Coal rock dynamic disaster early warning method and system based on Bayesian dynamic extreme value modeling

Abstract

The invention discloses a coal rock dynamic disaster early warning method and system based on Bayesian dynamic extreme value modeling, and relates to the technical field of coal rock dynamic disaster monitoring. The method comprises the following steps: carrying out segmentation and feature extraction on multi-source sensor signal data of a coal mining working face based on a preset time window; constructing a dynamic extreme event index and a dynamic covariable vector; the dynamic covariable vectors are combined and updated based on a multi-frequency sampling asynchronous fusion mechanism, and a dynamic covariable matrix is generated in combination with dynamic extreme event indexes; constructing a Bayesian dynamic extreme value model based on the dynamic covariable matrix; updating the Bayesian dynamic extremum model based on a Markov chain Monte Carlo method, and calculating a dynamic risk index of the coal mining working face based on the Bayesian dynamic extremum model; and performing disaster early warning on the coal mining working face based on the dynamic risk index. The technical problem that in the prior art, the capability of capturing the disaster instantaneous sudden change risk is insufficient is solved.

Description

Technical Field

[0001] This invention relates to the field of coal and rock dynamic disaster monitoring technology, and in particular to a method and system for early warning of coal and rock dynamic disasters based on Bayesian dynamic extreme value modeling. Background Technology

[0002] In deep coal mining, the coupling of multiple factors such as high ground stress, gas occurrence, and mining disturbance can easily induce coal and rock dynamic disasters such as rockbursts and coal and gas outbursts, seriously threatening mine safety. Existing monitoring and early warning systems generally rely on the deployment of multi-source sensors, including acoustic emission, electromagnetic radiation, and gas emission sensors, to assess disaster risk by monitoring precursory information such as coal and rock mass fracturing and gas migration. Coal and rock dynamic disasters are characterized by typical instantaneous changes, and multi-source sensor data suffers from problems such as inconsistent sampling frequencies, different dimensions, and asynchronous information. Effectively integrating and quantifying their comprehensive impact on extreme risks is a key challenge. Traditional early warning methods are mostly based on static thresholds or empirical models, making it difficult to effectively distinguish between normal fluctuations in the production process and true precursory extreme events of disasters, resulting in high false alarm and false negative rates. Furthermore, once deployed, models based on historical data have fixed parameters and lack the ability to self-update and adjust using real-time data streams, making it difficult to adapt to dynamic changes in mining face conditions. Summary of the Invention

[0003] To address the aforementioned technical problems in existing technologies, this invention provides a method and system for early warning of coal and rock dynamic disasters based on Bayesian dynamic extremum modeling. The technical solution is as follows:

[0004] On the one hand, a method for early warning of coal and rock dynamic disasters based on Bayesian dynamic extremum modeling is provided. The method includes: segmenting and extracting features from multi-source sensor signal data of coal mine working faces based on a preset time window to construct dynamic extreme event indicators and dynamic covariate vectors; combining and updating the dynamic covariate vectors based on a multi-frequency sampling asynchronous fusion mechanism, and generating a dynamic covariate matrix in combination with the dynamic extreme event indicators; constructing a Bayesian dynamic extremum model based on the dynamic covariate matrix; the Bayesian dynamic extremum model includes a cumulative probability distribution model of the dynamic extreme event indicators with respect to time-varying parameters and a time-varying parameter state equation of the time-varying parameters with respect to the dynamic covariate vectors; updating the Bayesian dynamic extremum model based on the Markov chain Monte Carlo method, and calculating the dynamic risk index of the coal mine working face based on the Bayesian dynamic extremum model; and providing disaster early warning for the coal mine working face based on the dynamic risk index.

[0005] Optionally, the method further includes: synchronously acquiring multi-source sensor signal data in the coal and rock mining face; the multi-source sensor signal data includes acoustic emission sensor time series data, electromagnetic radiation sensor time series data, gas concentration sensor time series data, and surrounding rock stress time series data.

[0006] Optionally, the dynamic extreme event index includes the maximum value of absolute gas emission within the preset time window; the dynamic covariate vector includes the cumulative energy value of acoustic emission signal, the average intensity value of electromagnetic radiation signal, and the maximum rate of change of multiple sensor signals within the preset time window.

[0007] Optionally, the dynamic covariate vector is combined and updated based on the multi-frequency sampling asynchronous fusion mechanism, and a dynamic covariate matrix is ​​generated by combining the dynamic extreme event indicators. This includes: after completely storing the dynamic covariate vector in a preset data cache, reading the latest value of the dynamic covariate vector from the preset data cache, and reading the dynamic extreme event indicators corresponding to the same time window, and combining them into a dynamic covariate matrix.

[0008] Optionally, based on the dynamic covariate matrix, a Bayesian dynamic extreme value model is constructed, including: modeling the dynamic extreme event index using a generalized extreme value distribution to obtain a cumulative probability distribution model of the dynamic extreme event index with respect to time-varying parameters; the time-varying parameters include time-varying location parameters, time-varying scale parameters, and shape parameters; the mathematical form of the cumulative probability distribution model includes: In the formula, G() represents the cumulative probability distribution function, and Z() represents the cumulative probability distribution function. t This refers to the dynamic extreme event index. The time-varying position parameter, The time-varying scale parameter is... Let t represent time, and let t be the shape parameter. Based on the dynamic covariate vector, a time-varying parameter state equation is constructed. The mathematical form of the time-varying parameter state equation includes: In the formula, , … , Represents the dynamic covariate vector Different components in , … , , … These are time-varying regression coefficients; the evolution of these time-varying regression coefficients follows a random walk process: , ,in , , , To control the variance hyperparameter of the evolution step size.

[0009] Optionally, the Bayesian dynamic extreme value model is updated based on the Markov chain Monte Carlo method, including: setting a normal prior distribution for the time-varying regression coefficients. , A uniform prior distribution is set for the shape parameters. An inverse gamma distribution is set as the prior distribution for the variance hyperparameter of the control evolution step size: , When there is data to be updated When a model update is triggered, samples are drawn from the posterior probability distributions of each parameter in the dynamic covariate matrix based on the Markov chain Monte Carlo method; the mean of the posterior distribution is calculated from the converged Markov chain Monte Carlo chain as the optimal estimate of the parameter at the current time.

[0010] Optionally, the dynamic risk index includes: the probability that the dynamic extreme event indicator exceeds a preset safety threshold within the current time window; the calculation formula for the dynamic risk index includes: In the formula, RC t G is the dynamic risk index. critcal The preset security threshold is defined as follows.

[0011] Optionally, based on the dynamic risk index, disaster early warning is issued for the coal mine working face, including: when the dynamic risk index is less than a first preset threshold, it is determined to be a first early warning level, and the coal mine working face is determined to be in a safe state; when the dynamic risk index is greater than or equal to the first preset threshold and less than a second preset threshold, it is determined to be a second early warning level, and warning and suggestion information is issued; when the dynamic risk index is greater than or equal to the second preset threshold, it is determined to be a third early warning level, and disaster emergency operations are performed.

[0012] Optionally, the method further includes: using the mining engineering plan as the base map, overlaying and displaying the real-time dynamic risk indicators of the coal mine working faces in each area in the form of a heat map; establishing an early warning decision database to store the input data, model parameter estimates, calculated dynamic risk indicators, and output early warning instructions corresponding to each early warning event; periodically performing retrospective analysis on the early warning decision log, and recalibrating the hyperparameters of the Bayesian dynamic extreme value model using historical data.

[0013] On the other hand, a coal and rock dynamic disaster early warning system based on Bayesian dynamic extreme value modeling is also provided to implement the coal and rock dynamic disaster early warning method based on Bayesian dynamic extreme value modeling provided in the embodiments of the present invention. The system includes: an extraction module, a generation module, a construction module, an update module, and an early warning module. The extraction module is used to segment and extract features from multi-source sensor signal data of the coal mine working face based on a preset time window, and construct dynamic extreme event indicators and dynamic covariate vectors. The generation module is used to combine and update the dynamic covariate vectors based on a multi-frequency sampling asynchronous fusion mechanism, and combine the dynamic extreme event indicators... The system generates a dynamic covariate matrix; the construction module is used to construct a Bayesian dynamic extreme value model based on the dynamic covariate matrix; the Bayesian dynamic extreme value model includes a cumulative probability distribution model of the dynamic extreme event index with respect to time-varying parameters and a time-varying parameter state equation of the time-varying parameters with respect to the dynamic covariate vector; the update module is used to update the Bayesian dynamic extreme value model based on the Markov chain Monte Carlo method, and calculate the dynamic risk index of the coal mine working face based on the Bayesian dynamic extreme value model; the early warning module is used to provide disaster early warning for the coal mine working face based on the dynamic risk index.

[0014] This invention provides a method and system for early warning of coal and rock dynamic disasters based on Bayesian dynamic extremum modeling. By establishing a Bayesian extremum model driven by dynamic covariates from multiple sources of sensors, it solves the key problem that traditional early warning methods cannot capture the risk of instantaneous changes in coal and rock disasters. It realizes the probabilistic dynamic assessment of extreme events that are precursors to disasters, and provides an interpretable and adaptive new method for real-time intelligent prevention and control of coal mine dynamic disasters. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart of a coal and rock dynamic disaster early warning method based on Bayesian dynamic extremum modeling provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of a coal and rock dynamic disaster early warning system based on Bayesian dynamic extremum modeling provided in an embodiment of the present invention. Detailed Implementation

[0017] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0018] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0019] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0020] Figure 1 This is a flowchart of a coal and rock dynamic disaster early warning method based on Bayesian dynamic extremum modeling, provided by an embodiment of the present invention. Figure 1 As shown, the method specifically includes the following steps:

[0021] Step S102: Based on a preset time window, segment and extract features from the multi-source sensor signal data of the coal mine working face to construct dynamic extreme event indicators and dynamic covariate vectors.

[0022] Step S104: Based on the multi-frequency sampling asynchronous fusion mechanism, the dynamic covariate vector is combined and updated, and a dynamic covariate matrix is ​​generated by combining dynamic extreme event indicators.

[0023] Step S106: Based on the dynamic covariate matrix, construct a Bayesian dynamic extremum model; the Bayesian dynamic extremum model includes a cumulative probability distribution model of dynamic extreme event indicators with respect to time-varying parameters and a time-varying parameter state equation of time-varying parameters with respect to the dynamic covariate vector.

[0024] Step S108: Update the Bayesian dynamic extreme value model based on the Markov chain Monte Carlo method, and calculate the dynamic risk index of the coal mine working face based on the Bayesian dynamic extreme value model.

[0025] Step S110: Based on the dynamic risk index, conduct disaster early warning for the coal mine working face.

[0026] Specifically, the method provided in this embodiment of the invention further includes: synchronously acquiring multi-source sensor signal data in the coal and rock mining face; the multi-source sensor signal data includes acoustic emission sensor time series data, electromagnetic radiation sensor time series data, gas concentration sensor time series data, and surrounding rock stress time series data.

[0027] Preferably, the acoustic emission sensor, electromagnetic radiation sensor, and gas concentration sensor should move with the coal and rock mining face, and the distance between the acoustic emission sensor and the electromagnetic radiation sensor and the coal and rock mining face should be 10 to 20 meters; the surrounding rock stress gauge should also be continuously updated as the working face advances.

[0028] Specifically, the dynamic extreme event index includes the maximum value of absolute gas emission within a preset time window; the dynamic covariate vector includes the cumulative energy value of acoustic emission signal, the average intensity value of electromagnetic radiation signal, and the maximum rate of change of multiple sensor signals within the preset time window.

[0029] In one optional embodiment provided by the present invention, a fixed duration is used. Multi-source sensor signal data is segmented into a preset time window, and extreme features of each sensor within the preset time window are extracted: Calculate the preset time window. Maximum absolute gas emission As a dynamic extreme event indicator value; extracting a preset time window. The cumulative energy value of the internal acoustic emission signal, the average intensity value of the electromagnetic radiation signal, and the preset time window for extracting other sensor signals. The maximum rate of change within the range together form the dynamic covariate vector. .

[0030] Specifically, step S104 further includes: after completely storing the dynamic covariate vector into the preset data cache, reading the latest value of the dynamic covariate vector from the preset data cache, and reading the dynamic extreme event index corresponding to the same time window, and combining them into a dynamic covariate matrix.

[0031] Specifically, a preset data cache for the dynamic covariate vector is established; when the feature values ​​of all dynamic covariates within the latest time window are calculated and stored in the cache, a model update is triggered; the latest values ​​of each dynamic covariate are retrieved from the preset data cache and combined to generate the dynamic covariate vector. And read the dynamic extreme event index value corresponding to the time window. After completing this model update, clear the cache area to prepare for subsequent data.

[0032] Specifically, step S106 includes: Step S1061: The generalized extreme value distribution is used to model the dynamic extreme event index, resulting in a cumulative probability distribution model of the dynamic extreme event index with respect to time-varying parameters. The time-varying parameters include time-varying location parameters, time-varying scale parameters, and shape parameters. The mathematical form of the cumulative probability distribution model includes: In the formula, G() represents the cumulative probability distribution function, and Z() represents the cumulative probability distribution function.t As an indicator of dynamic extreme events, For time-varying position parameters, For time-varying scale parameters, Here, t represents the shape parameter and time. Step S1062: Based on the dynamic covariate vector, construct the time-varying parameter state equation; the mathematical form of the time-varying parameter state equation includes: In the formula, , … , Represents a dynamic covariate vector Different components in , … , , … These are the time-varying regression coefficients; the evolution of the time-varying regression coefficients follows a random walk process: , ,in , , , To control the variance hyperparameter of the evolution step size.

[0033] Specifically, step S108 further includes the following steps: Step S1081, set a normal prior distribution for the time-varying regression coefficients: , To set a uniform prior distribution for the shape parameters To set the variance hyperparameter controlling the evolution step size, an inverse gamma distribution is used as the prior distribution: , ; Step S1082, when there is data to be updated When the model update is triggered, samples are drawn from the posterior probability distribution of each parameter in the dynamic covariate matrix based on the Markov chain Monte Carlo method. Step S1083: Calculate the mean of the posterior distribution from the converged Markov chain Monte Carlo chain, and use it as the optimal estimate of the parameters at the current time.

[0034] Specifically, the dynamic risk index includes: the probability that the dynamic extreme event index exceeds a preset safety threshold within the current time window; the calculation formula for the dynamic risk index includes: In the formula, RC t G is a dynamic risk index.critcal This is a preset safety threshold.

[0035] Preferably, step S110 further includes: When the dynamic risk index is less than the first preset threshold, it is determined to be the first warning level, and the coal mine working face is determined to be in a safe state. When the dynamic risk indicator is greater than or equal to the first preset threshold and less than the second preset threshold, it is determined to be at the second warning level, and a warning message and a suggestion message are issued. When the dynamic risk indicator is greater than or equal to the second preset threshold, it is determined to be at the third warning level, and disaster emergency operations are executed.

[0036] For example, when When the value is less than 0.05, it is classified as a Level I warning, and the system displays a safe status; when 0.05 ≤ When the value is less than 0.25, it is classified as a Level II warning. The system issues a warning message and recommends reducing mining speed and increasing manual inspections. When the value is ≥ 0.25, it is determined to be a Level III warning, and the system will automatically execute shutdown, power outage, and evacuation commands. Specifically, if there is a difference between the level warning result and the actual situation, the level warning dynamic risk index can be adjusted appropriately.

[0037] Preferably, the method provided in this embodiment of the invention further includes: Using the mining engineering plan as the base map, the real-time dynamic risk indicators of coal mining faces in each area are overlaid and displayed in the form of a heat map; Establish an early warning decision database to store the input data, model parameter estimates, calculated dynamic risk indicators, and output early warning instructions corresponding to each early warning event; Regularly conduct retrospective analysis of the early warning decision logs and use historical data to recalibrate the hyperparameters of the Bayesian dynamic extreme value model.

[0038] As described above, this invention provides a method for early warning of coal and rock dynamic disasters based on Bayesian dynamic extreme value modeling. By establishing a dynamic extreme value theoretical model driven by multi-source information, it solves the problem of insufficient ability of traditional methods to capture the risk of instantaneous changes in disasters, realizes the probabilistic and adaptive real-time assessment of the precursory extreme events of coal and rock dynamic disasters, and significantly improves the accuracy and reliability of early warning.

[0039] Figure 2 This is a schematic diagram of a coal and rock dynamic disaster early warning system based on Bayesian dynamic extremum modeling, according to an embodiment of the present invention. Figure 2 As shown, the system includes: extraction module 10, generation module 20, construction module 30, update module 40, and early warning module 50.

[0040] Specifically, the extraction module 10 is used to segment and extract features from multi-source sensor signal data of the coal mine working face based on a preset time window, and to construct dynamic extreme event indicators and dynamic covariate vectors. The generation module 20 is used to combine and update the dynamic covariate vector based on the multi-frequency sampling asynchronous fusion mechanism, and generate a dynamic covariate matrix by combining dynamic extreme event indicators; Module 30 is used to construct a Bayesian dynamic extremum model based on a dynamic covariate matrix. The Bayesian dynamic extremum model includes a cumulative probability distribution model of dynamic extreme event indicators with respect to time-varying parameters and a time-varying parameter state equation of time-varying parameters with respect to the dynamic covariate vector. The update module 40 is used to update the Bayesian dynamic extreme value model based on the Markov chain Monte Carlo method, and to calculate the dynamic risk index of the coal mine working face based on the Bayesian dynamic extreme value model. Early warning module 50 is used to provide disaster early warning for coal mine working faces based on dynamic risk index.

[0041] Specifically, the generation module 20 is also used to completely store the dynamic covariate vector into the preset data cache, read the latest value of the dynamic covariate vector from the preset data cache, and read the dynamic extreme event index corresponding to the same time window, and combine them into a dynamic covariate matrix.

[0042] Specifically, module 30 is also used for: A generalized extreme value distribution is used to model the dynamic extreme event index, resulting in a cumulative probability distribution model of the dynamic extreme event index with respect to time-varying parameters. These time-varying parameters include time-varying location parameters, time-varying scale parameters, and shape parameters. The mathematical form of the cumulative probability distribution model includes: In the formula, G() represents the cumulative probability distribution function, and Z() represents the cumulative probability distribution function. t As an indicator of dynamic extreme events, For time-varying position parameters, For time-varying scale parameters, Here, t represents the shape parameter and time. Based on dynamic covariate vectors, time-varying parameter state equations are constructed; the mathematical form of the time-varying parameter state equations includes: In the formula, , … , Represents a dynamic covariate vector Different components in , … , , … These are the time-varying regression coefficients; the evolution of the time-varying regression coefficients follows a random walk process: , ,in , , , To control the variance hyperparameter of the evolution step size.

[0043] Specifically, update module 40 is also used for: To establish a normal prior distribution for the time-varying regression coefficients: , To set a uniform prior distribution for the shape parameters To set the variance hyperparameter controlling the evolution step size, an inverse gamma distribution is used as the prior distribution: , ; When data is pending update When the model update is triggered, samples are drawn from the posterior probability distribution of each parameter in the dynamic covariate matrix based on the Markov chain Monte Carlo method. The mean of the posterior distribution is calculated from the convergent Markov chain Monte Carlo chain and used as the optimal estimate of the parameters at the current time step.

[0044] Specifically, the early warning module 50 is also used for: When the dynamic risk index is less than the first preset threshold, it is determined to be the first warning level, and the coal mine working face is determined to be in a safe state. When the dynamic risk indicator is greater than or equal to the first preset threshold and less than the second preset threshold, it is determined to be at the second warning level, and a warning message and a suggestion message are issued. When the dynamic risk indicator is greater than or equal to the second preset threshold, it is determined to be at the third warning level, and disaster emergency operations are executed.

[0045] Specifically, such as Figure 2 As shown, the system also includes a display and analysis module 60, used for: Using the mining engineering plan as the base map, the real-time dynamic risk indicators of coal mining faces in each area are overlaid and displayed in the form of a heat map; Establish an early warning decision database to store the input data, model parameter estimates, calculated dynamic risk indicators, and output early warning instructions corresponding to each early warning event; Regularly conduct retrospective analysis of the early warning decision logs and use historical data to recalibrate the hyperparameters of the Bayesian dynamic extreme value model.

[0046] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for early warning of coal and rock dynamic disasters based on Bayesian dynamic extremum modeling, characterized in that, The method includes: Based on a preset time window, multi-source sensor signal data from coal mine mining faces are segmented and feature extracted to construct dynamic extreme event indicators and dynamic covariate vectors. The dynamic covariate vector is combined and updated based on the multi-frequency sampling asynchronous fusion mechanism, and a dynamic covariate matrix is ​​generated by combining the dynamic extreme event index. Based on the dynamic covariate matrix, a Bayesian dynamic extremum model is constructed; the Bayesian dynamic extremum model includes the cumulative probability distribution model of the dynamic extreme event index with respect to the time-varying parameters and the time-varying parameter state equation of the time-varying parameters with respect to the dynamic covariate vector; The Bayesian dynamic extreme value model is updated based on the Markov chain Monte Carlo method, and the dynamic risk index of the coal mine working face is calculated based on the Bayesian dynamic extreme value model. Based on the dynamic risk index, disaster early warning is provided for the coal mine working face.

2. The method according to claim 1, characterized in that, The method further includes: Simultaneously collect multi-source sensor signal data in the coal and rock mining face; the multi-source sensor signal data includes time series data of acoustic emission sensor, time series data of electromagnetic radiation sensor, time series data of gas concentration sensor, and time series data of surrounding rock stress.

3. The method according to claim 1, characterized in that, The dynamic extreme event index includes the maximum value of absolute gas outflow within the preset time window; The dynamic covariate vector includes the cumulative energy value of the acoustic emission signal, the average intensity value of the electromagnetic radiation signal, and the maximum rate of change of multiple sensor signals within the preset time window.

4. The method according to claim 1, characterized in that, The dynamic covariate vector is combined and updated based on a multi-frequency sampling asynchronous fusion mechanism, and a dynamic covariate matrix is ​​generated by combining the dynamic extreme event indicators, including: After the dynamic covariate vector is completely stored in the preset data cache, the latest value of the dynamic covariate vector is read from the preset data cache, and the dynamic extreme event index corresponding to the same time window is read and combined into a dynamic covariate matrix.

5. The method according to claim 1, characterized in that, Based on the aforementioned dynamic covariate matrix, a Bayesian dynamic extremum model is constructed, including: A generalized extreme value distribution is used to model the dynamic extreme event index, resulting in a cumulative probability distribution model of the dynamic extreme event index with respect to time-varying parameters. These time-varying parameters include time-varying location parameters, time-varying scale parameters, and shape parameters. The mathematical form of the cumulative probability distribution model includes: In the formula, G() represents the cumulative probability distribution function, and Z() represents the cumulative probability distribution function. t This refers to the dynamic extreme event index. The time-varying position parameter, The time-varying scale parameter is... Here, t represents the shape parameter; Based on the dynamic covariate vector, a time-varying parameter state equation is constructed; the mathematical form of the time-varying parameter state equation includes: In the formula, , … , Represents the dynamic covariate vector Different components in , … , , … These are time-varying regression coefficients; the evolution of these time-varying regression coefficients follows a random walk process: , ,in , , , To control the variance hyperparameter of the evolution step size.

6. The method according to claim 5, characterized in that, The Bayesian dynamic extremum model is updated based on the Markov chain Monte Carlo method, including: A normal prior distribution is defined for the time-varying regression coefficients: , A uniform prior distribution is set for the shape parameters. An inverse gamma distribution is set as the prior distribution for the variance hyperparameter of the control evolution step size: , ; When data is pending update When the model update is triggered, samples are drawn from the posterior probability distribution of each parameter in the dynamic covariate matrix based on the Markov chain Monte Carlo method. The mean of the posterior distribution is calculated from the convergent Markov chain Monte Carlo chain and used as the optimal estimate of the parameters at the current time step.

7. The method according to claim 5, characterized in that, The dynamic risk index includes: the probability that the dynamic extreme event indicator exceeds a preset safety threshold within the current time window; the calculation formula for the dynamic risk index includes: In the formula, RC t G is the dynamic risk index. critcal The preset security threshold is defined as follows.

8. The method according to claim 1, characterized in that, Based on the dynamic risk index, disaster early warning is provided for the coal mine working face, including: When the dynamic risk index is less than the first preset threshold, it is determined to be the first early warning level, and the coal mine working face is determined to be in a safe state. When the dynamic risk indicator is greater than or equal to the first preset threshold and less than the second preset threshold, it is determined to be at the second warning level, and a warning message and a suggestion message are issued. When the dynamic risk indicator is greater than or equal to the second preset threshold, it is determined to be at the third warning level, and disaster emergency operations are executed.

9. The method according to claim 1, characterized in that, The method further includes: Using the mining engineering plan as the base map, the real-time dynamic risk indicators of coal mining faces in each area are overlaid and displayed in the form of a heat map; Establish an early warning decision database to store the input data, model parameter estimates, calculated dynamic risk indicators, and output early warning instructions corresponding to each early warning event; Regularly perform retrospective analysis on the early warning decision logs and recalibrate the hyperparameters of the Bayesian dynamic extreme value model using historical data.

10. A coal and rock dynamic disaster early warning system based on Bayesian dynamic extremum modeling, characterized in that, This system is used to implement the coal and rock dynamic disaster early warning method based on Bayesian dynamic extremum modeling as described in any one of claims 1-9; the system includes: an extraction module, a generation module, a construction module, an update module, and an early warning module; wherein... The extraction module is used to segment and extract features from multi-source sensor signal data of coal mine working face based on a preset time window, and to construct dynamic extreme event indicators and dynamic covariate vectors. The generation module is used to combine and update the dynamic covariate vector based on the multi-frequency sampling asynchronous fusion mechanism, and generate a dynamic covariate matrix by combining the dynamic extreme event index. The construction module is used to construct a Bayesian dynamic extremum model based on the dynamic covariate matrix; the Bayesian dynamic extremum model includes a cumulative probability distribution model of the dynamic extreme event index with respect to time-varying parameters and a time-varying parameter state equation of the time-varying parameters with respect to the dynamic covariate vector; The update module is used to update the Bayesian dynamic extreme value model based on the Markov chain Monte Carlo method, and to calculate the dynamic risk index of the coal mine working face based on the Bayesian dynamic extreme value model. The early warning module is used to provide disaster early warning for the coal mine working face based on the dynamic risk index.

Images