基于支持向量回归机的矿井突水量预测  

Prediction of Mine Water Inrush Quantity Based on Support Vector Regression

作　　者：秦洁璇[1] 李翠平[1] 李仲学[1] 赵怡晴[1] 

机构地区：[1]北京科技大学土木与环境工程学院金属矿山高效开采与安全教育部重点实验室,北京100083

出　　处：《中国安全科学学报》2013年第5期114-119,共6页China Safety Science Journal

基　　金：国家自然科学基金资助(51174032;51174260);教育部新世纪优秀人才支持计划资助项目(NCET-10-0225);中央高校基本科研业务费专项资金资助项目(FRF-TP-09-001A)

摘　　要：为更好地预测预防煤矿水害,遏制煤矿水害,针对煤层底板突水问题的非线性、小样本特点,通过研究支持向量回归机(SVR)原理,建立基于SVR的矿井底板突水量预测模型。从突水影响因素中选取属性特征,包括水压、含水层厚度、隔水层厚度、底板采动裂隙带深度和断层落差。用网格搜索法和5-折交叉验证法确定径向基核函数(RBF)及模型参数。运用模型对测试样本进行突水量预测。最后将构建的矿井突水SVR模型运用到国内某典型矿山的煤层底板突水预测中,对工作面煤层底板进行最大突水量预测。In order to better predict and prevent water disaster in coal mine,based on the feature that water inrush from coal floor is characterized by nonlinearity and small samples,a model for water inrush quantity prediction was built by studying the principle and characteristics of SVR.Firstly,attributive characters were chosen from the parameters which can be used to predict water inrush,including water pressure,aquifer thickness,aquiclude thickness,height of water flowing fractured zone and fault throw.Secondly,the radial basis function and parameters were gained by grid search and 5-folder cross validation.Then the water inrush quantity of test samples was predicted by SVR model.Finally,the model was used practically to predict the maximum water inrush quantity from a coal floor in a certain typical mine in China.

关 键 词：煤层底板 支持向量机(SVM) 支持向量回归机(SVR) 模型 突水量 

分 类 号：X936[环境科学与工程—安全科学]