Research On Identification Method Of Gas Emission Anomalies Based On Time Series Analysis

With the increase in the intensity and depth of coal mining in China,the number of casualties caused by gas disasters is still not to be underestimated.The abnormal warning and identification of gas emission anomalies still have great significance for coal mine safety.Although the monitoring platform of coal enterprises continuously monitors the gas status information,a single gas monitoring cannot fundamentally prevent the occurrence of gas disaster accidents.so the second use and feature mining of the time series are carried out.Identification of the type of gas emission anomalies can prevent and control gas disasters more effectively.This paper focuses on the pre-processing of time series,the capture of gas emission anomaly features,and the identification of gas emission anomaly categories.The main work is as follows:(1)Based on the pre-processing of time series anomalies and missing data,considering the noise problem of gas time series,a gas time series denoising algorithm based on improved wavelet threshold is proposed.Aiming at the disadvantage that the traditional wavelet threshold denoising method does not fully consider the impact of the change of the decomposition scale on the threshold,a new threshold definition principle is proposed and the threshold function is improved.The signal decomposition scale is introduced so that the threshold decreases with the increase of the decomposition scale,Do different quantization processing on wavelet coefficients of each scale,so as to achieve the purpose of protecting effective signals.Experimental analysis using gas time series shows that the denoising method proposed in this paper is superior to the traditional denoising method,and retains the effective signal components on the basis of eliminating noise interference.(2)From the perspective of statistical analysis principles,the Ryan-Joiner normal test method is used to analyze the probability distribution characteristics of gas time,and the distribution characteristics of gas time series are used as the basis for capturing the abnormal characteristics of gas emission,and whether the distribution characteristics of time series data are normal As a criterion for the occurrence of abnormal gas emission.The analysis of the experimental results shows that when there is no obvious fluctuation of gas emission and the change is normal,the time series follows a normal distribution;while during the abnormal gas emission period,the gas emission fluctuates significantly,and the gas time series does not follow the normal distribution.(3)According to different types of abnormal conditions,time series have different characteristics,and there are obvious differences between various test statistics.Based on the analysis of time series characteristics,the normal test,stationarity test and non-linear test methods are jointly used to obtain the gas time series feature information under different abnormal states.The least squares support vector machine model parameters are used to construct a classifier to classify and recognize the time series of abnormal gas emission.Through experimental comparison and analysis,the method in this paper has a high recognition accuracy for the abnormal gas emission.In this paper,Through multiple comparative experiments,the above-mentioned methods are verified and analyzed,which proves the effectiveness of those method.The research results have important guiding significance and auxiliary application value for coal mine safety and gas disaster prevention and control.