摘要
为提高预测模型的可靠性,实现对煤层未采区域瓦斯含量的精确预测,以山阳煤矿5#煤层为研究对象,进行未采区瓦斯含量的预测。运用瓦斯地质学和多元线性回归分析法,得出基岩厚度、煤层厚度和埋深是影响该矿瓦斯赋存的主要因素,并将其作为BP神经网络模型的输入端神经元,初步构建出瓦斯含量预测模型;结合地勘时期瓦斯钻孔的实际数据,进行网络训练,再对预测模型的可靠性进行检验。结果表明:该预测模型预测瓦斯含量,精度较高,效果较好,能满足工程要求。采用多元线性回归-BP神经网络可以对未开采区域煤层瓦斯含量进行准确预测,为矿井瓦斯灾害防治提供一定的参考依据。
In order to improve the reliability of the prediction model and achieve accurate prediction of gas content in the unmined area,taking the No.5 coal seam of Shanyang Mine as the research object,the gas content in the unmined area was predicted.Using gas geology and multiple linear regression analysis method,it is concluded that bedrock thickness,coal seam thickness and buried depth are the main factors affecting the gas occurrence in the mine,and as the input neuron of BP neural network model,the gas content prediction model is preliminarily constructed.Combined with the actual data of gas drilling in the geological exploration period,the network training was carried out,and then the reliability of the prediction model was tested.The results show that the prediction accuracy of the model can meet the engineering requirements.Using the multi linear regression BP neural network predicts the gas content of the coal seam in the unexploited area,which can provide some reference for the prevention and control of the gas disaster in the mine.
作者
高望
张岩
高帅帅
GAO Wang;ZHANG Yan;GAO Shuai-shuai(Lyuliang Emergency Management Bureau,Lyuliang 033300,China)
出处
《陕西煤炭》
2020年第1期77-80,共4页
Shaanxi Coal
关键词
瓦斯预测
BP神经网络
网络训练
多元线性回归
gas prediction
BP neural network
network training
multiple linear regression
