Search

CN-122024417-A - Multi-source data fusion-based tunneling face gas prediction method and system

CN122024417ACN 122024417 ACN122024417 ACN 122024417ACN-122024417-A

Abstract

The invention discloses a tunneling surface gas prediction method and a tunneling surface gas prediction system based on multi-source data fusion, which belong to the technical field of coal mine safety, and comprise the steps of acquiring geological exploration data and tunneling plan data and constructing a space-time risk field model; collecting real-time gas monitoring data and engineering working condition data, carrying out alignment fusion on the real-time gas monitoring data and engineering working condition data and geological risk codes in a time-space risk field model to generate a combined feature set, analyzing the combined feature set, identifying the inconsistency between a direct concentration data trend and an indirect evidence data trend to generate a contradiction signal sign, inputting the combined feature set and the contradiction signal sign into a dual-mode prediction engine to output a fusion prediction result, analyzing the fusion prediction result to generate grading early warning information, and triggering and controlling a preparation instruction according to the grading early warning information. According to the invention, a multisource data fusion and dual-mode prediction cooperative mechanism is adopted, and a space-time risk field model is combined, so that accurate prediction of gas concentration steady-state trend and sudden transition can be realized.

Inventors

  • WANG QIANG
  • YANG PEIPEI

Assignees

  • 晋城市综合检验检测中心

Dates

Publication Date
20260512
Application Date
20260413

Claims (10)

  1. 1. The tunneling face gas prediction method based on multi-source data fusion is characterized by comprising the following steps of: acquiring geological exploration data and tunneling plan data, constructing a space-time risk field model, and dynamically updating along with tunneling progress evolution, wherein the space-time risk field model outputs geological risk information of a tunneling area three-dimensional geological grid, and the geological risk information comprises gas occurrence attributes, geological complexity indexes and geological risk codes corresponding to tunneling positions; collecting real-time gas monitoring data and real-time engineering working condition data in a tunneling working environment, and carrying out alignment fusion on the real-time gas monitoring data, the real-time engineering working condition data and the geological risk code to generate a combined feature set; analyzing the combined feature set, identifying and quantifying the inconsistency between the direct concentration data trend and the indirect evidence data trend, and generating a contradictory signal sign; Inputting the combined feature set and the contradiction signal sign into a preset dual-mode prediction engine comprising a steady-state predictor and a transition detector, wherein the dual-mode prediction engine adjusts the prediction weights of the steady-state predictor and the transition detector according to the contradiction signal sign, performs cooperative prediction operation and outputs a fusion prediction result; Analyzing the fusion prediction result, generating hierarchical early warning information, and triggering a corresponding control preparation instruction according to the early warning level in the hierarchical early warning information.
  2. 2. The method for predicting gas of a tunneling surface based on multi-source data fusion according to claim 1, wherein the constructing a space-time risk field model and dynamically updating along with the evolution of the tunneling progress comprises: performing three-dimensional space discretization on the geological exploration data to generate a basic geological grid containing gas occurrence attributes; defining a dynamic prediction window in the basic geological grid according to the roadway track and the tunneling progress in the tunneling plan data; Constructing a space-time risk field model according to the basic geological grid and the dynamic prediction window; And extracting variation coefficients of geological parameters in the dynamic prediction window, generating a geological complexity index, and utilizing the geological complexity index to adaptively adjust the range and resolution of the dynamic prediction window and dynamically update the space-time risk field model.
  3. 3. The method for predicting gas in a heading face based on multi-source data fusion according to claim 1, wherein the generating a joint feature set comprises: Performing time window processing on the real-time gas monitoring data, and extracting gas concentration time sequence characteristics; Extracting engineering state characteristics of the running state of the heading machine from the real-time engineering working condition data; Inquiring the space-time risk field model according to the tunneling position, and extracting a geological risk code of the current position; And vectorizing and splicing the gas concentration time sequence characteristic, the engineering state characteristic and the geological risk code to form a combined characteristic set.
  4. 4. The method for predicting gas in a driving surface based on multi-source data fusion according to claim 1, wherein the generating contradictory signal signs comprises: separating direct concentration data and indirect evidence data from the combined feature set, wherein the indirect evidence data comprises at least one of acoustic emission signal features, coal wall temperature micro-variation features and ventilation pressure difference fluctuation features; Normalizing the change trend of the direct concentration data and the change trend of the indirect evidence data, and calculating to obtain a normalized deviation degree; acquiring a consistency threshold value obtained based on historical data statistical analysis; Comparing the degree of deviation to the consistency threshold, generating a valid contradictory signature when the degree of deviation exceeds the consistency threshold, and quantifying the intensity level thereof according to the degree to which the threshold is exceeded.
  5. 5. The method for predicting gas in a heading face based on multi-source data fusion according to claim 4, wherein the performing a collaborative prediction operation, outputting a fusion prediction result comprises: Acquiring a first threshold value for mode switching; When the contradiction signal sign is invalid or the intensity level of the contradiction signal sign is lower than a first threshold value of the mode switching, taking a first predicted value output by the steady-state predictor as a dominant prediction to carry out weighted fusion; When the contradiction signal mark is valid and the intensity level of the contradiction signal mark reaches or exceeds the first threshold value, the weight of a second predicted value output by the jump detector is increased, the space-time risk field model is triggered to be reversely optimized to correct the local geological attribute of the space-time risk field model, and the internal parameters of the steady-state predictor are updated by using the corrected local geological attribute; and carrying out dynamic weight fusion on the corrected first predicted value and the second predicted value which are output by the updated steady-state predictor, and generating a fusion predicted result.
  6. 6. The method of claim 5, wherein the triggering reverse optimizes the spatio-temporal risk field model to correct its local geological properties comprises: Taking the observation features in the combined feature set and the second predicted value as constraints to construct a geological parameter inversion optimization model; The geological parameter inversion optimization model is solved through iteration, and the local gas occurrence parameter correction quantity capable of explaining the current contradiction phenomenon is inverted; and applying the local gas occurrence parameter correction amount to a grid cell corresponding to the space-time risk field model to generate a dynamic evolution geological model.
  7. 7. The method for predicting gas in a heading face based on multi-source data fusion according to claim 1, wherein triggering the corresponding control preparation instruction according to the early warning level in the hierarchical early warning information comprises: when the early warning level is the first level, triggering an instruction for increasing the monitoring frequency of the target area sensor; When the early warning level is the second level, triggering a prompting instruction for reducing the tunneling speed and a state instruction for preparing to start the local enhanced ventilation device; And when the early warning level is the third level, triggering a forced instruction for controlling the suspension of the tunneling equipment and a sequence instruction for starting the emergency ventilation plan by interlocking.
  8. 8. The method for predicting gas in a driving surface based on multi-source data fusion according to claim 6, wherein the generating the dynamically evolving geological model further comprises: Packaging the combined feature set, the contradiction signal sign, the fusion prediction result and the change record of the dynamic evolution geological model into a prediction case; Storing the predicted cases to a preset predicted case library; And based on accumulated data in the prediction case library, periodically performing offline training and optimization on model parameters of the steady-state predictor and decision thresholds of the jump detector.
  9. 9. The method for predicting gas of a tunneling surface based on multi-source data fusion according to claim 1, wherein said analyzing the fusion prediction result, generating hierarchical early warning information comprises: Carrying out risk grade division on the fusion prediction result to obtain a preliminary risk grade; and carrying out confidence coefficient verification and working condition suitability correction on the preliminary risk level to generate grading early warning information.
  10. 10. A multi-source data fusion-based heading face gas prediction system applied to the multi-source data fusion-based heading face gas prediction method as claimed in any one of claims 1-9, wherein the system comprises: the data acquisition and model construction module is used for acquiring geological exploration data and tunneling plan data, constructing a space-time risk field model and dynamically updating along with tunneling progress evolution, wherein the space-time risk field model outputs geological risk information of a tunneling area three-dimensional geological grid, and the geological risk information comprises gas occurrence attributes, geological complexity indexes and geological risk codes corresponding to tunneling positions; The multi-source real-time data fusion module is used for collecting real-time gas monitoring data and real-time engineering working condition data in a tunneling working environment, aligning and fusing the real-time gas monitoring data, the real-time engineering working condition data and the geological risk code, and generating a combined feature set; the contradiction signal analysis and identification module is used for analyzing the combined characteristic set, identifying and quantifying the inconsistency between the direct concentration data trend and the indirect evidence data trend, and generating a contradiction signal mark; the dual-mode prediction engine module is used for inputting the combined feature set and the contradiction signal sign into a preset dual-mode prediction engine comprising a steady-state predictor and a transition detector, adjusting the prediction weights of the steady-state predictor and the transition detector according to the contradiction signal sign by the dual-mode prediction engine, executing cooperative prediction operation, and outputting a fusion prediction result; and the early warning decision and instruction generation module is used for analyzing the fusion prediction result, generating hierarchical early warning information and triggering a corresponding control preparation instruction according to the early warning level in the hierarchical early warning information.

Description

Multi-source data fusion-based tunneling face gas prediction method and system Technical Field The invention relates to the technical field of coal mine safety, in particular to a tunneling face gas prediction method and system based on multi-source data fusion. Background In underground projects such as coal mining and tunnel construction, a tunneling surface is a working surface where a roadway extends forward, and is also a high-risk area where harmful gases such as gas are gushed out and accumulated. As a flammable and explosive gas, abnormal surging or overrun accumulation of gas is one of the main causes of mine safety accidents. Therefore, the method accurately and timely predicts the change trend of the gas concentration of the tunneling surface, and is a key technical link for guaranteeing the life safety of operators and the property safety of equipment and realizing safe and efficient production. In the related technology, china patent publication No. CN121024694A discloses a coal gas content measuring method and device by utilizing the inversion of gas emission parameters of a tunneling surface, wherein the method comprises the steps of collecting gas concentration, wind speed, tunnel section parameters, geological parameters and environment parameters of the tunneling surface through a multi-parameter dynamic acquisition system, respectively calculating characteristic values of the gas emission quantity according to drilling and non-drilling operation shifts by adopting a dynamic inversion algorithm based on the collected parameters, inverting the gas content of a coal body in a current area, collecting a coal sample through a downhole one-stop measuring system and measuring the accurate gas content, optimizing the parameters of the dynamic inversion algorithm, inputting multi-source data into an intelligent inversion and early warning system to obtain a predicted value of the gas content in front of the tunneling surface and a gas protrusion risk level, and adjusting the collection frequency and an inversion model. However, the above prior art scheme has the following technical drawbacks. In the prior art, only geological parameters are used as static and fixed background condition references, a geological risk model which dynamically evolves along with the tunneling progress and is self-adaptively adjusted is not constructed, geological structure complexity and gas occurrence risk of an area which is not disclosed in front of the tunneling face cannot be effectively fused, and prediction is only limited to current monitoring data and is free of space foresight. When deviation exists between the actual geological condition and the initial knowledge of exploration, the prediction basis is distorted, so that the spatial positioning accuracy and trend prejudgement of gas concentration prediction are greatly reduced. The early warning logic in the prior art excessively depends on the direct trend analysis of the gas concentration historical data, and lacks a capturing and quantifying mechanism for trend inconsistency between multiple indirect evidence and the direct concentration data. The prior art can not identify the signals, is difficult to timely early warn nonlinear and sudden gas emission events, and has high safety risks of missing report and early warning delay. In the prior art, a single prediction model is adopted, a multimode collaborative prediction mechanism dynamically adjusted according to a risk situation is not available, and double requirements of daily steady state trend prediction and sudden risk jump detection cannot be met. The method has the advantages that the method lacks high-precision steady state prediction capability when the working condition is stable, and cannot be quickly switched to a detection mode sensitive to mutation when the gas mutation precursor appears, so that the model has strong rigidity and weak self-adaptive capability, and the prediction reliability and robustness under the complex working condition are insufficient. The prior art lacks an online reverse optimization and offline self-learning mechanism, cannot utilize real-time monitoring data and a prediction result to dynamically correct the model, cannot perform periodic training optimization on the prediction model based on historical working condition data, and cannot continuously improve prediction accuracy and decision logic along with data accumulation and working condition change, so that the environmental adaptability and accuracy of long-term operation are gradually degraded. Disclosure of Invention In order to solve the problems, the invention provides a tunneling surface gas prediction method and a tunneling surface gas prediction system based on multi-source data fusion, which adopt a multi-source data fusion and dual-mode prediction cooperative mechanism and combine a space-time risk field model to realize accurate prediction of gas concentration steady-state trend and sudden jump. Th