Coal mine geological disaster and environment early warning method based on multi-source monitoring data fusion

Abstract

The invention discloses a coal mine geological disaster and environment early warning method based on multi-source monitoring data fusion. The method comprises the following steps: determining a mine water disaster early warning evaluation index, a mine pressure disaster early warning evaluation index and a geological environment evolution early warning index; a corresponding first early warning threshold value is set; data are acquired and preprocessed, and a mine water disaster early warning model, a mine pressure disaster early warning model and a geological environment evolution early warning model are established; collecting new data to carry out iterative optimization on the model, and correcting an early warning threshold value; and integrating the iteratively optimized models into an early warning index system library, collecting and fusing various monitoring system and production dynamic data of the mine, and inputting the data into the early warning index system library for coal mine geological disaster and environment early warning analysis. The method solves the problems that in the prior art, multi-source monitoring data cannot be fully fused, an early warning model lacks self-learning and optimization capabilities, and an early warning threshold cannot be dynamically adjusted along with the change of a mining environment.

Description

Technical Field

[0001] The invention belongs to the technical field of coal mine safety, and in particular relates to a coal mine geological disaster and environment early warning method based on multi-source monitoring data fusion. Background Art

[0002] Mine water damage is one of the common safety hazards in mining areas. The water source in underground mines may come from groundwater, surface water or old kiln water. Once a water damage occurs, it will not only endanger the lives of miners, but also cause serious damage to mine facilities. Therefore, by establishing an effective mine water damage early warning system, real-time monitoring and analysis of mine hydrological data, it is possible to dynamically analyze potential water disaster risks and provide advance warning for emergency disposal of mine water damage. As coal mine development moves deeper underground, the problem of mine pressure disasters becomes more prominent. The stress distribution of surrounding rocks will change with the dynamics of mining, which may lead to dangerous situations such as movement and collapse of rock formations. The mine pressure disaster early warning system can predict the probability of mine pressure disasters by monitoring data such as ground stress and microseismicity, and provide mine workers with sufficient time to take emergency measures. In addition, as an important part of mining area environmental management, geological environment evolution, underground mining activities will have a certain impact on the shallow ecology of mining areas, such as surface collapse, landform landscape destruction, soil erosion, etc. The geological environment evolution early warning system uses remote sensing monitoring, surface deformation monitoring and other means to timely discover the changing trends of the geological environment, providing a scientific basis for the ecological protection and restoration of mining areas.

[0003] In the environmental management of mining areas, dynamic analysis and early warning of mine water hazards, mine pressure disasters and geological environment evolution are key links to ensure safe, efficient and green mining of coal mines. These early warning systems can timely warn of potential environmental risks and their evolution trends, avoid or reduce the occurrence of environmental incidents, and achieve the goal of coordinated development of mining area environmental protection and resource development. However, the traditional mining area environmental monitoring and early warning system has the following technical problems: the data source is single, and it cannot make full use of a variety of monitoring data to improve the accuracy of early warning; the early warning model lacks the ability of self-learning and optimization, and cannot adapt to the dynamic changes of the mining area caused by mining, which is out of touch with the actual production; the early warning threshold is fixed and cannot be adjusted according to the actual situation, which is easy to cause false alarms or missed alarms. Summary of the invention

[0004] The purpose of the present invention is to provide a coal mine geological disaster and environmental early warning method based on multi-source monitoring data fusion, which solves the problems existing in the prior art that multi-source monitoring data cannot be fully integrated, the early warning model lacks self-learning and optimization capabilities, and the early warning threshold cannot be dynamically adjusted with changes in the mining environment.

[0005] The technical solution adopted by the present invention is a coal mine geological disaster and environmental early warning method based on multi-source monitoring data fusion, which specifically includes the following steps: S1: Determine the mine water hazard warning evaluation index, the mine pressure disaster warning evaluation index and the geological environment evolution warning index; set the first mine water hazard warning threshold, the first mine pressure disaster warning threshold and the first geological environment evolution warning threshold; S2: Acquire and preprocess data, and establish mine water hazard warning model, mine pressure disaster warning model and geological environment evolution warning model according to the preprocessed data; S3: collect new data, iteratively optimize the models in S2, and modify the first mine water hazard warning threshold, the first mine pressure disaster warning threshold, and the first geological environment evolution warning threshold to the second mine water hazard warning threshold, the second mine pressure disaster warning threshold, and the second geological environment evolution warning threshold; S4: Integrate the iteratively optimized mine water hazard warning model, mine pressure disaster warning model and geological environment evolution warning model into the warning indicator system library, collect and integrate various monitoring systems and production dynamic data of the mine, and input them into the warning indicator system library for coal mine geological disaster and environmental warning analysis.

[0006] The present invention is also characterized in that: The mine water hazard warning evaluation indicators include hydrological monitoring data and abnormal microseismic fluctuation frequency. The hydrological monitoring data include water level and water quality. The mine pressure disaster warning evaluation indicators include ground stress change rate, microseismic time frequency and rock movement speed. The geological environment evolution warning indicators include surface deformation rate and the total amount of dissolved solids in surface water.

[0007] S2 specifically includes: Collect hydrological monitoring data, microseismic abnormal fluctuation frequency data and historical time series data, and after preprocessing to remove outliers, fill missing data and smooth noise, combine the mine water hazard warning evaluation index and the first mine water hazard warning threshold, and build a mine water hazard warning model based on a neural network model or a support vector machine; Collect ground stress change rate data, microseismic time frequency data and rock movement velocity data, and after preprocessing by removing outliers, filling missing data and smoothing noise, combine the mine pressure disaster warning evaluation index and the first mine pressure disaster warning threshold, and build a mine pressure disaster warning model based on the genetic algorithm model and the adjustment optimization model; The surface deformation velocity data and the total amount of dissolved solids in surface water were collected, and after preprocessing to remove outliers, fill in missing data, and smooth noise, the geological environment evolution early warning indicators and the first geological environment evolution early warning threshold were combined to construct a geological environment evolution early warning model based on parameter mining model and cumulative impact analysis.

[0008] The mine water hazard early warning model includes a multi-layer perceptron MLP; the multi-layer perceptron MLP is expressed as:

[0009] In the formula, X Represents the water level and water quality indicators received by the input layer; the hidden layer uses the weight matrix W and the bias term b Perform nonlinear transformation; the output layer outputs a binary classification result Y , indicating safe and dangerous status; f Represents the activation function, which is a type of sigmoid or ReLU.

[0010] The mine water hazard early warning model includes a linear support vector machine SVM, which is expressed as:

[0011] In the formula, represents the Lagrange multiplier, represents the sample label, is the kernel function, b is the bias term; represents the received water level and water quality index of the input layer, n represents the total number of indicators, and the output layer outputs a binary classification result. The symbol function The outputs 1 and 0 represent safety and danger respectively.

[0012] The mine pressure disaster early warning model includes a genetic algorithm model and an adjustment optimization model; The genetic algorithm model encodes the ground stress monitoring data, microseismic monitoring data and rock movement monitoring data into genetic individuals; iterates and optimizes through population initialization, fitness evaluation, selection, crossover and mutation; The adjustment optimization model includes a linear weighted optimization formula, which is expressed as:

[0013] In the formula, R is the early warning risk value, is the weight coefficient of the jth monitoring system, is the eigenvalue of the jth monitoring system.

[0014] The geological environment evolution early warning model includes parameter mining model and cumulative impact index; Parameter mining model, expressed as:

[0015] In the formula, is the kth key characterization parameter, is the remote sensing data related to the kth parameter, is the surface water data related to the kth parameter; The cumulative impact index is expressed as:

[0016] In the formula, represents the cumulative impact index, represents the change of the kth parameter at time t, Represents the weight coefficient of the kth parameter.

[0017] S3 specifically includes the following steps: S3.1: Acquire new mine data, including hydrological monitoring data, abnormal microseismic fluctuation frequency, including water level and water quality; ground stress change rate, microseismic time frequency and rock movement speed; surface deformation speed and total dissolved solids in surface water; perform data cleaning on the new mine data, including removing outliers, filling missing data and smoothing noise; S3.2: Divide the new mine data into training sets and validation sets in proportion, use the training sets to train the constructed mine water hazard warning model, mine pressure disaster warning model, and geological environment evolution warning model, and use the validation set and cross-validation methods to evaluate the performance of the trained models; S3.3: Use the trained models to predict the new mine data to obtain the prediction results of each prediction model; according to the prediction results of each model and the actual situation of the mine, revise the first mine water hazard warning threshold, the first mine pressure disaster warning threshold and the first geological environment evolution warning threshold to the second mine water hazard warning threshold, the second mine pressure disaster warning threshold and the second geological environment evolution warning threshold; S3.4: Use the data in S2 to verify the revised second mine water hazard warning threshold, the second mine pressure disaster warning threshold and the second geological environment evolution warning threshold to ensure their effectiveness and accuracy.

[0018] In S3.2, the training set is used to train the constructed mine water hazard warning model, mine pressure disaster warning model, and geological environment evolution warning model. The training function used in the process is expressed as:

[0019]

[0020]

[0021] In the formula, Train Represents the training function, which executes machine learning algorithms, including gradient descent, reverse algorithm, and convex optimization algorithm, to adjust model parameters; For new mine data, , , For the pre-training mine water hazard warning model, mine pressure disaster warning model, and geological environment evolution warning model, , , These are the trained mine water hazard warning model, mine pressure disaster warning model, and geological environment evolution warning model.

[0022] The beneficial effects of the present invention are: In the coal mine geological disaster and environmental early warning method based on multi-source monitoring data fusion, the present invention provides clear evaluation standards for subsequent monitoring and early warning by determining the early warning evaluation indicators of mine water hazards, mine pressure disasters and geological environment evolution. Setting early warning thresholds helps to quantify and quantitatively evaluate potential environmental risks, which is convenient for managers to understand and make decisions. Clear early warning thresholds help to take preventive measures in time and respond to possible environmental risks in advance. By collecting monitoring data from multiple sources, such as hydrology, microseismicity, rock movement, and ground stress, a rich information basis is provided for the early warning model. The integration of multi-source data helps to fully reflect the current status of coal mining and avoid the one-sidedness caused by relying on single information; at the same time, the mine water hazard early warning model, mine pressure disaster early warning model and geological environment evolution early warning model constructed based on neural network model, support vector machine, genetic algorithm model, adjustment optimization model, parameter mining model and cumulative impact analysis can better adapt to the complex changes in the mining environment, and can more comprehensively evaluate the environmental conditions of the mining area, thereby improving the accuracy of the early warning; further, through model training and learning, the early warning model can be continuously iterated and optimized to improve the accuracy and reliability of the early warning. The correction of the early warning threshold enables the system to be dynamically adjusted according to the actual environmental conditions to reduce false alarms and missed alarms. As the mining in the mining area changes dynamically, the optimized model and threshold are more in line with the actual situation on site and can be better applied to production. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a flow chart of the coal mine geological disaster and environmental early warning method based on multi-source monitoring data fusion of the present invention. DETAILED DESCRIPTION

[0024] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0025] The present invention is based on a coal mine geological disaster and environmental early warning method based on multi-source monitoring data fusion, such as Figure 1 As shown, the specific steps include: S1: Determine the mine water hazard warning evaluation indicators, mine pressure disaster warning evaluation indicators and geological environment evolution warning indicators; set the first mine water hazard warning threshold, the first mine pressure disaster warning threshold and the first geological environment evolution warning threshold.

[0026] The mine water hazard warning evaluation indicators include hydrological monitoring data and abnormal microseismic fluctuation frequency. The hydrological monitoring data include water level and water quality; water quality includes parameters such as pH value and dissolved oxygen. The mine pressure disaster warning evaluation indicators include the ground stress change rate, microseismic time frequency and rock movement speed; the geological environment evolution warning indicators include the surface deformation speed and the total amount of dissolved solids in surface water (i.e. mineralization). And based on the empirical values, the first mine water hazard warning threshold, the first mine pressure disaster warning threshold and the first geological environment evolution warning threshold are set. For example, the first mine pressure disaster warning threshold is set to a ground stress change rate of 5MPa / month, a microseismic event frequency of 3 times / day or a rock movement speed of 2cm / month.

[0027] By determining the early warning evaluation indicators for mine water hazards, mine pressure disasters and geological environment evolution, clear evaluation standards are provided for subsequent monitoring and early warning. Setting early warning thresholds helps to quantify potential environmental risks into specific values, which is easier for managers to understand and make decisions. Clear early warning thresholds help mining areas take preventive measures and respond to possible environmental risks in advance.

[0028] S2: Acquire and preprocess data, and establish mine water hazard warning model, mine pressure disaster warning model and geological environment evolution warning model according to the preprocessed data; Collect hydrological monitoring data, microseismic abnormal fluctuation frequency data and historical time series data, and after preprocessing to remove outliers, fill missing data and smooth noise, combine the mine water hazard warning evaluation index and the first mine water hazard warning threshold, and build a mine water hazard warning model based on a neural network model or a support vector machine; The mine water hazard warning model includes a multi-layer perceptron MLP; the neural network can learn the complex relationship between input and output and make predictions for new input data. The multi-layer perceptron MLP is represented as:

[0029] In the formula, X Represents the water level and water quality indicators received by the input layer; the hidden layer uses the weight matrix W and the bias term b Perform nonlinear transformation; the output layer outputs a binary classification result Y , indicating safe and dangerous status; f Represents the activation function, which is a type of sigmoid or ReLU.

[0030] Assume that the activation function of the output layer is sigmoid, and its output value is y (indicating the probability of a dangerous state). The initial threshold can be set to 0.5, that is, when y>0.5, the mine is considered to be in a dangerous state.

[0031] The mine flood warning model can also be a linear support vector machine SVM. Support vector machine (SVM) is a supervised learning model used for data classification and regression analysis. In the mine flood warning model, SVM is used to distinguish between safe and dangerous states. The linear support vector machine SVM is expressed as:

[0032] In the formula, represents the Lagrange multiplier, represents the sample label, is a kernel function (common kernel functions include linear kernel, polynomial kernel, radial basis function (RBF) kernel, etc.), b is a bias term used to adjust the position of the decision boundary; represents the received water level and water quality index of the input layer, n represents the total number of indicators, and the output layer outputs a binary classification result. The symbol function The output 1 and 0 represent safety and danger respectively. SVM optimizes the Lagrange multiplier and the bias term b , find an optimal linear decision boundary to distinguish between safe and dangerous states. In the mine water hazard early warning model, this means that SVM is able to distinguish between dangerous conditions that may lead to water damage and safe conditions.

[0033] Collect ground stress change rate data, microseismic time frequency data and rock movement velocity data, and after preprocessing by removing outliers, filling missing data and smoothing noise, combine the mine pressure disaster warning evaluation index and the first mine pressure disaster warning threshold, and build a mine pressure disaster warning model based on the genetic algorithm model and the adjustment optimization model; The mine pressure disaster early warning model includes a genetic algorithm model and an adjustment optimization model; The genetic algorithm model encodes the ground stress monitoring data, microseismic monitoring data and rock movement monitoring data into genetic individuals; by initializing the population, a group of individuals containing different monitoring data combinations are randomly generated; the fitness function is defined, taking into account factors such as the accuracy, timeliness and false alarm rate of the early warning; through selection, crossover and mutation operations, the population is iteratively optimized until the individual with the best performance is found. It can be specifically expressed as follows: Encoding genetic individuals: Encode the ground stress monitoring data, microseismic monitoring data and rock movement monitoring data into genetic individuals. Each individual can be represented as a vector, and each element of the vector corresponds to a value of monitoring data.

[0034] Initialize the population: Randomly generate a group of individuals containing different monitoring data combinations as the initial population. The population size can be set according to actual conditions, such as 100 individuals.

[0035] Define fitness function: The fitness function is used to evaluate the quality of individuals. In the mine pressure disaster warning model, the fitness function can take into account factors such as the accuracy, timeliness and false alarm rate of the warning. For example, a fitness function based on historical data can be defined, which scores the degree of conformity between the warning results calculated by the combination of individual monitoring data and the actual disaster situation.

[0036] On the basis of genetic algorithm, the adjustment optimization model is further used to refine the early warning plan. The adjustment optimization model includes a linear weighted optimization formula, which is expressed as:

[0037] In the formula, R is the early warning risk value, which integrates the data of various monitoring systems to assess the risk of mine pressure disasters; It is j The weight coefficient of a monitoring system reflects the importance of the monitoring system in early warning; It is the characteristic value of the jth monitoring system, representing the real-time monitoring data or processed key indicators of the system.

[0038] The surface deformation velocity data and the total amount of dissolved solids in surface water were collected, and after preprocessing to remove outliers, fill in missing data, and smooth noise, the geological environment evolution early warning indicators and the first geological environment evolution early warning threshold were combined to construct a geological environment evolution early warning model based on parameter mining model and cumulative impact analysis.

[0039] The geological environment evolution early warning model includes parameter mining model and cumulative impact index; The parameter mining model uses the collected surface deformation velocity data and the total dissolved solids data of surface water to identify the environmental parameters closely related to mining activities. The model is expressed as:

[0040] In the formula, It is k Key characterization parameters, which are important indicators for assessing geological environment changes; It is k Remote sensing data related to parameters, It is k Surface water data related to the parameters; In order to comprehensively evaluate the long-term impact of mining activities on the geological environment, it is necessary to construct a cumulative impact index, which is expressed as:

[0041] In the formula, represents the cumulative impact index, Indicates k Parameters in timet The amount of change, Indicates k The above cumulative impact index comprehensively considers the changes of various parameters in different time periods, as well as the importance of each parameter in the overall assessment (reflected by the weight coefficient).

[0042] S3: Collect new data, iteratively optimize the models in S2, and revise the first mine water hazard warning threshold, the first mine pressure disaster warning threshold, and the first geological environment evolution warning threshold to the second mine water hazard warning threshold, the second mine pressure disaster warning threshold, and the second geological environment evolution warning threshold.

[0043] The specific steps include: S3.1: Collect new data from various sensors, monitoring systems and field surveys in the mine. The new mine data include hydrological monitoring data, abnormal microseismic fluctuation frequency, hydrological monitoring data including water level and water quality; ground stress change rate, microseismic time frequency and rock movement speed; surface deformation speed and total dissolved solids in surface water; perform data cleaning on the new mine data, including removing outliers, filling missing data and smoothing noise to ensure data accuracy and reliability.

[0044] S3.2: Divide the new mine data into training sets and validation sets in proportion, use the training set to train the constructed mine water hazard warning model, mine pressure disaster warning model, and geological environment evolution warning model, and use the validation set and cross-validation method to evaluate the performance of the trained model.

[0045] The training set is used to train the constructed mine water hazard warning model, mine pressure disaster warning model, and geological environment evolution warning model. The training function used in the process is expressed as:

[0046]

[0047]

[0048] In the formula, Train Represents the training function, which executes machine learning algorithms, including gradient descent, reverse algorithm, and convex optimization algorithm, to adjust model parameters; For new mine data, , , For the pre-training mine water hazard warning model, mine pressure disaster warning model, and geological environment evolution warning model, , , These are the trained mine water hazard warning model, mine pressure disaster warning model, and geological environment evolution warning model.

[0049] For the mine water hazard warning model, if the multilayer perceptron (MLP) is used as the neural network model, the training set data is used to adjust the weight matrix W and the bias term b through the back propagation algorithm to minimize the loss function. If the linear support vector machine (SVM) is used as the classification model, the training set data is used to solve the optimization problem of the SVM through the convex optimization algorithm.

[0050] For the mine pressure disaster warning model, the specific training and iterative optimization process is as follows: Selection operation: According to the value of the fitness function, select individuals with better performance as parents to generate the next generation of individuals; you can also use roulette selection, tournament selection and other methods for selection.

[0051] Crossover operation: Perform a crossover operation on the selected parent individuals to generate new individuals. The crossover operation can be performed by single-point crossover, double-point crossover or uniform crossover.

[0052] Mutation operation: Perform mutation operation on newly generated individuals to increase the diversity of the population. The mutation operation can be achieved by randomly changing the values ​​of certain elements in the individuals.

[0053] Iterative optimization: Repeat the selection, crossover, and mutation operations until the set number of iterations is reached or other stopping conditions are met. In each iteration, a new population is generated and the performance of the individuals is evaluated by the fitness function. As the iterations proceed, the performance of the population will gradually improve, and eventually the individuals with the best performance will be obtained.

[0054] This can be illustrated by the following example: Assume the following historical data: Ground stress monitoring data (MPa): [2.0, 2.5, 3.0, ..., 4.8, 5.0]; Microseismic event frequency (times / day): [1, 2, 1, ..., 3, 4]; Rock migration speed (cm / month): [0.5, 1.0, 1.2, ..., 1.8, 2.0]; Corresponding disaster situation: [0, 0, 1, ..., 1, 1], where 0 indicates no disaster and 1 indicates a disaster.

[0055] When constructing a genetic algorithm model, the above data is encoded as genetic individuals and a population containing multiple individuals is initialized. By defining a fitness function (such as a function based on disaster prediction accuracy), the individuals in the population are evaluated and selected.

[0056] Define the fitness function as , where F(x) is the fitness value of individual x. A(x) is the number of disasters predicted correctly by individual x. B(x) is the number of false alarms (i.e., incorrectly predicting that a disaster will occur, but it did not actually occur) by individual x. C(x) is the number of omissions (i.e., failing to predict a disaster that actually occurred) by individual x. α、β and γ is a weight coefficient used to adjust the relative importance of prediction accuracy, false positives, and false negatives in the fitness function. These coefficients can be set according to actual conditions. For example, prediction accuracy may be considered more important than false positives and false negatives, so α Assign a larger value.

[0057] To make this more concrete, assume the following data: Individual x predicted 10 disasters, of which 7 were correct, 2 were false alarms, and 1 was missed.

[0058] Set the weight coefficient to α =1, β =0.5, γ = 1 (meaning the penalty for a false positive is half the reward for a correct prediction, while the penalty for a false negative is the same as the reward for a correct prediction).

[0059] According to the above fitness function, the fitness value of individual x can be calculated as: F(x)=1*7-0.5*2-1*1=7-1-1=5; This fitness value represents the comprehensive performance of individual x in predicting mine pressure disasters. The higher the fitness value, the better the performance of the individual. In the iterative process of the genetic algorithm, we tend to select individuals with higher fitness values ​​for crossover and mutation operations to generate a new generation of individuals with better performance.

[0060] The following results are obtained during the iterative optimization process: individuals in the initial population may have a lower fitness value, indicating that their early warning accuracy is low. As the iteration proceeds, individuals in the population are gradually optimized and their fitness values ​​gradually increase. The optimal individual may have the following monitoring data combination: ground stress change rate = 4.5MPa, microseismic event frequency = 2.5 times / day, rock movement speed = 1.8cm / month.

[0061] For the geological environment evolution early warning model, the model is trained and learned through genetic algorithms.

[0062] S3.3: Use the trained models to predict the new mine data and obtain the prediction results of each prediction model; according to the prediction results of each model and the actual situation of the mine, revise the first mine water hazard warning threshold, the first mine pressure disaster warning threshold and the first geological environment evolution warning threshold to the second mine water hazard warning threshold, the second mine pressure disaster warning threshold and the second geological environment evolution warning threshold.

[0063] For the mine water hazard warning model, after the model is trained, the warning threshold is adjusted according to the prediction. For example, if it is found that the model has a large number of missed reports during low-level water hazard events, you can try to lower the threshold; if it is found that the model has a large number of false alarms during high-level normal conditions, you can try to increase the threshold. By continuously adjusting the threshold until a balance point is found, cross-validation, grid search and other techniques can be used to adjust the threshold more finely. After multiple iterations of optimization, we determined a new warning threshold, the second mine water hazard warning threshold. This new threshold should be able to better balance the sensitivity and accuracy of the model and reduce false alarms and missed reports.

[0064] For the mine pressure disaster warning model, after the model is trained, the warning threshold is adjusted according to the prediction. Corresponding to the example of S3.2, the following results are obtained during the iterative optimization process: the individuals in the initial population may have a lower fitness value, indicating that their warning accuracy is low. As the iteration proceeds, the individuals in the population are gradually optimized and the fitness value gradually increases. The optimal individual may have the following monitoring data combination: ground stress change rate = 4.5MPa, microseismic event frequency = 2.5 times / day, rock movement speed = 1.8cm / month.

[0065] For the geological environment evolution early warning model, after the model is trained, the warning threshold is adjusted according to the prediction.

[0066] S3.4: Use the data in S2 to verify the revised second mine water hazard warning threshold, the second mine pressure disaster warning threshold and the second geological environment evolution warning threshold to ensure their effectiveness and accuracy.

[0067] S4: Integrate the iteratively optimized mine water hazard warning model, mine pressure disaster warning model and geological environment evolution warning model into the warning indicator system library, collect and integrate various monitoring systems and production dynamic data of the mine, and input them into the warning indicator system library for coal mine geological disaster and environmental warning analysis.

[0068] After automatically collecting and integrating various monitoring systems and production dynamic data of the mine, the data are input into the early warning indicator system library to conduct coal mine geological disaster and environmental early warning analysis.

[0069] For the mine water hazard warning model, water level and water quality indicators are input, and after processing, they are compared with the second mine water hazard warning threshold to output a binary classification result, indicating a safe or dangerous state; for the neural network threshold is a limit of the probability value given by the output layer (for example, when the probability is greater than 0.5, it is considered to be a dangerous state); for the support vector machine, the threshold is a limit of the decision function value (for example, when the decision function value is greater than 0, it is considered to be a dangerous state).

[0070] For the mine pressure disaster warning model, the ground stress change rate, microseismic event frequency and rock movement speed indicators are input, and compared with the second mine pressure disaster warning threshold after processing. If the indicator is greater than the threshold, a warning signal is issued. For example, after model training, the second mine pressure disaster warning threshold is a ground stress change rate of 3.8MPa / month, a microseismic event frequency of 2 times / day, and a rock movement speed of 1.5cm / month. The corresponding weight coefficients are 0.5, 0.3, and 0.2, respectively. The warning risk value R=0.5×3.8+0.3×2+0.2×1.5=2.6, and the system sets a threshold (such as 2.5 is the boundary between safety and danger) to determine whether the period is in a dangerous state. In this example, since the warning risk value 2.6 is greater than the threshold 2.5, a warning signal is issued.

[0071] For the geological environment evolution early warning model, the surface deformation rate and the total amount of dissolved solids in surface water are input, and after processing, they are compared with the second geological environment evolution early warning threshold. If one of the indicators is greater than the threshold, an early warning signal is issued. For example, after model training, the second geological environment evolution early warning threshold is the surface deformation rate V = 4cm / year, the total amount of dissolved solids in surface water U = 250mg / L, and the input is the surface deformation rate V = 3.8cm / year, the total amount of dissolved solids in surface water U = 260mg / L, because the total amount of dissolved solids in surface water is greater than the threshold, so an early warning signal needs to be issued. If the surface deformation rate or the total amount of dissolved solids in surface water has continued to rise in the past few periods and is close to the new early warning threshold, then we may need to take preventive measures in advance to ensure the safety and stability of the mining area.

[0072] Example 1 Coal mine geological disaster and environmental early warning method based on multi-source monitoring data fusion, such as Figure 1 As shown, the specific steps include: S1: Determine the mine water hazard warning evaluation index, the mine pressure disaster warning evaluation index and the geological environment evolution warning index; set the first mine water hazard warning threshold, the first mine pressure disaster warning threshold and the first geological environment evolution warning threshold; S2: Acquire and preprocess data, and establish mine water hazard warning model, mine pressure disaster warning model and geological environment evolution warning model according to the preprocessed data; S3: collect new data, iteratively optimize the models in S2, and modify the first mine water hazard warning threshold, the first mine pressure disaster warning threshold, and the first geological environment evolution warning threshold to the second mine water hazard warning threshold, the second mine pressure disaster warning threshold, and the second geological environment evolution warning threshold; S4: Integrate the iteratively optimized mine water hazard warning model, mine pressure disaster warning model and geological environment evolution warning model into the warning indicator system library, collect and integrate various monitoring systems and production dynamic data of the mine, and input them into the warning indicator system library for coal mine geological disaster and environmental warning analysis.

[0073] Example 2 Based on Example 1, mine water hazard warning evaluation indicators include hydrological monitoring data and abnormal microseismic fluctuation frequency. The hydrological monitoring data include water level and water quality; mine pressure disaster warning evaluation indicators include ground stress change rate, microseismic time frequency and rock movement speed; geological environment evolution warning indicators include surface deformation speed and the total amount of dissolved solids in surface water.

[0074] Example 3 On the basis of Example 2, S2 specifically includes: Collect hydrological monitoring data, microseismic abnormal fluctuation frequency data and historical time series data, and after preprocessing to remove outliers, fill missing data and smooth noise, combine the mine water hazard warning evaluation index and the first mine water hazard warning threshold, and build a mine water hazard warning model based on a neural network model or a support vector machine; Collect ground stress change rate data, microseismic time frequency data and rock movement velocity data, and after preprocessing by removing outliers, filling missing data and smoothing noise, combine the mine pressure disaster warning evaluation index and the first mine pressure disaster warning threshold, and build a mine pressure disaster warning model based on the genetic algorithm model and the adjustment optimization model; The surface deformation velocity data and the total amount of dissolved solids in surface water were collected, and after preprocessing to remove outliers, fill in missing data, and smooth noise, the geological environment evolution early warning indicators and the first geological environment evolution early warning threshold were combined to construct a geological environment evolution early warning model based on parameter mining model and cumulative impact analysis.

[0075] Example 4 Based on Example 3, the mine water hazard early warning model includes a multi-layer perceptron MLP; the multi-layer perceptron MLP is expressed as:

[0076] In the formula, X Represents the water level and water quality indicators received by the input layer; the hidden layer uses the weight matrix W and the bias term b Perform nonlinear transformation; the output layer outputs a binary classification resultY , indicating safe and dangerous status; f Represents the activation function, which is a type of sigmoid or ReLU.

[0077] The mine water hazard warning model can also be a linear support vector machine SVM, which is expressed as:

[0078] In the formula, represents the Lagrange multiplier, represents the sample label, is the kernel function, b is the bias term; represents the received water level and water quality index of the input layer, n represents the total number of indicators, and the output layer outputs a binary classification result. The symbol function The outputs 1 and 0 represent safety and danger respectively.

[0079] The mine pressure disaster early warning model includes a genetic algorithm model and an adjustment optimization model; The genetic algorithm model encodes the ground stress monitoring data, microseismic monitoring data and rock movement monitoring data into genetic individuals; iterates and optimizes through population initialization, fitness evaluation, selection, crossover and mutation; The adjustment optimization model includes a linear weighted optimization formula, which is expressed as:

[0080] In the formula, R is the early warning risk value, is the weight coefficient of the jth monitoring system, is the eigenvalue of the jth monitoring system.

[0081] The geological environment evolution early warning model includes parameter mining model and cumulative impact index; Parameter mining model, expressed as:

[0082] In the formula, is the kth key characterization parameter, is the remote sensing data related to the kth parameter, is the surface water data related to the kth parameter; The cumulative impact index is expressed as:

[0083] In the formula, represents the cumulative impact index, represents the change of the kth parameter at time t, Represents the weight coefficient of the kth parameter.

[0084] Example 5 On the basis of Example 4, S3 specifically includes the following steps: S3.1: Acquire new mine data, including hydrological monitoring data, abnormal microseismic fluctuation frequency, including water level and water quality; ground stress change rate, microseismic time frequency and rock movement speed; surface deformation speed and total dissolved solids in surface water; perform data cleaning on the new mine data, including removing outliers, filling missing data and smoothing noise; S3.2: Divide the new mine data into a training set and a validation set in a ratio of 9:1. Use the training set to train the constructed mine water hazard warning model, mine pressure disaster warning model, and geological environment evolution warning model. Use the validation set and cross-validation method to evaluate the performance of the trained model. S3.3: Use the trained models to predict the new mine data to obtain the prediction results of each prediction model; according to the prediction results of each model and the actual situation of the mine, revise the first mine water hazard warning threshold, the first mine pressure disaster warning threshold and the first geological environment evolution warning threshold to the second mine water hazard warning threshold, the second mine pressure disaster warning threshold and the second geological environment evolution warning threshold; S3.4: Use the data in S2 to verify the revised second mine water hazard warning threshold, the second mine pressure disaster warning threshold and the second geological environment evolution warning threshold to ensure their effectiveness and accuracy.

[0085] Example 6 On the basis of Example 5, the training set is used in S3.2 to train the constructed mine water hazard warning model, mine pressure disaster warning model, and geological environment evolution warning model. The training function used in the process is expressed as:

[0086]

[0087]

[0088] In the formula, Train Represents the training function, which executes machine learning algorithms, including gradient descent, reverse algorithm, and convex optimization algorithm, to adjust model parameters; For new mine data, , , For the pre-training mine water hazard warning model, mine pressure disaster warning model, and geological environment evolution warning model, , , These are the trained mine water hazard warning model, mine pressure disaster warning model, and geological environment evolution warning model.

Claims

1. A coal mine geological disaster and environmental early warning method based on multi-source monitoring data fusion, characterized in that: The specific steps include: S1: Determine the mine water hazard warning evaluation index, the mine pressure disaster warning evaluation index and the geological environment evolution warning index; set the first mine water hazard warning threshold, the first mine pressure disaster warning threshold and the first geological environment evolution warning threshold; S2: Acquire and preprocess data, and establish mine water hazard warning model, mine pressure disaster warning model and geological environment evolution warning model according to the preprocessed data; S3: collect new data, iteratively optimize the models in S2, and modify the first mine water hazard warning threshold, the first mine pressure disaster warning threshold, and the first geological environment evolution warning threshold to the second mine water hazard warning threshold, the second mine pressure disaster warning threshold, and the second geological environment evolution warning threshold; S4: Integrate the iteratively optimized mine water hazard warning model, mine pressure disaster warning model and geological environment evolution warning model into the warning indicator system library, collect and integrate various monitoring systems and production dynamic data of the mine, and input them into the warning indicator system library for coal mine geological disaster and environmental warning analysis.

2. The coal mine geological disaster and environmental early warning method based on multi-source monitoring data fusion according to claim 1 is characterized in that: The mine water hazard warning evaluation indicators include hydrological monitoring data and abnormal microseismic fluctuation frequency. The hydrological monitoring data include water level and water quality; the mine pressure disaster warning evaluation indicators include ground stress change rate, microseismic time frequency and rock movement speed; the geological environment evolution warning indicators include surface deformation speed and the total amount of dissolved solids in surface water.

3. The coal mine geological disaster and environmental early warning method based on multi-source monitoring data fusion according to claim 1 is characterized in that: S2 specifically includes: Collect hydrological monitoring data, microseismic abnormal fluctuation frequency data and historical time series data, and after preprocessing to remove outliers, fill missing data and smooth noise, combine the mine water hazard warning evaluation index and the first mine water hazard warning threshold, and build a mine water hazard warning model based on a neural network model or a support vector machine; Collect ground stress change rate data, microseismic time frequency data and rock movement velocity data, and after preprocessing by removing outliers, filling missing data and smoothing noise, combine the mine pressure disaster warning evaluation index and the first mine pressure disaster warning threshold, and build a mine pressure disaster warning model based on the genetic algorithm model and the adjustment optimization model; The surface deformation velocity data and the total amount of dissolved solids in surface water were collected, and after preprocessing to remove outliers, fill in missing data, and smooth noise, the geological environment evolution early warning indicators and the first geological environment evolution early warning threshold were combined to construct a geological environment evolution early warning model based on parameter mining model and cumulative impact analysis.

4. The coal mine geological disaster and environmental early warning method based on multi-source monitoring data fusion according to claim 3 is characterized in that: The mine water hazard early warning model includes a multi-layer perceptron MLP; the multi-layer perceptron MLP is expressed as: In the formula, X Represents the water level and water quality indicators received by the input layer; the hidden layer uses the weight matrix W and the bias term b Perform nonlinear transformations; The output layer outputs a binary classification result Y , indicating safe and dangerous status; f Represents the activation function, which is a type of sigmoid or ReLU.

5. The coal mine geological disaster and environmental early warning method based on multi-source monitoring data fusion according to claim 3 is characterized in that: The mine water hazard early warning model includes a linear support vector machine SVM, and the linear support vector machine SVM is expressed as: In the formula, represents the Lagrange multiplier, represents the sample label, is the kernel function, b is the bias term; represents the received water level and water quality index of the input layer, n represents the total number of indicators, and the output layer outputs a binary classification result. The symbol function The outputs 1 and 0 represent safety and danger respectively.

6. The coal mine geological disaster and environmental early warning method based on multi-source monitoring data fusion according to claim 3 is characterized in that: The mine pressure disaster early warning model includes a genetic algorithm model and an adjustment optimization model; The genetic algorithm model encodes the ground stress monitoring data, microseismic monitoring data and rock movement monitoring data into genetic individuals; iterates and optimizes through population initialization, fitness evaluation, selection, crossover and mutation; The adjustment optimization model includes a linear weighted optimization formula, which is expressed as: In the formula, R is the early warning risk value, is the weight coefficient of the jth monitoring system, is the eigenvalue of the jth monitoring system.

7. The coal mine geological disaster and environmental early warning method based on multi-source monitoring data fusion according to claim 3 is characterized in that: The geological environment evolution early warning model includes a parameter mining model and a cumulative impact index; The parameter mining model is expressed as: In the formula, is the kth key characterization parameter, is the remote sensing data related to the kth parameter, is the surface water data related to the kth parameter; The cumulative impact index is expressed as: In the formula, represents the cumulative impact index, represents the change of the kth parameter at time t, Represents the weight coefficient of the kth parameter.

8. The coal mine geological disaster and environmental early warning method based on multi-source monitoring data fusion according to claim 1 is characterized in that: S3 specifically includes the following steps: S3.1: Acquire new mine data, including hydrological monitoring data, abnormal microseismic fluctuation frequency, and hydrological monitoring data including water level and water quality; ground stress change rate, microseismic time frequency, and rock movement speed; surface deformation speed and total dissolved solids in surface water; perform data cleaning on the new mine data, including removing outliers, filling missing data, and smoothing noise; S3.2: Divide the new mine data into a training set and a validation set in proportion, use the training set to train the constructed mine water hazard warning model, mine pressure disaster warning model, and geological environment evolution warning model, and use the validation set and cross-validation method to evaluate the performance of the trained model; S3.3: Use the trained models to predict the new mine data to obtain the prediction results of each prediction model; according to the prediction results of each model and the actual situation of the mine, revise the first mine water hazard warning threshold, the first mine pressure disaster warning threshold and the first geological environment evolution warning threshold to the second mine water hazard warning threshold, the second mine pressure disaster warning threshold and the second geological environment evolution warning threshold; S3.4: Use the data in S2 to verify the revised second mine water hazard warning threshold, the second mine pressure disaster warning threshold and the second geological environment evolution warning threshold to ensure their effectiveness and accuracy.

9. The method for coal mine geological disaster and environmental early warning based on multi-source monitoring data fusion according to claim 8, characterized in that: In S3.2, the training set is used to train the constructed mine water hazard warning model, mine pressure disaster warning model, and geological environment evolution warning model. The training function used in the process is expressed as: In the formula, Train Represents the training function, which executes machine learning algorithms, including gradient descent, reverse algorithm, and convex optimization algorithm, to adjust model parameters; For new mine data, , , For the pre-training mine water hazard warning model, mine pressure disaster warning model, and geological environment evolution warning model, , , These are the trained mine water hazard warning model, mine pressure disaster warning model, and geological environment evolution warning model.

Images