A goaf coal spontaneous combustion early warning method, system, device and medium

Abstract

The present application relates to the technical field of coal mining safety monitoring, and discloses a goaf coal spontaneous combustion early warning method, system, device and medium. The method comprises the following steps: obtaining basic indexes of coal samples of different coal seams in a mine in a coal spontaneous combustion process and gas indexes of coal spontaneous combustion gas products to construct a basic data set; training a coal temperature prediction model constructed in advance by taking the basic data set as input and corresponding coal sample temperature as output, and iteratively optimizing hyperparameters of the coal temperature prediction model by using a particle swarm optimization algorithm during the training process; inversely calculating real-time coal sample temperature of a goaf of the mine according to real-time gas indexes of coal spontaneous combustion gas products of the goaf of the mine by using the trained coal temperature prediction model; combining a pre-trained time series model to determine an evolution law of historical coal sample temperature by using a time series of the coal sample temperature obtained by historical inversion, so as to predict a time series of future coal sample temperature of the goaf of the mine and perform coal spontaneous combustion early warning.

Description

Technical Field

[0001] This invention relates to the field of coal mining safety monitoring technology, and in particular to a method, system, equipment and medium for early warning of spontaneous combustion of coal in goaf areas. Background Technology

[0002] As the main energy source in the energy structure, coal mining has long faced the severe challenge of spontaneous combustion disasters. Among these, spontaneous combustion fires in goaf areas account for a prominent proportion of all types of coal mine fires. This not only seriously restricts the safe and efficient mining of coal and threatens the safety of people's lives and property, but also causes coal resource depletion and ecological environmental pollution.

[0003] Because the goaf is a closed environment with complex and irregular air leakage channels, the high-temperature point of coal spontaneous combustion is highly concealed. Traditional temperature sensors cannot penetrate deep into the goaf to directly measure the actual coal temperature. Therefore, the current mine site generally uses a bundled tube monitoring system to extract the gas inside the goaf. By analyzing the concentration evolution characteristics of the characteristic gases during the coal spontaneous combustion process, the state of coal spontaneous combustion can be indirectly determined. This technology has become the mainstream technical means for coal mine fire prevention and extinguishing monitoring.

[0004] Existing methods often rely on a single or a few key indicators, but these are easily affected by complex operating conditions such as underground blasting operations, gas outbursts, and ventilation fluctuations. Therefore, they cannot accurately identify the state of spontaneous combustion in coal, leading to a high false alarm rate. Furthermore, existing early warning methods are "state-response type," only able to estimate the real-time coal temperature in the goaf based on current gas concentration data. They lack in-depth analysis of the development patterns of spontaneous combustion and cannot predict coal temperature evolution trends in advance. This makes them unsuitable for the practical needs of proactive early warning and prevention of spontaneous combustion disasters in coal mines, resulting in a passive approach to fire prevention and extinguishing measures and an inability to effectively curb the spread of the disaster.

[0005] In summary, existing technologies cannot accurately provide early warning of spontaneous combustion of coal in goaf areas of coal mines. Summary of the Invention

[0006] The purpose of this invention is to provide a method, system, equipment, and medium for early warning of spontaneous combustion of coal in goaf areas, which can solve the above-mentioned technical problems.

[0007] An embodiment of the present invention provides a method for early warning of spontaneous combustion of coal in goaf areas, comprising the following steps: To construct a basic dataset, we need to obtain the basic indicators of coal samples from different coal seams in the mine during the spontaneous combustion process, as well as the gas indicators of the spontaneous combustion gas products. The basic indicators are used to indicate the physicochemical properties and spontaneous combustion tendency of the coal samples, and the gas indicators include the concentration of multiple single gases and the concentration ratio of different gases in the composite gas. Using the basic dataset as input and the corresponding coal sample temperature as output, a pre-built coal temperature prediction model is trained, and the hyperparameters of the coal temperature prediction model are iteratively optimized using the particle swarm optimization algorithm during the training process. The real-time coal sample temperature in the goaf is inverted by using a trained coal temperature prediction model based on the real-time gas indicators of the coal spontaneous combustion gas products in the goaf. By combining a pre-trained time series model with the coal sample temperature time series obtained through historical inversion, the evolution law of historical coal sample temperature can be determined to predict the future coal sample temperature time series in the goaf of the mine, and to provide early warning of coal spontaneous combustion.

[0008] Furthermore, prior to training the pre-built coal temperature prediction model, the following steps are also included: The K-means clustering algorithm was used to perform unsupervised clustering analysis on the optimized basic dataset to determine the temperature range for graded early warning of coal spontaneous combustion fires and the corresponding indicator data for each level of early warning. Following the inversion of the real-time coal sample temperature in the goaf of the mine, the following is also included: Based on the temperature range for graded early warning of coal spontaneous combustion fires and the corresponding indicator data for each level of early warning, the early warning level and indicator data corresponding to the real-time coal sample temperature in the goaf of the mine are determined in order to conduct early warning of coal spontaneous combustion.

[0009] Furthermore, the K-means clustering algorithm is used to perform unsupervised clustering analysis on the optimized basic dataset to determine the temperature range for graded early warning of coal spontaneous combustion fires and the corresponding indicator data for each level of early warning, including: Based on the basic indicators of the coal sample, the critical temperature and drying cracking temperature of the coal sample during the spontaneous combustion process were determined; among them, the critical temperature and drying cracking temperature are the key node temperatures at which significant physicochemical changes occur in the coal sample during the spontaneous combustion process. Based on the critical temperature and cracking temperature of the coal sample, clustering index data are selected from the optimized basic dataset and used as the sample dataset. Using the Euclidean distance between indicator data in the sample dataset as a similarity measure, the K-means clustering algorithm is adopted. Based on the preset number of clusters, the algorithm aims to minimize the intra-cluster variance by iteratively allocating indicator data and updating cluster centers to obtain the clustering results of the sample dataset.

[0010] Furthermore, the pre-built coal temperature prediction model includes a limit gradient boosting model, a support vector regression model, and a random forest model; The process of using a trained coal temperature prediction model to invert the real-time coal sample temperature in the goaf based on real-time gas indicators of spontaneous combustion gas products from the coal, including: The k-fold cross-validation method was used to evaluate the trained extreme gradient boosting model, support vector regression model and random forest model respectively, so as to determine the target coal temperature prediction model based on the regression performance and classification performance of the extreme gradient boosting model, support vector regression model and random forest model respectively. The target coal temperature prediction model, after being trained, is used to invert the real-time coal sample temperature in the goaf based on the real-time gas indicators of the spontaneous combustion gas products of coal in the goaf.

[0011] Furthermore, before training the pre-built coal temperature prediction model using the basic dataset as input and the corresponding coal sample temperature as output, the following steps are also included: Pearson correlation coefficient analysis was used to determine the degree of linear correlation between each index data in the basic dataset and the coal sample temperature, in order to optimize the basic dataset.

[0012] Furthermore, the basic indicators of the coal sample include the coal sample origin, the coal seam where the coal sample is located, the coal type, moisture, ash content, volatile matter, fixed carbon, coal sample temperature, furnace temperature, time, and temperature crossover point; The concentrations of the various gases and the concentration ratios of the different gases are the concentrations of O2, CO, CO2, and C2H4, as well as the C2H4 / CH4 ratio, CO2 / CO ratio, and CO / O2 ratio.

[0013] Furthermore, the temperature range for classifying and issuing early warnings for spontaneous combustion of coal is as follows: coal sample temperature T < 30℃ indicates low risk; 0℃ ≤ T < 60℃ indicates general risk; 60℃ ≤ T < 100℃ indicates significant risk; 100℃ ≤ T < 150℃ indicates major risk; and T ≥ 150℃ indicates extremely high risk.

[0014] Embodiments of the present invention also provide an early warning system for spontaneous combustion of coal in goaf areas, comprising: The data acquisition module is used to acquire the basic indicators of coal samples from different coal seams in the mine during the coal spontaneous combustion process, as well as the gas indicators of the coal spontaneous combustion gas products, obtained through coal spontaneous combustion experiments, in order to construct a basic dataset. Among them, the basic indicators are used to indicate the physicochemical properties and spontaneous combustion tendency of the coal samples, and the gas indicators include the concentration of multiple single gases and the concentration ratio of different gases in the composite gas. The model training module is used to train the pre-built coal temperature prediction model with the optimized basic dataset as input and the corresponding coal sample temperature as output. During the training process, the particle swarm optimization algorithm is used to iteratively optimize the hyperparameters of the coal temperature prediction model. The coal temperature prediction module is used to invert the real-time coal sample temperature in the goaf area of ​​the mine based on the real-time gas index of the coal spontaneous combustion gas products in the goaf area of ​​the mine using a trained coal temperature prediction model. The advanced early warning module is used to combine a pre-trained time series model with the coal sample temperature time series obtained by historical inversion to determine the evolution law of historical coal sample temperature, so as to predict the future coal sample temperature time series in the goaf of the mine and carry out advanced early warning of coal spontaneous combustion.

[0015] Embodiments of the present invention also provide a computer device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the above-described method for early warning of spontaneous combustion of coal in goaf areas.

[0016] Embodiments of the present invention also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-described method for early warning of spontaneous combustion of coal in goaf areas.

[0017] The method for early warning of spontaneous combustion of coal in goaf areas provided by this invention has at least the following beneficial effects: This invention combines fundamental indicators of coal samples with gaseous indicators of coal spontaneous combustion products as data for coal temperature prediction. The fundamental indicators of the coal sample indicate its physicochemical properties and spontaneous combustion tendency. The gaseous indicators include the concentrations of various single gases and the concentration ratios of different gases in a complex gas mixture, not limited to single gas concentrations or multiple gas concentrations. This enriches the gaseous indicators and adds to the fundamental indicators of the coal sample, effectively constructing characteristic data for accurate coal temperature prediction. Furthermore, based on the aforementioned characteristic data and a coal temperature prediction model that predicts the coal temperature of mine goaf areas in real time, a time series model is combined with historical coal temperature sequences over multiple days as input to directly output continuous coal temperature prediction values ​​for the goaf areas over the next few days. This enables early warning of coal spontaneous combustion, buying time for on-site intervention. Attached Figure Description

[0018] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this invention, illustrate exemplary embodiments of the invention and are used to explain the invention, but do not constitute an undue limitation of the invention. In the drawings:

[0019] Figure 1 A flowchart of an early warning method for spontaneous combustion of coal in goaf areas provided by the present invention; Figure 2 A technical roadmap provided for this invention; Figure 3 A diagram illustrating the gas production pattern of coal sample heating is provided for this invention. Figure 4 A comparison chart of root mean square error values ​​for each model during the iteration process, provided by the present invention; Figure 5 This invention provides a comparison chart of the root mean square error values ​​of each model on different coal types; Figure 6 This invention provides a comparison chart of the classification performance of various models on a test set; Figure 7 A flowchart for ARIMA modeling is provided by this invention; Figure 8 This invention provides an ARIMA model for predicting the future coal temperature trend. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0021] The technical solutions provided by the various embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0022] One embodiment of the present invention relates to a method for early warning of spontaneous combustion of coal in goaf areas. The specific process of the method for early warning of spontaneous combustion of coal in goaf areas in this embodiment can be as follows: Figure 1 As shown, it includes: Step 101: Obtain the basic indicators of coal samples from different coal seams in the mine during the coal spontaneous combustion process and the gas indicators of coal spontaneous combustion gas products obtained through coal spontaneous combustion experiments, so as to construct a basic dataset; wherein, the basic indicators are used to indicate the physicochemical properties and spontaneous combustion tendency of the coal samples, and the gas indicators include the concentration of multiple single gases and the concentration ratio of different gases in the composite gas. Step 102: Using the basic dataset as input and the corresponding coal sample temperature as output, train the pre-built coal temperature prediction model, and use the particle swarm optimization algorithm to iteratively optimize the hyperparameters of the coal temperature prediction model during the training process. Step 103: Using the trained coal temperature prediction model, the real-time coal sample temperature in the goaf is inverted based on the real-time gas index of the coal spontaneous combustion gas products in the goaf. Step 104: Combining the pre-trained time series model, the evolution law of historical coal sample temperature is determined through the coal sample temperature time series obtained by historical inversion, so as to predict the future coal sample temperature time series in the goaf of the mine and carry out early warning of coal spontaneous combustion.

[0023] The following is a detailed description of the implementation details of the early warning method for spontaneous combustion of coal in goaf areas in this embodiment. The following content is only for the convenience of understanding and is not necessary for implementing this solution.

[0024] In step 101, coal samples from different coal seams in the mine are collected to conduct coal spontaneous combustion programmed heating experiments and industrial analysis experiments. The experimental coal samples must be crushed and screened, separating the original outer shell and selecting the core portion. The samples are then crushed under a nitrogen protective atmosphere, and the particle sizes are mixed for the experiment. Before the experiment, the samples should be vacuum dried and sealed for storage. Data obtained from the experiments, including coal sample origin, coal seam, coal type, moisture content, ash content, volatile matter, fixed carbon, coal sample temperature, furnace temperature, time, and temperature crossover points, are collected to construct basic parameters for the coal samples. These parameters are combined with coal spontaneous combustion gas indices, which include the concentrations of various single gases and the concentration ratios of different gases in a composite gas mixture (such as the concentrations of O2, CO, CO2, and C2H4, as well as the C2H4 / CH4 ratio, CO2 / CO ratio, and CO / O2 ratio), to construct a basic dataset.

[0025] In one example, after obtaining the basic dataset, missing value imputation and normalization preprocessing are performed on the basic dataset. Then, the correlation strength between each indicator data and coal temperature is calculated by Pearson correlation coefficient analysis. The indicator data with correlation coefficients that meet the preset threshold are selected as the early warning indicator features for coal spontaneous combustion, thereby optimizing the basic dataset.

[0026] Missing value imputation involves filling in missing values ​​using the mean of the missing parameter data, and setting missing values ​​of invalid data to 0. Due to differences in the units and dimensions of the data, normalization preprocessing is required to ensure good comparability between the data.

[0027] The specific formula for calculating the Pearson correlation coefficient is as follows: ; In the formula, , For features in the dataset, For time hour The specific value, For time hour The specific value, For time hour The average value, For time hour The average value.

[0028] When the absolute value of the correlation coefficient is between 0.8 and 1.0, the correlation is extremely strong; when the absolute value of the correlation coefficient is between 0.5 and 0.8, the correlation is strong; when the absolute value of the correlation coefficient is between 0.2 and 0.5, the correlation is moderate; and when the absolute value of the correlation coefficient is between 0 and 0.2, the correlation is weak.

[0029] In step 102, the determined coal spontaneous combustion early warning index features are used as model inputs and the corresponding coal temperature is used as model outputs to construct three basic regression prediction models: Limiting Gradient Boosting (XGBoost), Support Vector Regression (SVR), and Random Forest (RF). The Particle Swarm Optimization (PSO) algorithm is then used to perform global optimization iteration on the core hyperparameters of the XGBoost, SVR, and RF models to obtain the optimal hyperparameter combination for each model.

[0030] The basic dataset for the prediction model includes a training set and a test set. The split ratio can be flexibly determined according to the size of the dataset. When the sample size is sufficient, an 8:2 or 7:3 ratio is preferred to ensure the sample coverage of the training set and the validation effectiveness of the test set.

[0031] The PSO algorithm iteratively optimizes the core hyperparameters of XGBoost, SVR, and RF models globally. The iteration process relies on the Particle Swarm Optimization algorithm to continuously update the particle velocity and position parameters, driving the hyperparameters of each model to gradually converge towards the global optimum. To achieve this convergence, several key parameters need to be appropriately set and selected, with inertia weights and acceleration coefficients being the core parameters. Inertia weights characterize the strength with which particles maintain their original velocity during the search process, while acceleration coefficients primarily affect the particle's exploration rate and direction of motion within the search space. The scientific selection of these parameters plays a decisive role in the PSO algorithm's performance and convergence characteristics. Optimizing these key parameters effectively enhances the algorithm's adaptability to different optimization problems and improves its generalization performance. Reasonable parameter control allows the PSO algorithm to better adapt to various optimization scenarios and possesses a more robust global search capability.

[0032] Among them, the particle swarm optimization functions are the particle velocity update and position update functions. , ; In the formula, k is the number of iterations. This represents the position of the particle after k iterations. Let the velocity be the particle's velocity after k iterations. , As a learning factor, , As an acceleration factor, For inertial weights, , These are the local optimal solution and the global optimal solution of the particle swarm, respectively.

[0033] The limiting gradient boosting consists of n base models. Model loss function Model objective function , ; In the formula, For the i-th sample, Let K be the predicted value for the i-th sample, and K be the number of base models. This is the k-th base model.

[0034] The objective function of support vector machine regression Model constraints , , The target value predicted by the model ; In the formula, Here, C represents the model's weight vector, and C is the regularization parameter used to control the model's complexity. and 'b' is a slack variable, allowing some data points to not meet the interval condition. 'N' is the number of training data points, and 'b' is the bias term of the model. For a predefined interval threshold, Input features The mapping.

[0035] Random Forest Sample Gini Index Function The sample set S is divided into two subsets S1 and S2, and the Gini index for the division is... ; In the formula, n is the number of sample categories, S is the dataset, and P i Let S be the proportion of the i-th type of sample in S.

[0036] The core hyperparameters of the model include the number of particles, local acceleration coefficient, global acceleration coefficient, inertia weight, and number of iterations. The optimal hyperparameter combination is determined by setting a fitness function for the fit, and updating the individual and global optimum values ​​through particle fitness calculations during iteration. When the root mean square error reaches its minimum, if subsequent iterations trigger overfitting, the iteration is terminated. At this point, the model converges to the point of minimum root mean square error, and the optimal hyperparameter combination is finally determined. Based on the theories of PSO and XGBoost algorithms, PSO is applied to the parameter optimization of the XGBoost regressor to construct the PSO-XGBoost model. Similarly, PSO-SVR and PSO-RF models are constructed.

[0037] The steps for building the PSO-XGBoost model are as follows: (1) For the various index data obtained in the experiment, a series of data preprocessing operations such as data cleaning, data integration and data transformation are carried out in sequence; (2) Calculate the Pearson correlation coefficient for the preprocessed dataset, and select indicators with high correlation to coal temperature based on the magnitude of the correlation coefficient; (3) According to an appropriate ratio, the selected index data and coal temperature data are divided into training set and test set, a suitable fitness function is constructed, and the individual optimal value and global optimal value of the particles are initialized. (4) The PSO algorithm is used to optimize the hyperparameters of the model. The velocity and position of the particles are continuously updated through the iterative process to search for the global optimal solution of the hyperparameters. The fitness value of the particles is calculated in each iteration, and the individual optimal value and the global optimal value of the population are dynamically updated accordingly until the preset iteration termination condition is met and the optimization stops. (5) Select the optimal parameter values ​​to construct the parameter-optimized XGBoost model, train it on the test set, and obtain the coal temperature prediction results; Understandably, the PSO-SVR and PSO-RF coal spontaneous combustion fire early warning models are modeled in the same way.

[0038] In the specific implementation, the coal temperature prediction models obtained by the above methods include the extreme gradient boosting model, the support vector regression model, and the random forest model. The k-fold cross-validation method is used to evaluate the extreme gradient boosting model, the support vector regression model, and the random forest model respectively. Based on the regression performance and classification performance of the extreme gradient boosting model, the support vector regression model, and the random forest model, the target model is determined for the final prediction.

[0039] k-fold cross-validation divides the training set into k independent subsets of equal size. In each iteration, one subset is selected as the validation set, and the remaining k-1 subsets are used as the training set. This process is repeated k times to ensure that each subset participates in both training and validation. The value of k can be determined based on the size of the dataset, or k-fold validation can be performed multiple times and the average value is taken.

[0040] The regression performance evaluation metric includes the root mean square error (RMSE), and the formula is as follows: Mean Absolute Error (MAE), the formula is: Mean Absolute Percentage Error (MAPE), the formula is as follows: The coefficient of determination (R²) is calculated using the following formula: ; In the formula, n is the total amount of sample data, and j is the amount of sample data. This is the predicted coal temperature value. This represents the actual coal temperature. This represents the average of the actual coal temperature values.

[0041] Classification performance evaluation metrics include model accuracy (F1-SCORE), the formula is as follows: Precision, the formula is: Accuracy, the formula is: Recall rate, the formula is: ; In the formula, TP represents the number of true positive samples, TN represents the number of true negative samples, FP represents the number of false positive samples, and FN represents the number of false negative samples.

[0042] In step 103, the target coal temperature prediction model is input using test set data, and the predicted coal temperature is output for model verification. Then, the real-time gas data of the mine bundle tube monitoring system is obtained and input into the verified target coal temperature prediction model to invert the current coal temperature in the goaf.

[0043] In step 104, by combining a time series model and based on historical predicted coal temperature sequences, the coal temperature evolution trend and risk level of the goaf area in the following 7 days are predicted, so as to achieve early warning of spontaneous combustion of coal in the goaf area.

[0044] The time series model uses an autoregressive integral moving average (ARIMA) model. The input to the ARIMA model is a series of coal temperature data from goaf inversion over several consecutive days, including timestamps, and the output is the continuous coal temperature forecast for the next 7 days.

[0045] In some embodiments, before training the coal temperature prediction model, the K-means clustering algorithm is used to perform unsupervised clustering analysis on the optimized basic dataset to determine the temperature range for graded early warning of coal spontaneous combustion fires and the corresponding indicator data for each warning level. Then, after outputting the predicted coal temperature, the warning level and indicator data corresponding to the real-time coal sample temperature in the mine goaf are determined based on the temperature range for graded early warning of coal spontaneous combustion fires and the corresponding indicator data for each warning level, for the purpose of issuing early warnings for coal spontaneous combustion. That is, the corresponding risk levels are divided according to the determined graded early warning temperature ranges, and then the predicted coal temperature and risk levels are compared with the actual data to verify the accuracy of the target coal temperature prediction model.

[0046] Unsupervised clustering analysis employs the K-means clustering algorithm, using an iterative optimization strategy to assign data points to clusters, aiming to minimize intra-cluster variance, until the cluster centers tend to stabilize or the preset number of iterations is reached. In the SPSS environment, the K-means clustering method is selected, and a phylogenetic map is constructed based on coal temperature and related indicators. Finally, the classification results are organized according to the temperature sequence to form the K-means clustering discrimination results of the coal spontaneous combustion process.

[0047] Specifically, based on the critical temperature and cracking temperature of the coal sample (coal-oxygen composite characteristics), clustering index data are selected from the optimized basic dataset as the sample dataset. The Euclidean distance between the index data in the sample dataset is used as the similarity measure. The K-means clustering algorithm is adopted. According to the preset number of clusters, during the iteration process, the sum of squared errors of data points within the cluster gradually decreases. The index data is continuously allocated and the cluster centers are updated through iteration until the stopping condition is reached (such as the cluster centers no longer changing significantly or reaching the predetermined number of iterations), and the clustering results of the sample dataset are obtained.

[0048] The basic steps are as follows: (1) Select the number of clusters K; (2) Randomly select K from all data points as the initial cluster centers for the clustering process; (3) Assign each data point to the cluster containing the nearest cluster center; (4) Calculate the new cluster center for each cluster, which is the mean of all data points in the corresponding cluster; (5) Repeat (3) and (4) and continue the iteration process until the cluster center no longer changes significantly, or the loop is exited when the number of iterations reaches the preset threshold.

[0049] The graded warning temperature ranges are divided into four categories based on K-means++ cluster analysis: green (low risk), blue (moderate risk), yellow (significant risk), orange (major risk), and red (extremely high risk). Coal sample temperature T < 30℃ indicates low risk; 0℃ ≤ T < 60℃ indicates moderate risk; 60℃ ≤ T < 100℃ indicates significant risk; 100℃ ≤ T < 150℃ indicates major risk; and T ≥ 150℃ indicates extremely high risk.

[0050] The beneficial effects of this invention are as follows: (1) Scientific and objective feature extraction: Through Pearson correlation analysis, redundant gas features unrelated to coal temperature changes are eliminated, and an index system combining single components and composite components is established, improving the quality of model input; combined with the K-means++ algorithm, the subjectivity of manually setting temperature thresholds is overcome, making the early warning classification more in line with the inherent laws of the data. (2) Significantly improved prediction accuracy: The Particle Swarm Optimization (PSO) algorithm is introduced to automatically optimize the hyperparameters of models such as XGBoost globally. The PSO algorithm has a fast convergence speed and strong optimization ability, effectively avoiding overfitting or underfitting problems caused by manual parameter tuning. The optimized target model has extremely high nonlinear fitting ability and generalization when inverting the hidden coal temperature in the goaf. (3) Realized true early warning: It breaks through the limitation of traditional models that can only predict the current state. Based on the nonlinear inversion of the current coal temperature, this invention innovatively connects the time series model. By utilizing historical coal temperature data to construct a time series, the coal temperature path for the next 7 days can be predicted in a rolling manner, providing a valuable time window for on-site fire prevention and extinguishing intervention measures such as grouting and nitrogen injection.

[0051] In a specific embodiment, the 507 fully mechanized longwall face in a certain mining area is used as the application background. This longwall face adopts the total caving method for roof management, the coal seam has a spontaneous combustion tendency of Class II, the spontaneous ignition period is 42 days, and the risk of spontaneous combustion of coal in the goaf is relatively high. The advanced early warning method for spontaneous combustion of coal in the goaf based on particle swarm optimization algorithm provided in this embodiment, the overall technical route of which can be found in [reference needed]. Figure 2 The process is strictly divided into three stages: data acquisition and feature engineering, algorithm model construction and optimization evaluation, and application of early warning in mines. The specific steps are as follows:

[0052] Phase 1: Data Acquisition and Characterization. Fresh coal samples were collected from the coal seam of the 507 fully mechanized longwall face in this mine. The original outer shell of the coal samples was removed, and the core portion was selected. The samples were crushed and screened under a nitrogen protective atmosphere, then vacuum-dried, sealed, and stored before being sent to the laboratory. A programmed temperature rise experiment was conducted using a coal spontaneous combustion test platform to obtain basic parameters such as moisture, ash, volatile matter, and fixed carbon content of the coal samples. During the experiment, the concentrations of gases such as O2, CO, CO2, C2H4, CH4, and C2H6 produced by coal spontaneous combustion at different temperature points were continuously analyzed and recorded using gas chromatography to obtain the gas production patterns of the coal samples during temperature rise. (See [link to relevant documentation]). Figure 3An initial coal temperature prediction dataset was generated by integrating all experimental data. This dataset was then divided into training and testing sets at a 4:1 ratio. The initial dataset underwent data cleaning to remove outliers and invalid values ​​caused by instrument errors. All data were normalized to eliminate the influence of differences in units and dimensions. The Pearson correlation coefficient formula was used to calculate the linear correlation between each gas indicator variable and the coal temperature variable. Feature variables with an absolute correlation coefficient greater than 0.2 were extracted, ultimately establishing seven early warning indicator gas input features: the concentrations of O2, CO, CO2, and C2H4, as well as the alkane ratios C2H4 / CH4, CO2 / CO, and CO / O2. Sample datasets were selected based on the coal-oxygen composite characteristics (critical temperature, cracking temperature) observed in the programmed heating experiment. The K-means++ clustering algorithm was used, with a specified number of clusters k=5, and iterative clustering was performed using the Euclidean distance between samples as a similarity metric. Ultimately, the spontaneous combustion process of coal is scientifically divided into five graded warning temperature ranges: Green warning (low risk): T<30℃; Blue warning (moderate risk): 30℃≤T<60℃; Yellow warning (significant risk): 60℃≤T<100℃; Orange warning (major risk): 100℃≤T<150℃; and Red warning (extremely high risk): T≥150℃.

[0053] Phase Two: Three machine learning algorithms—Limited Gradient Boosting Tree (XGBoost), Support Vector Regression (SVR), and Random Forest (RF)—were established as foundational candidate models. The core parameters of the Particle Swarm Optimization (PSO) algorithm were initialized, setting the population size to 30, the maximum number of iterations to 100, and the learning factors c1=c2=2.0. The root mean square error (RMSE) of the model on the training set was used as the particle fitness function. The PSO algorithm drives the particles to continuously update their velocity and position within the hyperparameter space, achieving global optimization of the core hyperparameters of each foundational model. For the XGBoost model, the focus was on finding the optimal learning_rate (search range 0.01~0.3), n_estimators (number of iterations, range 50~500), and max_depth (tree depth, range 3~10). The changes in the root mean square error of each model during the iteration process are detailed in [link to documentation]. Figure 4 After global iterative optimization, the optimal hyperparameter combinations for PSO-XGBoost, PSO-SVR, and PSO-RF are output respectively. Five-fold cross-validation is used to rigorously train and evaluate the three optimized models, and the mean RMSE, mean absolute error (MAE), and coefficient of determination (R²) of each model are statistically analyzed. 2 For a comparison of the root mean square error values ​​of each model on different coal types, please refer to [link / reference]. Figure 5 The comparison revealed that the PSO-XGBoost model has the smallest error and the highest R-value.2 Therefore, the PSO-XGBoost model was established as the final optimal performing model. Test set data was input into the PSO-XGBoost model to obtain coal temperature prediction results. Combined with the graded early warning thresholds determined in the first stage, the predicted coal temperature was converted into the corresponding risk level. The prediction results were compared and verified with the true labels. For a comparison of the classification performance of each model on the test set, please refer to [link to relevant documentation]. Figure 6 The results show that the risk level classification accuracy of the PSO-XGBoost model reaches over 90%, the model is robust, and meets the needs of on-site early warning.

[0054] Phase 3: The trained PSO-XGBoost model is deployed to the mine fire prevention and extinguishing monitoring system of the 507 fully mechanized longwall face in the mine. The system receives real-time gas concentration data from the underground goaf bundled tube monitoring system. After automatically calculating the composite gas index, all early warning characteristic indicators are input into the PSO-XGBoost model to achieve real-time and accurate inversion of the temperature of the hidden coal in the deep goaf, and automatically outputs the risk level of spontaneous combustion of coal in the goaf (e.g., when the inverted coal temperature is 55℃, the system automatically lights up a "blue warning" signal). The system backend simultaneously starts the autoregressive integral moving average (ARIMA(p,d,q)) time series module. For the ARIMA model modeling process, please refer to [link to ARIMA model modeling process]. Figure 7 The coal temperature data from the past 15 days, derived from the PSO-XGBoost model, was input as a stationary time series into the ARIMA model. The model order was determined using the autocorrelation function (ACF) and partial autocorrelation function (PACF). The ordered ARIMA model was then used to extrapolate to future timelines, outputting predicted coal temperatures for the next 1 to 7 days in the goaf. The trend prediction curve for future coal temperatures using the ARIMA model can be found in [reference needed]. Figure 8 If the ARIMA model predicts that the coal temperature in the goaf of the working face will exceed 60℃ and enter the "yellow warning (higher risk)" range on the 5th day, the host computer system will immediately trigger an advanced audible and visual alarm. Based on this advanced warning information, mine fire prevention and extinguishing technicians can formulate fire prevention and extinguishing intervention plans in advance, such as nitrogen injection for equalization and injection of three-phase foam, to complete targeted prevention and control during the incubation period of coal spontaneous combustion fire, thereby achieving proactive defense against coal spontaneous combustion disasters in the goaf and effectively avoiding fire accidents.

[0055] The steps of the various methods described above are only for clarity. In practice, they can be combined into one step or some steps can be split into multiple steps. As long as they include the same logical relationship, they are all within the protection scope of this invention. Adding insignificant modifications or introducing insignificant designs to the algorithm or process, without changing the core design of the algorithm and process, are also within the protection scope of this invention.

[0056] Another embodiment of the present invention relates to an early warning system for spontaneous combustion of coal in goaf areas. This embodiment of the early warning system for spontaneous combustion of coal in goaf areas includes: The data acquisition module is used to acquire the basic indicators of coal samples from different coal seams in the mine during the coal spontaneous combustion process, as well as the gas indicators of the coal spontaneous combustion gas products, obtained through coal spontaneous combustion experiments, in order to construct a basic dataset. Among them, the basic indicators are used to indicate the physicochemical properties and spontaneous combustion tendency of the coal samples, and the gas indicators include the concentration of multiple single gases and the concentration ratio of different gases in the composite gas. The model training module is used to train the pre-built coal temperature prediction model with the optimized basic dataset as input and the corresponding coal sample temperature as output. During the training process, the particle swarm optimization algorithm is used to iteratively optimize the hyperparameters of the coal temperature prediction model. The coal temperature prediction module is used to invert the real-time coal sample temperature in the goaf area of ​​the mine based on the real-time gas index of the coal spontaneous combustion gas products in the goaf area of ​​the mine using a trained coal temperature prediction model. The advanced early warning module is used to combine a pre-trained time series model with the coal sample temperature time series obtained by historical inversion to determine the evolution law of historical coal sample temperature, so as to predict the future coal sample temperature time series in the goaf of the mine and carry out advanced early warning of coal spontaneous combustion.

[0057] It is not difficult to see that this embodiment is a system embodiment corresponding to the above method embodiments, and this embodiment can be implemented in conjunction with the above method embodiments. The relevant technical details and technical effects mentioned in the above embodiments are still valid in this embodiment, and will not be repeated here to reduce repetition. Accordingly, the relevant technical details mentioned in this embodiment can also be applied to the above embodiments.

[0058] It is worth mentioning that all modules involved in this embodiment are logical modules. In practical applications, a logical unit can be a physical unit, a part of a physical unit, or a combination of multiple physical units. Furthermore, to highlight the innovative aspects of this invention, this embodiment does not introduce units that are not closely related to solving the technical problem proposed by this invention; however, this does not mean that other units are absent from this embodiment.

[0059] Another embodiment of the present invention relates to a computer device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the early warning method for spontaneous combustion of coal in goaf in the above embodiments.

[0060] The memory and processor are connected via a bus, which can include any number of interconnecting buses and bridges, connecting various circuits of one or more processors and memories. The bus can also connect various other circuits, such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and will not be described further herein. The bus interface provides an interface between the bus and the transceiver. The transceiver can be a single element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices over a transmission medium. Data processed by the processor is transmitted over the wireless medium via an antenna, which further receives data and transmits it to the processor.

[0061] The processor manages the bus and general processing, and also provides various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory is used to store data used by the processor during operation.

[0062] Another embodiment of the present invention relates to a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the method embodiments described above.

[0063] That is, those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. This program is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0064] Those skilled in the art will understand that the above embodiments are specific embodiments for implementing the present invention, and in practical applications, various changes can be made to them in form and detail without departing from the spirit and scope of the present invention.

Claims

1. A method for early warning of spontaneous combustion of coal in goaf areas, characterized in that, The method includes: To construct a basic dataset, we need to obtain the basic indicators of coal samples from different coal seams in the mine during the spontaneous combustion process, as well as the gas indicators of the spontaneous combustion gas products. The basic indicators are used to indicate the physicochemical properties and spontaneous combustion tendency of the coal samples, and the gas indicators include the concentration of multiple single gases and the concentration ratio of different gases in the composite gas. Using the basic dataset as input and the corresponding coal sample temperature as output, a pre-built coal temperature prediction model is trained, and the hyperparameters of the coal temperature prediction model are iteratively optimized using the particle swarm optimization algorithm during the training process. The real-time coal sample temperature in the goaf is inverted by using a trained coal temperature prediction model based on the real-time gas indicators of the coal spontaneous combustion gas products in the goaf. By combining a pre-trained time series model with the coal sample temperature time series obtained through historical inversion, the evolution law of historical coal sample temperature can be determined to predict the future coal sample temperature time series in the goaf of the mine, and to provide early warning of coal spontaneous combustion.

2. The method for early warning of spontaneous combustion of coal in goaf areas according to claim 1, characterized in that, Before training the pre-built coal temperature prediction model, the following is also included: The K-means clustering algorithm was used to perform unsupervised clustering analysis on the optimized basic dataset to determine the temperature range for graded early warning of coal spontaneous combustion fires and the corresponding indicator data for each level of early warning. Following the inversion of the real-time coal sample temperature in the goaf of the mine, the following is also included: Based on the temperature range for graded early warning of coal spontaneous combustion fires and the corresponding indicator data for each level of early warning, the early warning level and indicator data corresponding to the real-time coal sample temperature in the goaf of the mine are determined in order to conduct early warning of coal spontaneous combustion.

3. The method for early warning of spontaneous combustion of coal in goaf areas according to claim 2, characterized in that, The K-means clustering algorithm is used to perform unsupervised clustering analysis on the optimized basic dataset to determine the temperature range for graded early warning of coal spontaneous combustion fires and the corresponding indicator data for each level of warning, including: Based on the basic indicators of the coal sample, the critical temperature and drying cracking temperature of the coal sample during the spontaneous combustion process were determined; among them, the critical temperature and drying cracking temperature are the key node temperatures at which significant physicochemical changes occur in the coal sample during the spontaneous combustion process. Based on the critical temperature and cracking temperature of the coal sample, clustering index data are selected from the optimized basic dataset and used as the sample dataset. Using the Euclidean distance between indicator data in the sample dataset as a similarity measure, the K-means clustering algorithm is adopted. Based on the preset number of clusters, the algorithm aims to minimize the intra-cluster variance by iteratively allocating indicator data and updating cluster centers to obtain the clustering results of the sample dataset.

4. The method for early warning of spontaneous combustion of coal in goaf areas according to claim 1, characterized in that, The pre-built coal temperature prediction model includes the extreme gradient boosting model, the support vector regression model, and the random forest model; The process of using a trained coal temperature prediction model to invert the real-time coal sample temperature in the goaf based on real-time gas indicators of spontaneous combustion gas products from the coal, including: The k-fold cross-validation method was used to evaluate the trained extreme gradient boosting model, support vector regression model and random forest model respectively, so as to determine the target coal temperature prediction model based on the regression performance and classification performance of the extreme gradient boosting model, support vector regression model and random forest model respectively. The target coal temperature prediction model, after being trained, is used to invert the real-time coal sample temperature in the goaf based on the real-time gas indicators of the spontaneous combustion gas products of coal in the goaf.

5. The method for early warning of spontaneous combustion of coal in goaf areas according to claim 1, characterized in that, Before training the pre-built coal temperature prediction model using the basic dataset as input and the corresponding coal sample temperature as output, the following steps are also included: Pearson correlation coefficient analysis was used to determine the degree of linear correlation between each index data in the basic dataset and the coal sample temperature, in order to optimize the basic dataset.

6. The method for early warning of spontaneous combustion of coal in goaf areas according to claim 1, characterized in that, The basic indicators of the coal sample include the coal sample origin, coal seam where the coal sample is located, coal type, moisture, ash content, volatile matter, fixed carbon, coal sample temperature, furnace temperature, time, and temperature crossover point; The concentrations of the various gases and the concentration ratios of the different gases are the concentrations of O2, CO, CO2, and C2H4, as well as the C2H4 / CH4 ratio, CO2 / CO ratio, and CO / O2 ratio.

7. The method for early warning of spontaneous combustion of coal in goaf areas according to claim 3, characterized in that, The temperature range for classifying and issuing early warnings for spontaneous combustion of coal is as follows: coal sample temperature T < 30℃ indicates low risk; 0℃ ≤ T < 60℃ indicates general risk; 60℃ ≤ T < 100℃ indicates significant risk; 100℃ ≤ T < 150℃ indicates major risk; and T ≥ 150℃ indicates extremely high risk.

8. A pre-warning system for spontaneous combustion of coal in goaf areas, characterized in that, The system includes: The data acquisition module is used to acquire the basic indicators of coal samples from different coal seams in the mine during the coal spontaneous combustion process, as well as the gas indicators of the coal spontaneous combustion gas products, obtained through coal spontaneous combustion experiments, in order to construct a basic dataset. Among them, the basic indicators are used to indicate the physicochemical properties and spontaneous combustion tendency of the coal samples, and the gas indicators include the concentration of multiple single gases and the concentration ratio of different gases in the composite gas. The model training module is used to train the pre-built coal temperature prediction model with the optimized basic dataset as input and the corresponding coal sample temperature as output. During the training process, the particle swarm optimization algorithm is used to iteratively optimize the hyperparameters of the coal temperature prediction model. The coal temperature prediction module is used to invert the real-time coal sample temperature in the goaf area of ​​the mine based on the real-time gas index of the coal spontaneous combustion gas products in the goaf area of ​​the mine using a trained coal temperature prediction model. The advanced early warning module is used to combine a pre-trained time series model with the coal sample temperature time series obtained by historical inversion to determine the evolution law of historical coal sample temperature, so as to predict the future coal sample temperature time series in the goaf of the mine and carry out advanced early warning of coal spontaneous combustion.

9. A computer device, characterized in that, include: At least one processor; And a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the early warning method for spontaneous combustion of coal in goaf as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the method for early warning of spontaneous combustion of coal in goaf areas as described in any one of claims 1 to 7.

Images