data , usually less than or greater than The data points are considered as outliers.

[0055] At the same time, the K-nearest neighbor (KNN) distance method can also be used to eliminate outliers. The principle of the K-nearest neighbor distance method is to calculate the distance (such as Euclidean distance) between each data point in the mine data set and other data points. The distribution of the K nearest neighbor data points around the data point is used to determine whether it is an outlier. If the distance between a data point and its K nearest neighbor data points is significantly greater than the distance between other points, then it may be an outlier.

[0056] For example, the temperature sensor data in a mine. Assuming that there are multiple temperature sensors in an area, the data of each sensor is regarded as a data point to data point, and the Euclidean distance is calculated. Then, for each data point, find its K nearest neighbor data points, and calculate the average distance between the data point and its K nearest neighbor data points. If this average distance is greater than a certain threshold, where the threshold can be determined by analysis of normal data or expert experience, the data point is regarded as an outlier and removed.

[0057] The Euclidean distance satisfies the following formula: , In the formula, is the Euclidean distance, and For different data points, is the position of the data point in the data sequence, is the number of data points.

[0058] The missing value interpolation algorithm is used to interpolate missing values ​​in mine data, including mean interpolation, median interpolation, mode interpolation, regression-based interpolation and multiple interpolation. Interpolation of missing values ​​can ensure the integrity of the data set, avoid analysis bias or insufficient model training caused by missing data, and enhance the availability of data and the generalization ability of neural network models.

[0059] In this implementation, the mean interpolation method is used to interpolate missing values. When there are missing values ​​in the data set, the mean of the variable is used to fill the missing values. This method is simple and intuitive, and is suitable for situations where the data distribution is relatively uniform and the proportion of missing values ​​is relatively small.

[0060] The mean interpolation method interpolates missing values ​​to meet the following formula: , In the formula, Indicates the pending The The data of the parameters, Indicates The The data of the parameters, Indicates The The data of the parameters, Indicates the dimension sequence number of the parameter.

[0061] In an optional implementation, the missing values ​​can be interpolated using the median interpolation method. The principle of the median interpolation method is similar to the mean interpolation method, but the median is used to fill in the missing values. The median is the value in the middle after the data is sorted. For data with skewed distribution or data with outliers, the median can better represent the center position of the data than the mean. The specific steps of this method include: sorting the non-missing values ​​of the variables with missing values; finding the median. If the number of data points is odd, the median is the middle number; if the number of data points is even, the median is the average of the two middle numbers; replace the missing values ​​with the median.

[0062] In another optional embodiment, a regression-based interpolation method can also be used to interpolate missing values. The principle of the regression-based interpolation method is to establish a regression model and use other related variables to predict missing values. For example, in mine data, a regression model of mine temperature and ventilation volume, depth and other variables can be established. When temperature data is missing, the values ​​of other variables are used to predict the missing temperature value through the regression model.

[0063] In general, the removal of outliers and interpolation of missing values ​​in the mine data ensures that the subsequent gas concentration prediction model can receive clean, complete and consistent training data, thereby improving the accuracy and reliability of model prediction. Based on the results of the first data processing and the second data processing, multi-dimensional target data of mine gas is obtained.

[0064] Step S202, extracting features from the multi-dimensional target data to generate the pre-processed data.

[0065] In this implementation, wavelet packet transform is used to extract features from multidimensional target data and convert time series data into frequency domain features to capture subtle features of gas concentration changes.

[0066] Among them, the wavelet packet transform satisfies the following formula: , In the formula, Indicates the data to be analyzed. represents the translation parameter, represents the result of wavelet packet transform, It is the wavelet packet function, which is obtained by multiple decompositions of the wavelet function.

[0067] In an optional implementation, principal component analysis can also be used to extract features from multidimensional target data. Principal component analysis is a dimensionality reduction technique that converts multiple related variables into a few unrelated principal components. For multivariate data such as mine temperature and ventilation conditions, principal component analysis can extract the main feature information, reduce the data dimension, simplify the subsequent analysis and modeling process, and retain most of the data variation information.

[0068] In another optional embodiment, independent component analysis can be used to extract features from multidimensional target data. Independent component analysis can separate independent source signals from mixed signals. In mine data, multi-channel data such as temperature and ventilation can be regarded as mixed signals. Independent components can be separated by independent component analysis. Each component represents a source signal or feature pattern with specific physical meaning, which helps to mine potential information and features in the data.

[0069] Step S3, introducing the Sparrow Algorithm, optimizing the Sparrow Algorithm, and constructing a parameter optimization model.

[0070] The sparrow algorithm is a new type of swarm intelligence optimization algorithm inspired by the foraging and anti-predator behavior of sparrows. The sparrow algorithm was developed based on the foraging and anti-predator behavior of sparrows. In nature, when sparrows forage in groups, they will explore food sources (similar to finding the optimal solution to an optimization problem), obtain food, and adopt corresponding strategies when encountering dangers (such as threats from natural enemies). This biological group behavior pattern is abstracted into a mathematical model and used to solve various optimization problems.

[0071] Among them, the sparrow algorithm is introduced, the sparrow algorithm is optimized, and the construction of the parameter optimization model specifically includes the following sub-steps: Step S301, constructing a discoverer location update optimization model based on the sparrow algorithm.

[0072] In the sparrow algorithm, the fitness value is used to evaluate the quality of the sparrow individual (i.e., the quality of the solution), which directly affects the position update strategy of the discoverer, joiner, and alerter. In the present invention, the design of the fitness function is based on the root mean square error (RMSE) of the mine gas concentration prediction error, and its calculation formula is as follows: , in, Fitness function, is the root mean square error of the mine gas monitoring and early warning model corresponding to the current sparrow individual on the training set. The larger the fitness value, the smaller the model prediction error and the better the individual. The fitness value ranges from (0,1). The fitness value of the best individual approaches 1, and the worst individual approaches 0.

[0073] Among them, constructing a discoverer location update optimization model based on the sparrow algorithm specifically includes the following sub-steps: Step S30101, according to the finder position update method of the sparrow algorithm, set the first optimization condition and the second optimization condition. In this embodiment, the optimization condition is set to balance the search strategy and improve the success rate of finding the optimal solution, and the first optimization condition satisfies , using a global search strategy, the finder moves to the area between the optimal individual and the population average, expanding the search range, and the second optimization condition is satisfied , enhance the diversity of local search and avoid falling into local optimality, is a random variable between 0 and 1.

[0074] In an optional embodiment, the first optimization condition satisfies , the second optimization condition is satisfied , is a random variable between 0 and 1.

[0075] In another optional embodiment, the first optimization condition satisfies , the second optimization condition is satisfied , is a random variable between 0 and 1.

[0076] Therefore, the range of optimization conditions can be flexibly selected according to the actual situation. It should be noted here that is a random variable between 0 and 1. As two different strategies, according to the probability distribution, which strategy can better solve a certain problem? Then, a larger proportion can be allocated to the optimization condition corresponding to the strategy.

[0077] Step S30102, optimizing the finder location update method to obtain a finder location update optimization algorithm.

[0078] In this implementation, when the discoverer moves over a large area, two different location update modes are set to meet different location update strategies.

[0079] The first location update method satisfies the following formula: , in, For the A sparrow in the The iteration The value of the dimension, is the number of sparrows, is the current iteration number, represents the currently found global optimal position, is the maximum number of iterations, is the number of sparrows in the population, is the average value of all sparrow positions during iteration, A random number between 0 and 1.

[0080] The second position update method enables it to perform more effective global search to enhance its global search capability and avoid local optimal solutions, thereby improving the overall performance of the algorithm. The second position update method satisfies the following formula: , in, For the A sparrow in the The iteration The value of the dimension, represents the currently found global optimal position, is the step size calculation parameter, The value range of is between 0.1 and 1.5, and when , the step size is small, which is suitable for fine searching of the current area; when , the step size increases, and the ability to jump out of the local optimum is enhanced. , is a random variable between 0 and 1, represents the standard deviation of the adjusted distribution, is a random variable between 0 and 1, Indicates A sparrow in the The random solution in the search range is searched at the iteration. represents the spiral search angle parameter, ensuring the diversity of the search path, A random number between 0 and 1.

[0081] The standard deviation of the adjusted distribution satisfies the following formula: , in, represents the standard deviation of the adjusted distribution, Gamma function, represents the shape parameter of the conditioning distribution.

[0082] When the finder performs a fine search move, the following formula is satisfied; , in, For the A sparrow in the The iteration The value of the dimension, is the number of sparrows, is the current iteration number, For the A sparrow in the The iteration The value of the dimension, is a normally distributed random number, is the variable matrix to be optimized.

[0083] Step S30103: construct a first optimization model based on the first optimization condition and the finder position update optimization algorithm.

[0084] Among them, the first optimization model satisfies the following formula: , in, is the warning value, is the safety threshold, and the other character definitions are the same as the above character definitions.

[0085] , indicating that the environment is safe, the discoverer can search in a wider area near the current position, and its search range gradually shrinks with the increase of the number of iterations, which is achieved by multiplying by an exponential function less than 1. This allows the algorithm to quickly explore a large range of solution space in the early stage, and gradually focus on better areas in the later stage.

[0086] , indicating that there may be danger or that the food resources in the current area are not rich enough. The discoverer needs to lead the population to other safer or more promising areas. At this time, the discoverer's position update will be added with a random number that follows a normal distribution. and the variable matrix to be optimized to achieve a larger position shift, thereby guiding the population to avoid danger or find a better foraging area.

[0087] Step S30104: construct a second optimization model based on the second optimization condition and the finder position update optimization algorithm.

[0088] , The character definition is the same as the above character definition.

[0089] Step S30105, combining the first optimization model and the second optimization model to construct the discoverer location update optimization model.

[0090] Among them, the discoverer location update optimization model satisfies the following formula: , in, For the A sparrow in the The iteration The value of the dimension, is the number of sparrows, is the current iteration number, represents the currently found global optimal position, is the maximum number of iterations, is the number of sparrows in the population, is the average value of all sparrow positions during iteration, is a random number between 0 and 1, is a random variable between 0 and 1, is the step size calculation parameter, is a random variable between 0 and 1, represents the standard deviation of the adjusted distribution, is a random variable between 0 and 1, Indicates A sparrow in the The random solution in the search range is searched at the iteration. represents the spiral search angle parameter, For the A sparrow in the The iteration The value of the dimension, is a normally distributed random number, is the variable matrix to be optimized, is the warning value, is the safety threshold.

[0091] The above spiral search angle parameters satisfy the following formula: , represents the spiral search constant, Represents the search space dimension.

[0092] The above optimization conditions are introduced, namely Parameters can dynamically balance between global search and local search, which helps the algorithm achieve a better balance between exploration and development, thereby improving the algorithm's search efficiency and the quality of the solution, and is used to determine which strategy to use when the sparrow updates its position. The optimization condition satisfies the following formula: ,in is a fixed number of search cycles in [1,10], Represents the search space dimension.

[0093] Step S302: construct a joiner location update optimization model based on the sparrow algorithm.

[0094] Among them, the joiner position update optimization model satisfies the following formula: , in, For the A sparrow in the The iteration The value of the dimension, is a normally distributed random number, is the current iteration number, is the number of sparrows, For the dimension, is the position of the sparrow in the worst condition, For the A sparrow in the The iteration The value of the dimension, is the total number of the population, represents the search space, For the The best position at iteration .

[0095] Step S303: constructing a sentinel position update optimization model based on the sparrow algorithm.

[0096] In this implementation, the multi-path optimization and adaptive search strategies are introduced into the sentinel position update, which can effectively replace or enhance the early warning stage in the sparrow algorithm. Especially when dealing with complex optimization problems, the adaptive search strategy can effectively avoid falling into the local optimal trap.

[0097] Among them, the guard position update optimization model satisfies the following formula: , in, Indicates A sparrow in The value of the dimension, is the linear reduction factor, is the initial value of the linear reduction factor, is the current iteration number, is the final value of the linear reduction factor, is the maximum number of iterations, is the population size, is the number of sparrows, is the upper boundary of the dimensional search space, is the lower boundary of the dimensional search space, is the interaction force function between individuals, Expressed as A sparrow in The value of the dimension, Indicates The position of a sparrow, Expressed as A sparrow in The value of the dimension, Indicates The position of a sparrow, Expressed as The sparrow and The distance between sparrows, is the best individual in the current population.

[0098] The above linear reduction strategy helps the algorithm to perform global search in the early stage and local search in the later stage, thereby balancing the exploration and development capabilities of the algorithm.

[0099] The above-mentioned interaction force function between individuals satisfies the following formula: , is the interaction force function between individuals, is the distance between individuals, is the gravitational strength, is the gravitational scale range.

[0100] Step S304: construct a parameter optimization model according to the discoverer location update optimization model, the joiner location update optimization model and the sentinel location update optimization model.

[0101] Step S4, constructing a mine gas monitoring and early warning model, and optimizing the hyperparameters of the mine gas monitoring and early warning model using the parameter optimization model.

[0102] In this embodiment, a mine gas monitoring and early warning model is constructed based on a wide neural network. The hyperparameters of the mine gas monitoring and early warning model include the number of layers, the number of neurons in each layer, and the learning rate. The value range of the hyperparameters is determined based on experience and detailed investigation.

[0103] The value range of each hyperparameter constitutes a solution space. The parameter optimization model finds the optimal combination according to the fitness value of each solution, outputs the optimal monitoring and early warning parameters, and then uses the optimal monitoring and early warning parameters as the hyperparameters of the mine gas monitoring and early warning model. During training, the training of the mine gas monitoring and early warning model can be accelerated, and the performance of the mine gas monitoring and early warning model can be improved.

[0104] Specifically, the position of the sparrow population is first initialized, and the sparrows are sorted according to their individual fitness values. The sparrows corresponding to the 30% with the highest fitness values ​​in the population are selected as discoverers, and the rest are selected as joiners.

[0105] The parameter optimization model starts to iterate. The discoverer first updates its position, and then the joiner updates its position. After the joiner's position is updated, a 5% to 10% ratio of alerts is randomly generated between the discoverer and the joiner. Then the alerter completes the position update, and the above process is repeated until the maximum number of iterations is completed or the updated fitness value changes within the threshold range, completing the iterative calculation. Finally, the optimal solution is output to obtain the optimal monitoring and early warning parameters.

[0106] It should be noted that 30% discoverers and 70% joiners are an embodiment of the present invention, and are not a limitation of the present invention. In an optional embodiment, the ratio between the two can be flexibly selected according to different problems.

[0107] In an optional embodiment, the proportion of alerters is not fixed, usually between 5% and 20%. It should be noted that the proportion of alerters should not be too large or too small. If the proportion of alerters is too large, the algorithm will become too sensitive. In the sparrow algorithm, the main responsibility of the alerters is to sense danger and guide the population to change the search strategy. If the proportion of alerters is too high, they will frequently send out alarm signals, causing the entire sparrow population to change the search direction and range too frequently, thereby missing the true optimal solution and affecting the accuracy of the algorithm. If the proportion of alerters is too small, the algorithm's ability to perceive the local optimum will be weakened. In the sparrow algorithm, the alerters play a role similar to that of an early warning system, which can detect adverse conditions in the search space. Due to the lack of sufficient alerters to issue alarms and guide the population to conduct more extensive exploration, the algorithm may stay near the local optimal solution for too long. In the search space, the local optimal solution is like some small "traps" that will make the algorithm mistakenly believe that the optimal solution has been found.

[0108] The proportion of discoverers is fixed. Discoverers are individuals with higher fitness in the sparrow population. They usually represent sparrows with higher energy reserves and the ability to find food more effectively. Discoverers are responsible for finding food and providing foraging areas and directions for the entire sparrow population, leading the population's foraging activities.

[0109] The proportion of joiners is also fixed. Joiners are individuals in the population that rely on discoverers to obtain information about food resources. Joiners usually account for the majority of the population. They adjust their positions by observing the behavior and position of discoverers to obtain better food resources, which can be understood as a better solution in this embodiment.

[0110] A wide neural network is an artificial neural network architecture that features a wide hidden layer. The "wide" here refers to the relatively large number of neurons in the hidden layer. A wide neural network focuses on increasing the number of neurons in each layer to improve the network's expressive power.

[0111] Due to the large number of hidden layer neurons, wide neural networks can fit very complex functional relationships. For some highly nonlinear data distributions, it can better capture the characteristics and patterns in the data.

[0112] Compared with deep neural networks, wide neural networks may have higher computational efficiency during training and inference. Because there is no need to propagate information through too many layers, problems such as gradient vanishing or gradient exploding are reduced. At the same time, the key to wide neural networks is that they do not require complex back-propagation algorithms, but instead learn by directly calculating output weights, which makes wide neural networks more efficient when processing large-scale data sets.

[0113] The structure of a wide neural network is relatively simple, usually with fewer layers. This makes it easier to understand the role of each neuron and the working principle of the network as a whole.

[0114] The width neural network satisfies the following formula: , in, is the input data, are randomly generated feature map weights, is the bias term, is the feature mapping function, is the feature vector after mapping.

[0115] The prediction output satisfies the following formula: , in, is the feature matrix that combines the input features and the enhanced nodes, is the output weight obtained by pseudo-inverse calculation, is the predicted output of the network.

[0116] The performance of wide neural networks is sensitive to the choice of hyperparameters. The data optimized by the sparrow algorithm is used as the hyperparameters of the wide neural network, which improves the efficiency of model construction.

[0117] At the same time, wide neural networks may fall into local optimal solutions during training, resulting in the inability to find the global optimal model parameters. The parameter optimization model has a strong global search capability. The optimized hyperparameters can guide the wide neural network to jump out of the local optimal solution and explore a wider parameter space, thereby finding better model parameters and improving the performance and generalization ability of the model.

[0118] The optimized hyperparameters provide better initial parameters for the training of wide neural networks, avoiding invalid searches and iterations of wide neural networks during training, improving training efficiency, thereby accelerating the convergence of training and reducing training time.

[0119] In an optional embodiment, a mine gas monitoring and early warning model can also be constructed using a long short-term memory network, and the data optimized by the parameter optimization model is used as the hyperparameter of the mine gas monitoring and early warning model to speed up model training and improve model performance. The specific implementation process is similar to the above steps and will not be described in detail here.

[0120] Step S5, using the preprocessed data to train and test the mine gas monitoring and early warning model, and using the mine gas monitoring and early warning model to perform mine gas monitoring and early warning.

[0121] The method of using the preprocessed data to train and test the mine gas monitoring and early warning model specifically includes the following sub-steps: Step S501, dividing the preprocessed data into training data and test data.

[0122] In this embodiment, the preprocessed data is divided into training data and test data in a ratio of 7:3. In the training and testing process of the model, in order to ensure the stability and authenticity of the model performance, the training data and the test data cannot be confused. The ratio of training data to test data can be flexibly selected according to different actual conditions.

[0123] Step S502: training the mine gas monitoring and early warning model using the training data.

[0124] In this embodiment, the mine gas monitoring and early warning model usually converges to a relatively stable state after training. This means that after multiple iterations of optimization of the parameters (such as weights and biases) of the mine gas monitoring and early warning model, the loss function (such as mean square error loss, cross entropy loss, etc.) on the training data no longer decreases significantly.

[0125] Step S503, testing the trained mine gas monitoring and early warning model.

[0126] Among them, testing the trained mine gas monitoring and early warning model specifically includes the following sub-steps: Step S50301, obtaining a mine gas monitoring and early warning comparison model.

[0127] In this embodiment, multiple existing mine gas monitoring and early warning models are found, including a mine gas monitoring and early warning model that is not optimized by the sparrow algorithm but is directly trained by a wide neural network using training data, and a mine gas monitoring and early warning model that is trained by training data using an unimproved long short-term memory network LSTM and an improved long short-term memory network CNN-BILSTM.

[0128] Step S50302, obtaining the true value of the gas concentration and the true average value of the gas concentration based on the test data.

[0129] In this embodiment, the gas concentration data in the test data is extracted to obtain the true value of the gas concentration, and the true value of the gas concentration is summed and averaged to obtain the true average value of the gas concentration.

[0130] Step S50303, input the test data into the mine gas monitoring and early warning model and the mine gas monitoring and early warning comparison model, and output the predicted value of gas concentration.

[0131] In this embodiment, the data in the test data are input into each model to realize the simulation prediction of mine gas by each model, and the prediction result, that is, the predicted value of gas concentration, is output.

[0132] Step S50304, substituting the predicted value of the gas concentration and the actual value of the gas concentration into the mean absolute error expression to obtain a mean absolute error data set.

[0133] In this embodiment, the mean absolute error expression satisfies the following formula: , in, represents the mean absolute error, represents the number of samples, Indicates The true value of the concentration of the samples, Indicates The predicted concentration of each sample.

[0134] Step S50305, substituting the predicted value of the gas concentration and the actual value of the gas concentration into a root mean square error expression to obtain a root mean square error data set.

[0135] In this embodiment, the root mean square error expression satisfies the following formula: , in, represents the root mean square error, represents the number of samples, Indicates The true value of the concentration of the samples, Indicates The predicted concentration of each sample.

[0136] Step S50306, substitute the predicted value of gas concentration, the actual value of gas concentration and the actual average value of gas concentration into the determination coefficient expression to obtain a determination coefficient data set.

[0137] In this embodiment, the determination coefficient expression satisfies the following formula: , in, represents the coefficient of determination, represents the number of samples, Indicates The true value of the concentration of the samples, Indicates The predicted concentration value of samples, Represents the average value of the true concentration.

[0138] Step S50307, ​​obtaining the performance test result of the mine gas monitoring and early warning model by analyzing the absolute error data set, the root mean square error data set and the determination coefficient data set.

[0139] In this embodiment, the smaller the data in the absolute error data set and the data in the root mean square error data set are, the better the prediction effect of the model is.

[0140] In the data set of the coefficient of determination, when When , it means that the model fits the data completely accurately. In the regression model, the predicted value is exactly the same as the true value, and all data points are on the regression line or regression plane. This is an ideal fit state, indicating that the model can perfectly explain the changes in the dependent variable.

[0141] In the data set of the coefficient of determination, when When , it means that there is an error between the model prediction result and the true value, which is normal. The closer the determination coefficient is to 1, the better the prediction effect of the model. Taking the model of the present invention as an example, the determination coefficient is 0.9635, indicating that the small error is likely to be caused by measurement error, unconsidered dependent variables and random noise. A higher determination coefficient value means that the model can make good use of independent variables such as ventilation volume and temperature to explain the changes in gas concentration, and the prediction results are more reliable.

[0142] By analyzing the data in the absolute error data set, the root mean square error data set and the determination coefficient data set, it can be obtained that the model of the present invention has the most accurate prediction effect. The performance comparison between the mine gas monitoring and early warning model of the present invention and other models is shown in the following table: MAE RMSE R2 Mine gas monitoring and early warning model 8.123 10.269 0.9635 BLS 13.2697 21.6975 0.9128 LSTM 12.362 14.8113 0.9088 CNN-BILSTM 13.1442 15.2666 0.9425 The comparison of prediction and evaluation results shows that the mine gas monitoring and early warning model obtained by improving the parameter optimization model can effectively improve the prediction accuracy, and the root mean square error is reduced from 21.695 before the improvement to 10.269, and it has better optimization effect than the unimproved long short-term memory network (LSTM); at the same time, the root mean square error of the improved long short-term memory network (CNN-BILSTM) is 15.2666, which is higher than the root mean square error of 10.269 of the width learning neural network based on the parameter optimization model. In summary, the width learning neural network model based on the parameter optimization model has better fitting effect on the data than other models. In addition, the R2 before the improved algorithm is optimized is 0.9128, and the R2 after the improved algorithm is optimized is 0.9635. Combined with the comparison chart of the predicted value and the true value of each model, it can be seen that the present invention has the most accurate prediction result for the mine gas concentration, which provides new ideas and technical means for the prediction of different types of mine gas concentrations.

[0143] In summary, the parameter optimization model of the present invention can effectively search and determine the optimal value of the hyperparameter in the mine gas monitoring and early warning model, improve the generalization ability and prediction accuracy of the model, and secondly, the mine gas monitoring and early warning model combines the global search ability of the optimized sparrow algorithm and the efficient learning ability of the wide neural network, and can maintain low computational complexity and fast response time when processing a large amount of complex data, which is particularly important for real-time monitoring of changes in mine gas concentration. The optimized model can more accurately capture the pattern and trend of changes in gas concentration, reduce false alarms and missed alarms, and improve the reliability of the early warning system. Therefore, the present invention effectively overcomes the various shortcomings of existing mine gas concentration monitoring and early warning technologies and has a high industrial utilization value.

[0144] In an optional embodiment, the performance of the mine gas monitoring and early warning model is evaluated based on an independent test set, which requires obtaining new mine data that cannot participate in the training of the wide neural network.

[0145] It is relatively simple to evaluate the performance of the mine gas monitoring and early warning model through an independent test set. It only requires inputting new mine data into the mine gas monitoring and early warning model, and then analyzing the error between the predicted gas concentration result and the actual gas concentration.

[0146] In another optional embodiment, the performance of the mine gas monitoring and early warning model is evaluated simultaneously according to cross validation and an independent test set. Here, in order to avoid data interference, the independent test set should be unused data. By evaluating the performance of the model by two different methods at the same time, it can be ensured that the model has good generalization.

[0147] In one or some other embodiments of the present invention, before executing the step: using the mine gas monitoring and early warning model to perform mine gas monitoring and early warning, the performance of the mine gas monitoring and early warning model is evaluated based on cross-validation and / or independent test sets. The specific design steps and methods are basically similar to the above and will not be repeated here.

[0148] Finally, the mine gas monitoring and early warning model is deployed to the mine gas monitoring system. The model can be deployed on a local server or a cloud server and integrated with the mine's sensor data acquisition system. After the model is deployed, the mine gas monitoring data and the model's prediction results are continuously collected to monitor the actual application effect of the model. If it is found that the prediction accuracy of the model decreases or a large number of false positives and false negatives occur, the reasons need to be analyzed in a timely manner. And based on new data and feedback information, the model is regularly updated and optimized. This may include operations such as retraining the model, adjusting the model structure or parameters, and updating feature engineering to ensure that the model can adapt to changes in the mine environment and new data patterns.

[0149] On the other hand, the present invention also provides a mine gas monitoring and early warning system, comprising: a processor, an input device, an output device and a memory, wherein the processor, the input device, the output device and the memory are interconnected, wherein the memory is used to store a computer program, and the computer program includes program instructions, and the processor is configured to call the program instructions to execute relevant steps of a relevant embodiment of a mine gas monitoring and early warning method of the present invention.

[0150] In the mine gas monitoring and early warning system provided by the present invention, each functional component can be integrated into one processing component, or each component can exist physically separately, or two or more components can be integrated into one component. The above-mentioned integrated components can be implemented in the form of hardware or in the form of software functions.

[0151] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and specification of the present invention.

Claims

1. A mine gas monitoring and early warning method, characterized in that: The method comprises: Obtaining mine data related to mine gas explosions; Performing data preprocessing on the mine data to generate preprocessed data; Introducing the Sparrow Algorithm, optimizing the Sparrow Algorithm, and constructing a parameter optimization model; Constructing a mine gas monitoring and early warning model, and optimizing the hyperparameters of the mine gas monitoring and early warning model using the parameter optimization model; The preprocessed data is used to train and test the mine gas monitoring and early warning model, and the mine gas monitoring and early warning model is used to perform mine gas monitoring and early warning.

2. A mine gas monitoring and early warning method according to claim 1, characterized in that: The preprocessing of the mine data to generate preprocessed data includes: Eliminating outliers and interpolating missing values ​​from the mine data to generate multidimensional target data; Extract features from the multi-dimensional target data to generate the pre-processed data.

3. A mine gas monitoring and early warning method according to claim 1, characterized in that: The method of introducing the sparrow algorithm, optimizing the sparrow algorithm, and constructing a parameter optimization model includes: Constructing a discoverer position update optimization model based on the sparrow algorithm; Constructing a joiner position update optimization model based on the sparrow algorithm; Constructing a sentinel position update optimization model based on the sparrow algorithm; A parameter optimization model is constructed according to the discoverer location update optimization model, the joiner location update optimization model and the sentinel location update optimization model.

4. A mine gas monitoring and early warning method according to claim 3, characterized in that: The constructing of the discoverer location update optimization model based on the sparrow algorithm includes: According to the finder position update method of the sparrow algorithm, a first optimization condition and a second optimization condition are set; Optimizing the finder location update method to obtain a finder location update optimization algorithm; Based on the first optimization condition and the finder position update optimization algorithm, construct a first optimization model; Based on the second optimization condition and the finder position update optimization algorithm, construct a second optimization model; The finder location update optimization model is constructed by combining the first optimization model and the second optimization model.

5. A mine gas monitoring and early warning method according to claim 3, characterized in that: The finder location update optimization model satisfies the following formula: ， in, For the A sparrow in the The first iteration The value of the dimension, is the number of sparrows, is the current iteration number, represents the currently found global optimal position, is the maximum number of iterations, is the number of sparrows in the population, is the average value of all sparrow positions during iteration, is a random number between 0 and 1, is a random variable between 0 and 1, is the step size calculation parameter, is a random variable between 0 and 1, represents the standard deviation of the adjusted distribution, is a random variable between 0 and 1, Indicates A sparrow in the The random solution in the search range is searched at the iteration. represents the spiral search angle parameter, For the A sparrow in the The first iteration The value of the dimension, is a normally distributed random number, is the variable matrix to be optimized, is the warning value, is the safety threshold.

6. A mine gas monitoring and early warning method according to claim 3, characterized in that: The joiner location update optimization model satisfies the following formula: ， in, For the A sparrow in the The first iteration The value of the dimension, is a normally distributed random number, is the current iteration number, is the number of sparrows, For the dimension, is the position of the sparrow in the worst condition, For the A sparrow in the The first iteration The value of the dimension, is the total number of the population, represents the search space, For the The best position at iteration .

7. A mine gas monitoring and early warning method according to claim 3, characterized in that: The guard position update optimization model satisfies the following formula: ， in, Indicates A sparrow in The value of the dimension, is the linear reduction factor, is the initial value of the linear reduction factor, is the current iteration number, is the final value of the linear reduction factor, is the maximum number of iterations, is the population size, is the number of sparrows, is the upper boundary of the dimensional search space, is the lower boundary of the dimensional search space, is the interaction force function between individuals, Expressed as A sparrow in The value of the dimension, Indicates The position of a sparrow, Expressed as A sparrow in The value of the dimension, Indicates The position of a sparrow, Expressed as The sparrow and The distance between sparrows, is the best individual in the current population.

8. A mine gas monitoring and early warning method according to claim 1, characterized in that: The using the pre-processed data to train and test the mine gas monitoring and early warning model includes: Dividing the preprocessed data into training data and test data; Using the training data to train the mine gas monitoring and early warning model; The trained mine gas monitoring and early warning model is tested.

9. A mine gas monitoring and early warning method according to claim 8, characterized in that: The testing of the trained mine gas monitoring and early warning model includes: Obtain a mine gas monitoring and early warning comparison model; According to the test data, a true value of gas concentration and a true average value of gas concentration are obtained; Inputting the test data into the mine gas monitoring and early warning model and the mine gas monitoring and early warning comparison model, and outputting a predicted value of gas concentration; Substituting the predicted value of gas concentration and the true value of gas concentration into a mean absolute error expression to obtain a mean absolute error data set; Substituting the predicted value of gas concentration and the true value of gas concentration into a root mean square error expression to obtain a root mean square error data set; Substituting the predicted value of gas concentration, the actual value of gas concentration and the actual average value of gas concentration into a determination coefficient expression to obtain a determination coefficient data set; The performance test results of the mine gas monitoring and early warning model are obtained by analyzing the absolute error data set, the root mean square error data set and the determination coefficient data set.

10. A mine gas monitoring and early warning system, characterized in that: include: A processor, an input device, an output device and a memory, wherein the processor, the input device, the output device and the memory are interconnected, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is configured to call the program instructions to execute a mine gas monitoring and early warning method as described in any one of claims 1 to 9.

Images