CAJ | 학술논문

为了准确预测回采工作面的瓦斯浓度,提出云自适应粒子群算法优化Elman神经网络的瓦斯浓度动态预测新方法。利用井下无线传感器网络监测系统采集的回采工作面瓦斯浓度时间序列作为样本,并对其进行数据降噪和相空间重构等预处理。采用CAPSO算法对Elman神经网络的权值、阈值进行寻优运算,建立了回采工作面瓦斯浓度动态预测模型。通过对MATLAB仿真得出结果研究表明：该模型的平均相对变动ARV值为0.000357,相对均方根误差RRMSE值为0.1056,对回采工作面的瓦斯浓度预测结果合理且可为矿井瓦斯防治工作提供有效理论依据。
위료준학예측회채공작면적와사농도,제출운자괄응입자군산법우화Elman신경망락적와사농도동태예측신방법。이용정하무선전감기망락감측계통채집적회채공작면와사농도시간서렬작위양본,병대기진행수거강조화상공간중구등예처리。채용CAPSO산법대Elman신경망락적권치、역치진행심우운산,건립료회채공작면와사농도동태예측모형。통과대MATLAB방진득출결과연구표명：해모형적평균상대변동ARV치위0.000357,상대균방근오차RRMSE치위0.1056,대회채공작면적와사농도예측결과합리차가위광정와사방치공작제공유효이론의거。
In order to accurately predict the gas concentration in mining working face, a new dynamic prediction method of gas concentration which based on the cloud adaptive particle swarm algorithm on optimizing Elman neural network was proposed. The samples came from gas concentration time series which were collected by the under-ground wireless sensor network system in working face. The pretreatment of data noise reduction,phase space recon-struction and so on was carried. The weights and threshold of Elman neural network was optimized by CAPSO algo-rithm,and the dynamic prediction model of gas concentration in mining working face was established. Through the research on the MATLAB simulation results show that:The model of the relative change of average ARV value is 0.000 357,the relative root mean square error of the RRMSE value is 0.1056,the heading stope gas concentration prediction provides an effective theoretical basis for reasonable results and for mine gas prevention and control work.