CN-121999902-A - LSTM-based coal mine gas concentration time sequence prediction method and system

CN121999902ACN 121999902 ACN121999902 ACN 121999902ACN-121999902-A

Abstract

The invention relates to the technical field of coal mine gas concentration prediction, in particular to a coal mine gas concentration time sequence prediction method and system based on LSTM. And acquiring multi-source data such as mining machinery pose time sequence, microseismic monitoring event sequence, roadway environment monitoring and the like, constructing a damage-permeability evolution equation by combining geological information, inverting a coal rock mass dynamic stress damage field, calculating a gas dynamic permeability field and an emission source strong field, and finally predicting the future gas concentration distribution by combining historical data through a space-time prediction model. Real-time space positioning of a coal mine gas emission source is realized.

Inventors

CUI FAN
ZHANG LEI
SUN ZHENYAN
XU SHURONG
ZHANG QINGHUA
CHEN GONGHUA
WANG RAN
ZHANG GUIXIN
LIU ZHIWEI
MO LIANHONG
Ren yanchuan
YUAN ZHONGFENG

Assignees

中国矿业大学(北京)
贵州安晟能源有限公司
中煤科工集团重庆研究院有限公司

Dates

Publication Date: 20260508
Application Date: 20260127

Claims (10)

1. The LSTM-based coal mine gas concentration time sequence prediction method is characterized by comprising the following steps of: Collecting multi-source monitoring data of a target mining face, wherein the multi-source monitoring data at least comprise mining machine pose time sequence data, microseismic monitoring event sequences and roadway environment monitoring data; Obtaining geological information of a target mining face, and constructing a damage-permeability evolution equation of the target mining face based on the geological information; Based on the mining machine pose time sequence data and the microseismic monitoring event sequence, obtaining a coal rock mass dynamic stress damage field taking the current moment as a reference according to inversion of geological information; according to the dynamic stress injury field of the coal rock mass, combining a preset injury-permeability evolution equation, and calculating to obtain a gas dynamic permeability field; Solving and obtaining a gas dynamic emission source strong field at the current moment through a preset source strong calculation rule based on the gas dynamic permeability field, the gas pressure data in the roadway environment monitoring data and the coal gas content distribution updated according to the mining progress; And taking the strong field of the gas dynamic emission source as input, and carrying out gas concentration distribution prediction in a future period by utilizing a space-time prediction model in combination with historical gas concentration time sequence data.
2. The LSTM based coal mine gas concentration timing prediction method of claim 1, wherein multi-source monitoring data of a target mining face is collected, the multi-source monitoring data at least including mining machine pose timing data, microseismic monitoring event sequences, and roadway environment monitoring data, including: Acquiring position and posture time sequence data of the mining machine at a first sampling frequency through a positioning sensor and a posture sensor which are arranged on the mining machine, wherein the position and posture time sequence data of the mining machine at least comprise a space three-dimensional coordinate, a cutting drum rotating speed and a travelling speed; Acquiring a sequence of microseismic monitoring events at a second sampling frequency through microseismic sensor arrays distributed on a target mining face and surrounding rock, wherein each microseismic monitoring event in the sequence of microseismic monitoring events at least comprises occurrence time, space coordinates and release energy; Collecting roadway environment monitoring data at a third sampling frequency through an environment sensor arranged in a roadway, wherein the roadway environment monitoring data at least comprise gas concentration, gas pressure and wind speed; And establishing a unified space-time coordinate system, and performing time stamp alignment and space registration on the mining machine pose time sequence data, the microseismic monitoring event sequence and the roadway environment monitoring data to form synchronous multi-source monitoring data.
3. The LSTM based coal mine gas concentration timing prediction method of claim 1, wherein obtaining geological information of a target mining face and constructing an impairment-permeability evolution equation of the target mining face based on the geological information comprises: Obtaining geological information of a target mining face and coal and rock mass test data; acquiring an original permeability basic value of a coal rock mass according to the geological information; fitting according to the coal rock mass test data to obtain a coupling coefficient; extracting a damage coefficient of a target mining face from the dynamic stress damage field of the coal rock mass; and constructing the damage-permeability evolution equation based on the original permeability basic value, the coupling coefficient and the damage coefficient.
4. The LSTM-based coal mine gas concentration time sequence prediction method according to claim 1, wherein obtaining a coal rock mass dynamic stress damage field based on a current moment according to the geological information inversion based on the mining machine pose time sequence data and the microseismic monitoring event sequence comprises: establishing an initial geomechanical numerical model of the target mining face based on the geological information; converting the mining machine pose time sequence data into dynamic boundary load input to the initial geomechanical numerical model; The space coordinates and the release energy distribution of the microseismic monitoring event sequence in the current time window are used as observation data to be input into a data assimilation algorithm; And driving the initial geomechanical numerical model to update the state by using the data assimilation algorithm according to the observed data, matching a theoretical microseismic energy field output by the initial geomechanical numerical model with the observed data, and inverting and outputting the dynamic stress damage field of the coal rock mass, wherein the dynamic stress damage field comprises stress values and damage coefficients of each point in space.
5. The LSTM-based coal mine gas concentration timing prediction method of claim 1, wherein calculating the gas dynamic permeability field according to the coal rock mass dynamic stress injury field in combination with a preset injury-permeability evolution equation includes: Substituting the damage coefficient of each spatial position point in the dynamic stress damage field of the coal rock mass into the damage-permeability evolution equation; And calculating the gas dynamic permeability corresponding to each space point in parallel, and generating a gas dynamic permeability field which is isomorphic with the space of the dynamic stress injury field.
6. The LSTM-based coal mine gas concentration time sequence prediction method of claim 1, wherein solving and obtaining a gas dynamic emission source strong field at a current moment through a preset source strong calculation rule based on the gas dynamic permeability field, the gas pressure data in the roadway environment monitoring data and the coal gas content distribution updated according to the mining progress comprises: acquiring comprehensive conductance coefficients of a plurality of space coordinate points of the target mining face based on the geological information of the target mining face and the roadway environment monitoring data; Acquiring dynamic permeability of a plurality of space coordinate points of a target mining face from the gas dynamic permeability field; Acquiring gas pressure gradient absolute values of a plurality of space coordinate points of a target mining face based on the roadway environment monitoring data; According to the position and posture time sequence data of the mining machine and the geological information of the target mining face, calculating and obtaining residual gas contents of a plurality of space coordinate points of the target mining face; Calculating a plurality of dynamic gas emission sources of a plurality of space coordinate points based on the comprehensive conductance coefficient, the dynamic permeability, the absolute value of the gas pressure gradient and the residual gas content of the plurality of space coordinate points of the target mining face; and obtaining a strong field of the gas dynamic emission source based on the plurality of gas dynamic emission sources.
7. The LSTM-based coal mine gas concentration timing prediction method according to claim 6, wherein calculating a gas dynamic emission source intensity based on the integrated conductance coefficient, the dynamic permeability, the absolute value of the gas pressure gradient, and the residual gas content of a plurality of spatial coordinate points of a target mining face includes: and presetting a source intensity calculation mathematical model, wherein in the source intensity calculation mathematical model, the gas dynamic emission source intensity is positively correlated with the comprehensive conductance coefficient, the dynamic permeability, the absolute value of the gas pressure gradient and the residual gas content.
8. The LSTM-based coal mine gas concentration timing prediction method according to claim 6, wherein calculating residual gas contents of a plurality of spatial coordinate points of the target mining face according to the mining machine pose timing data and geological information of the target mining face includes: Dividing a coal body of a target mining face into three dynamic areas in a three-dimensional space according to the pose time sequence data of the mining machine in combination with geological information of the target mining face, wherein the three dynamic areas comprise a mined-out area, a current mining pressure relief influence area and an original undisturbed area; For the mined-out area, setting the residual gas content to be zero; For a current mining pressure relief influence area, calculating the real-time residual gas content of the current mining pressure relief influence area according to the time of the current mining pressure relief influence area from the start of mining pressure relief and the space distance between the current mining pressure relief influence area and a mining working face by using a preset gas content dynamic attenuation function, wherein the gas content dynamic attenuation function comprises a time attenuation factor and a space attenuation factor; And for the original undisturbed region, adopting an original gas content value determined by a geological survey report as the residual gas content.
9. The LSTM-based coal mine gas concentration timing prediction method of claim 1, wherein the gas concentration distribution prediction in a future period by using a space-time prediction model in combination with historical gas concentration timing data by taking the gas dynamic emission source strong field as an input includes: building a space-time prediction model, wherein the space-time prediction model is a mixed model formed by a long-term memory network and a convolutional neural network; Constructing space-time input feature vectors, wherein the space-time input feature vectors comprise historical gas concentration time sequence data and a gas dynamic emission source strong field; Recombining the space-time input feature vectors into a feature matrix with space dimension, inputting the feature matrix into a convolutional neural network module, and outputting to obtain an advanced space feature map; and expanding the advanced space feature map according to the time dimension, inputting the expanded advanced space feature map into a long-term and short-term memory network module, and outputting a continuous gas concentration predicted value sequence of each monitoring point in a future preset period.
10. An LSTM based coal mine gas concentration timing prediction system, wherein the system is configured to implement an LSTM based coal mine gas concentration timing prediction method as claimed in any one of claims 1 to 9, the system comprising: the multi-source monitoring data acquisition module is used for acquiring multi-source monitoring data of a target mining face, wherein the multi-source monitoring data at least comprise mining machine pose time sequence data, microseismic monitoring event sequences and roadway environment monitoring data; The permeability equation construction module is used for acquiring geological information of the target mining face and constructing a damage-permeability evolution equation of the target mining face based on the geological information; The stress damage distribution deduction module is used for obtaining a coal rock mass dynamic stress damage field taking the current moment as a reference according to inversion of the geological information based on the mining machine pose time sequence data and the microseismic monitoring event sequence; the dynamic permeability distribution calculation module is used for calculating and obtaining a gas dynamic permeability field according to the coal rock mass dynamic stress injury field and in combination with a preset injury-permeability evolution equation; the dynamic emission source intensity calculation module is used for solving and obtaining a gas dynamic emission source intensity field at the current moment through a preset source intensity calculation rule based on the gas pressure data in the gas dynamic permeability field and the roadway environment monitoring data and the coal gas content distribution updated according to the mining progress; The gas concentration distribution prediction module is used for predicting the gas concentration distribution in the future period by using a space-time prediction model in combination with historical gas concentration time sequence data by taking the strong field of the gas dynamic emission source as input.

Description

LSTM-based coal mine gas concentration time sequence prediction method and system Technical Field The invention relates to the technical field of coal mine gas concentration prediction, in particular to a coal mine gas concentration time sequence prediction method and system based on LSTM. Background In the existing coal mine gas concentration time sequence prediction method, a technology for inverting a stress field based on microseismic data and predicting the total gas emission amount of a working face through an empirical formula is provided, such as the technology disclosed in the patent application with the publication number of CN114810213A, but the technology has the defects of incapability of spatially positioning the gas emission source, poor decision support capability, disjointing with real-time mining actions and the like. Disclosure of Invention Aiming at the defects of incapability of spatially positioning a gas emission source, poor decision support capability, disjointing from real-time mining actions and the like in the prior art, the invention provides an LSTM-based coal mine gas concentration time sequence prediction method and system for solving the problems. The technical scheme for solving the technical problems is as follows: in a first aspect, the invention provides an LSTM-based coal mine gas concentration time sequence prediction method, which comprises the following steps: Collecting multi-source monitoring data of a target mining face, wherein the multi-source monitoring data at least comprise mining machine pose time sequence data, microseismic monitoring event sequences and roadway environment monitoring data; Obtaining geological information of a target mining face, and constructing a damage-permeability evolution equation of the target mining face based on the geological information; Based on the mining machine pose time sequence data and the microseismic monitoring event sequence, obtaining a coal rock mass dynamic stress damage field taking the current moment as a reference according to inversion of geological information; according to the dynamic stress injury field of the coal rock mass, combining a preset injury-permeability evolution equation, and calculating to obtain a gas dynamic permeability field; Solving and obtaining a gas dynamic emission source strong field at the current moment through a preset source strong calculation rule based on the gas dynamic permeability field, the gas pressure data in the roadway environment monitoring data and the coal gas content distribution updated according to the mining progress; And taking the strong field of the gas dynamic emission source as input, and carrying out gas concentration distribution prediction in a future period by utilizing a space-time prediction model in combination with historical gas concentration time sequence data. Optionally, collecting multi-source monitoring data of the target mining face, where the multi-source monitoring data at least includes mining machine pose time sequence data, microseismic monitoring event sequence, and roadway environment monitoring data, and includes: Acquiring position and posture time sequence data of the mining machine at a first sampling frequency through a positioning sensor and a posture sensor which are arranged on the mining machine, wherein the position and posture time sequence data of the mining machine at least comprise a space three-dimensional coordinate, a cutting drum rotating speed and a travelling speed; Acquiring a sequence of microseismic monitoring events at a second sampling frequency through microseismic sensor arrays distributed on a target mining face and surrounding rock, wherein each microseismic monitoring event in the sequence of microseismic monitoring events at least comprises occurrence time, space coordinates and release energy; Collecting roadway environment monitoring data at a third sampling frequency through an environment sensor arranged in a roadway, wherein the roadway environment monitoring data at least comprise gas concentration, gas pressure and wind speed; And establishing a unified space-time coordinate system, and performing time stamp alignment and space registration on the mining machine pose time sequence data, the microseismic monitoring event sequence and the roadway environment monitoring data to form synchronous multi-source monitoring data. Optionally, obtaining geological information of the target mining face, and constructing an injury-permeability evolution equation of the target mining face based on the geological information, including: Obtaining geological information of a target mining face and coal and rock mass test data; acquiring an original permeability basic value of a coal rock mass according to the geological information; fitting according to the coal rock mass test data to obtain a coupling coefficient; extracting a damage coefficient of a target mining face from the dynamic stress damage field of the coal