煤层瓦斯含量PSO-BP神经网络预测模型及其应用  

Study on PSO-BP neural network prediction method of coal seam gas content and its application

作　　者：林海飞 高帆[1] 严敏 白杨 肖鹏[1,2] 谢行俊 LIN Haifei;GAO Fan;YAN Min;BAI Yang;XIAO Peng;XIE Xingjun(College of Safety Science and Engineering,Xi'an University of Science and Technology,Xi'an Shaanxi 710054,China;Key Laboratory of Western Mine Exploitation and Hazard Prevention,Ministry of Education,Xizan University of Science and Technology,Xi'an Shaanxi 710054,China)

机构地区：[1]西安科技大学安全科学与工程学院,陕西西安710054 [2]教育部西部矿井开采及灾害防治重点实验室,陕西西安710054

出　　处：《中国安全科学学报》2020年第9期80-87,共8页China Safety Science Journal

基　　金：国家自然科学基金资助(51674192,51874236);陕西省杰出青年项目(2020JC-48);陕西省联合基金培育重点项目(2019JLP-02)。

摘　　要：为提高煤层瓦斯含量预测的科学性及准确性,提出基于粒子群算法(PSO)优化误差反向传播(BP)神经网络的瓦斯含量预测模型(PSO-BP模型);研究试验矿井煤层瓦斯含量与埋深、煤厚、底板标高以及测点与断层垂距等因素之间关系;对比分析该模型与多元线性回归模型和BP神经网络模型的预测结果。结果表明:随着埋深、煤厚及测点与断层垂距的增大,瓦斯含量变大,煤层底板标高增大,瓦斯含量变小;PSO-BP神经网络预测模型相对误差为2.4%~4.8%(平均3.1%),多元线性回归模型为2.3%~77.4%(平均27.7%),BP神经网络预测模型为7.5%~14.8%(平均10.2%),PSO-BP神经网络预测模型预测精度最高。In order to make more scientific and accurate prediction of coal seam gas content,a predication model was proposed based on PSO-BP neural network.Firstly,relationship between gas content and buried depth,seam thickness,floor elevation and vertical distance from fault to measuring points were analyzed.Then,prediction results of this model,multiple linear regression model,and BP neural network model were compared and analyzed.The results show that as burial depth,seam thickness,and vertical distance grow,gas content increases,but it will decrease when floor elevation increases.And relative error of PSO-BP neural network model is 2.4%-4.8%(3.1%on average),that of multiple linear regression model is 2.3%-77.4%(27.7%on average),and BP neural network is 7.5%-14.8%(10.2%on average),showing PSO-BP model has the highest prediction accuracy.

关 键 词：瓦斯含量 粒子群算法(PSO) 误差反向传播(BP)神经网络 预测模型 

分 类 号：X936[环境科学与工程—安全科学]