Detection Of Coal Mine Gas Anomalies Based On Multi-Source Information Fusion And Causes

Coal mine safety management is one of the difficulties and emphases faced by coal enterprise safety management.Gas anomaly is a major risk hidden danger factor inducing coal mine gas disaster,and it is also one of the key contents of coal mine intelligent construction.Therefore,how to efficiently and accurately detect gas anomalies,accurately identify the state of potential risks,and then make scientific early warning has become a research issue in coal mine safety production process monitoring.With the advancement of intelligent construction of coal mines,a large amount of monitoring data covering the whole process has been accumulated.The mechanism of underground gas migration is complex,the gas monitoring data is interfered by underground production and environmental factors,and some problems such as false alarms and false negatives of gas sensors occur from time to time,so there is a certain risk in abnormal detection.Based on LZ intelligent coal mine of China Coal Group,Anhui Province,this paper adopts multi-parameter monitoring data fusion,deep learning and cause analysis methods to conduct in-depth research on gas anomaly detection and early warning.The main contents are as follows:(1)Through the analysis of the underground situation and monitoring multiparameter data,the abnormal characteristics of mine gas are introduced,and the environmental variable data such as underground gas,temperature,wind speed,carbon monoxide,mine pressure,upper corner and gas pre-extraction in return air lane are collected.The pretreatment method of gas concentration monitoring data is put forward.Through the completion of missing data and the time series of gas concentration after the original monitoring data is denoised,the missing values and noise data are processed by cubic exponential smoothing and wavelet denoising methods,and the overall statistical law of the original monitoring data is obtained,which ensures the authenticity and integrity of the data,and denoises the noise data to eliminate the data interference and obtain high reliability,thus laying a data foundation for the follow-up work.(2)Aiming at the massive gas monitoring data in coal mines,based on information fusion technology,the entropy weighted data fusion algorithm is adopted to fuse the gas concentration data in the upper corner,the gas concentration data in the working face and the gas concentration data in the return air lane,and then a BP neural network model is constructed to test the accuracy of the fusion algorithm,and the average absolute and root mean square errors of the data are calculated respectively,and the accuracy of the data before and after fusion is compared.The results show that,compared with the data before fusion,the prediction accuracy of the data after entropy weighted fusion is improved by about 45% and 50% in the training set,and by about 49% and 45% in the test set.(3)After processing the multi-parameter time series data of mine gas,a gas anomaly detection model based on prediction model and residual threshold method is constructed.The long-term and short-term memory network(LSTM)model is applied to the prediction of gas time series,and the MAE and RMSE of the model are calculated to be0.175 and 0.021,respectively,with a small error.On the basis of the prediction model,the deviation degree of residual threshold method is used to judge the abnormal state.In order to verify the effectiveness of the model,ROC curve performance evaluation method is introduced to verify the effect of the model.The AUC value of the model is 96.6%,which indicates that the gas anomaly detection model has high accuracy.(4)In order to implement the evaluation of mine safety state,based on the analysis of the correlation between multi-parameter gas anomaly data,a mine gas risk cause model based on the Apriori association rule algorithm is constructed to find out the frequent item set,identify the cause factors of the anomaly,and determine the large influencing factors of the gas anomaly.The results of association rules are taken as the basis of risk early warning analysis,and the threshold is determined to divide the gas risk classification and early warning level,so as to realize the fine hierarchical early warning analysis,and provide decision support for the risk precontrol of coal mine gas disaster.Figure 35 Table 17 Reference 101...