摘要
煤层瓦斯含量是矿井瓦斯灾害防治及煤层气开发的基础参数,为提高煤层瓦斯含量预测的科学性及准确性,提出了基于WPA-BP神经网络的煤层瓦斯含量预测模型,并将其与DGC瓦斯含量直接测定结果对比分析;构建煤层瓦斯含量因素指标体系;对WPA-BP预测模型不断迭代训练,使其预测值与真实值绝对误差在1%以下;最后利用该预测模型对临近工作面煤层瓦斯含量进行预测,并将预测结果与DGC测定瓦斯含量对比分析。结果表明:随着指标因素增大,瓦斯含量变大;WPA-BP神经网络预测模型相对误差为0.06%~12.92%(平均1.83%);对比分析表明,预测模型预测结果比DGC直接测定的瓦斯含量高,主要是由于损失量的计算有误差导致的,应用深度学习预测煤层瓦斯含量可矫正煤层瓦斯含量测定的准确性。
Coalbed methane content is the basic parameter of mine gas disaster prevention and development of coalbed methane,in order to improve the scientificity and accuracy of coalbed methane content prediction,proposes a coalbed methane content prediction model based on WPA-BP neural network,and compares it with the direct determination of DGC gas content.Construct an index system for coal bed methane content factors;the WPA-BP prediction model is continuously iteratively trained so that the absolute error error between the predicted value and the true value is less than 1%;finally,the prediction model was used to predict the methane content of the coal seam near the working surface,and the prediction results were compared with the gas content measured by DGC.The results show that with the increase of the index factors,the gas content becomes larger.The relative error of the WPA-BP neural network prediction model is 0.06%-12.92%(average 1.83%);comparative analysis shows that the prediction model predicts that the gas content is higher than that directly determined by DGC,mainly due to the error in the calculation of the loss amount,and the accuracy of the determination of the coalbed methane content can be corrected by using deep learning to predict the coalbed methane content.
作者
陈杰
任金武
朱喜旺
CHEN Jie;REN Jinwu;ZHU Xiwang(Henan Energy and Chemical Industry Group Co.,Ltd.,Yonghua Energy Co.,Ltd.,Luoyang 471000,China)
出处
《煤炭技术》
CAS
北大核心
2022年第9期143-147,共5页
Coal Technology
