基于支持向量回归机的煤层瓦斯含量预测研究  

Study on Prediction of Coal Seam Gas Content Based on Support Vector Regression

作　　者：聂百胜[1,2] 戴林超[1,2] 颜爱华[2,3] 杨华[1,2] 

机构地区：[1]中国矿业大学(北京)煤炭资源与安全开采国家重点实验室,北京100083 [2]中国矿业大学(北京)资源与安全工程学院 [3]国家安全生产监督管理总局研究中心,北京100713

出　　处：《中国安全科学学报》2010年第6期28-32,共5页China Safety Science Journal

基　　金：国家"十一五"科技支撑计划(2007BAK22B05);教育部新世纪优秀人才支持计划资助(NCET-07-0799);河南省煤矿瓦斯与火灾防治重点实验室开放基金项目(HKLGF200907)

摘　　要：为了对煤层瓦斯含量进行准确预测,应用支持向量回归机(SVR)理论建立煤层瓦斯含量预测模型,结合现场实测数据利用支持向量机(SVM)工具箱进行模型的求解及预测,并从均方根误差、希尔不等系数和平均绝对百分误差3个不同误差指标与人工神经网络预测模型进行比较分析。研究结果表明:SVR模型其预测精度及可行性高于神经网络模型,而且运算快,实时性较好,用于煤层瓦斯含量的预测较理想,具有良好的应用前景,可以为煤矿瓦斯防治提供理论依据。In order to accurately predict coal seam gas content,the theory of SVR is applied to establishing the prediction model of coal seam gas content,and SVM toolbox is used to solve the model and prediction with the measured data.From the three different error indicators of root-mean-square error,hill inequality coefficient and mean absolute percent error,a comparison and analysis is made with artificial neural network prediction model.The results show that,the accuracy and feasibility of SVR model prediction is much higher than that of the neural network model,and its computing speed is more satisfactory than the latter in terms of real-time.The model can better forecast the coal seam gas content and has a good prospect of application.It provides a theoretical basis for the prevention and control of coal gas.

关 键 词：煤层瓦斯含量 支持向量回归机(SVR) SVM工具箱 误差指标 预测 

分 类 号：X936[环境科学与工程—安全科学]