基于数据挖掘技术的煤与瓦斯突出预测研究
|
摘要
煤与瓦斯突出严重威胁着煤矿安全生产和矿工的生命安全,能够对突出发生的危险性进行准确预测并及时采取预防措施显得尤为重要。煤与瓦斯突出会导致巷道中瓦斯浓度变化出现异常,由于瓦斯浓度具有连续性,能够跟踪反映突出这一动态过程,因此可以利用瓦斯浓度变化特征作为突出预测的指标。在装备有煤矿瓦斯监测监控系统的矿井中瓦斯浓度是可测的,瓦斯监控系统中强大的数据库技术能够存储大量的瓦斯浓度历史数据,为研究瓦斯浓度变化规律与突出发生危险性之间的关系提供了丰富的数据资源。但目前的煤矿瓦斯监测系统缺乏对数据作进一步分析的功能,只能对瓦斯浓度超限进行报警,对于可能发生的煤与瓦斯突出灾害不具备提前预报能力,造成大量数据资源的浪费。针对这一问题提出了将广泛用于经济学领域的数据挖掘技术引入对瓦斯浓度数据的分析中,试图挖掘出浓度变化异常与煤与瓦斯突出之间的关系,并建立相应的突出预测模型,使其能够根据一定时间段内瓦斯浓度变化特征进行分类,实现对是否存在煤与瓦斯突出危险进行预测,为矿井管理人员提供决策的指导。
通过建立数据挖掘过程模型确定各阶段要完成的工作。在数据准备阶段,从大量瓦斯浓度历史数据中甄选出部分突出发生前10小时内瓦斯浓度时间序列和正常情况下10小时内瓦斯浓度时间序列作为样本,利用ARIMA方法对样本序列进行建模,并提取模型系数作为样本的特征向量。在数据挖掘阶段,采用支持向量机这种机器学习方法作为挖掘算法,并利用样本特征向量对其进行训练,得到基于支持向量机的突出预测模型。最后对该模型和基于BP神经网络的预测模型进行仿真比较,结果表明基于支持向量机的预测模型具有较高的预测性能,因此利用数据挖掘技术建立的支持向量机预测模型能够对煤与瓦斯突出的危险性进行预测。
Coal mine safe production and miners’security are threatened by coal and gas outburst, and accurate prediction of the danger of coal and gas outburst and taking preventive measures timely are particularly important. The gas concentration is abnormity in the lane when the coal and gas outburst happens, the dynamic process of the coal and gas outburst is reflected by the continuity gas concentration, so characteristics of the gas concentration is used as the critical value of coal and gas outburst prediction. The gas concentration is detectable in the coal mine where the coal mine gas monitoring system is established, large amounts of the gas concentration data are stored in the system by strongly database technology, wealthy data resources is provided for making out the relation between the discipline of gas concentration and the danger of coal and gas outburst. But the coal mine gas monitoring system lacks of the function of further analysis, and could alarm only when the gas over limit, do not have the early warning capabilities for the potential hazards of coal and gas outburst, so causes the waste of large amounts of data. Data mining which is widely used in economics is introduced for this problem and tries to excavate the relation between abnormal concentration and coal and gas outburst, the prediction model is established, the characteristics of the gas concentration could be classified, the danger of coal and gas outburst is forecast, and the guidance to decision-making is provided for mine management.
Task of various stages is defined by establishment of data mining model. In data preparation phase of data mining process, some of gas concentration time series within ten hours before coal and gas outburst happening and some of gas concentration time series within ten hours before coal and gas outburst does not happen are made as samples get from large amounts of historical gas concentration data, the time series is modeled by ARIMA method, the parameters of model are used as feature vectors. In data mining phase, the machine learning methods of support vector machine is used as mining algorithm, the prediction model of coal and gas outburst based on support vector machine is trained by samples feature vectors. Finally, this prediction model is compared with the prediction model based on BP artificial neural network by simulation experiment, the prediction model based on support vector machine having the higher prediction performance is proved by experimental results, the danger of coal and gas outburst could be predicted by the prediction model based on support vector machine established by data mining.
通过建立数据挖掘过程模型确定各阶段要完成的工作。在数据准备阶段,从大量瓦斯浓度历史数据中甄选出部分突出发生前10小时内瓦斯浓度时间序列和正常情况下10小时内瓦斯浓度时间序列作为样本,利用ARIMA方法对样本序列进行建模,并提取模型系数作为样本的特征向量。在数据挖掘阶段,采用支持向量机这种机器学习方法作为挖掘算法,并利用样本特征向量对其进行训练,得到基于支持向量机的突出预测模型。最后对该模型和基于BP神经网络的预测模型进行仿真比较,结果表明基于支持向量机的预测模型具有较高的预测性能,因此利用数据挖掘技术建立的支持向量机预测模型能够对煤与瓦斯突出的危险性进行预测。
Coal mine safe production and miners’security are threatened by coal and gas outburst, and accurate prediction of the danger of coal and gas outburst and taking preventive measures timely are particularly important. The gas concentration is abnormity in the lane when the coal and gas outburst happens, the dynamic process of the coal and gas outburst is reflected by the continuity gas concentration, so characteristics of the gas concentration is used as the critical value of coal and gas outburst prediction. The gas concentration is detectable in the coal mine where the coal mine gas monitoring system is established, large amounts of the gas concentration data are stored in the system by strongly database technology, wealthy data resources is provided for making out the relation between the discipline of gas concentration and the danger of coal and gas outburst. But the coal mine gas monitoring system lacks of the function of further analysis, and could alarm only when the gas over limit, do not have the early warning capabilities for the potential hazards of coal and gas outburst, so causes the waste of large amounts of data. Data mining which is widely used in economics is introduced for this problem and tries to excavate the relation between abnormal concentration and coal and gas outburst, the prediction model is established, the characteristics of the gas concentration could be classified, the danger of coal and gas outburst is forecast, and the guidance to decision-making is provided for mine management.
Task of various stages is defined by establishment of data mining model. In data preparation phase of data mining process, some of gas concentration time series within ten hours before coal and gas outburst happening and some of gas concentration time series within ten hours before coal and gas outburst does not happen are made as samples get from large amounts of historical gas concentration data, the time series is modeled by ARIMA method, the parameters of model are used as feature vectors. In data mining phase, the machine learning methods of support vector machine is used as mining algorithm, the prediction model of coal and gas outburst based on support vector machine is trained by samples feature vectors. Finally, this prediction model is compared with the prediction model based on BP artificial neural network by simulation experiment, the prediction model based on support vector machine having the higher prediction performance is proved by experimental results, the danger of coal and gas outburst could be predicted by the prediction model based on support vector machine established by data mining.
引文
[1]吴华帮,许石青.煤与瓦斯突出机理及其预测技术研究[J].矿业快报,2008,24(6):47-50.
[2]白新华,贾天让,张子敏等.新密煤田瓦斯动力现象分析[J].煤田地质与勘探,2009, 37(4):19-21.
[3]韩军,张宏伟,霍丙杰.向斜构造煤与瓦斯突出机理探讨[J].煤炭学报,2008,32(8):908-912.
[4]陈清华,张国枢,秦汝祥等.基于瓦斯涌出异常的煤与瓦斯突出预报软件开发[J].煤田地质与勘探,2007,35(3):18-21.
[5]陈祖云.煤与瓦斯突出前兆的非线性预测及支持向量机识别研究[D].北京:中国矿业大学图书馆,2009,5.
[6]程远平,俞启香.中国煤矿区域性瓦斯治理技术的发展[J].采矿与安全工程学报,2007,24(4):383-390.
[7]蔡寒宇,魏国营,辛新平.石门短导硐快速揭煤防突技术研究与应用[J].河南理工大学学报(自然科学版),2007,26(5):489-492.
[8]张天军.富含瓦斯煤岩体采掘失稳非线性力学机理研究[D].西安:西安科技大学图书馆,2009,05,30.
[9]许文全,赵恩来,马衍坤等.钻屑量采样技术分析与改进[J].煤田地质与勘探,2008, 36(1):78-80
[10] Ahmed,M.,and J.W,Smith,Biogenic methane generation in the degradation of eastern Australian Permain coals[J].Organic Geochemistry,2001,32(1):809-816.
[11]周丕昌,刘万伦,李伟.大河边向斜地质构造对煤与瓦斯突出的影响[J].采矿与安全工程学报.2009,26(1):55-59.
[12]孙忠强.苏昭贵.张金锋.煤与瓦斯突出预测预报技术研究现状及发展趋势[J].能源技术与管理,2008(2):56-65.
[13]何秋学,王恩元,魏建平.煤岩电磁辐射的力-电耦合模型[J].科技导报,2007,25(17):46-49.
[14] Kovalenlo V.,Yarovoy A.G.,Ligthart L P.A novel cluter suppression algorithm for landmine detection with GPR[J].IEEE Transaction on Geoscience and Remote Sensing,2007,45(11):3740-3751.
[15]肖长亮,刘有哲.矿用变频器对监测系统影响的解决方案[J].煤矿安全,2008,39(4):89-91.
[16]刘章平.地震检波器串并组合性能分析[J].江汉石油科技,2007,3(9):16-18.
[17] Ian H Witten,Eibe Frank.Data Mining: Practical machine Learning Tools and Techniques[M].San Francisco: Morgan Kaufmann,2005.
[18]孙莹,胡学钢.基于频繁概念格的序列模式发现研究[J].计算机科学,2004,1(31):169-170.
[19] Qian Wan,Aijun An.Discovering Transitional Patterns and Their Significant Milestones in Transaction Databases[J].IEEE Transaction on Knowledge and Data Engineering,2009,21(12):1692-1707.
[20]陈宗海.基于复杂过程简化模型的DHP学习控制[J].控制与决策,2006,21(10):1087-1091.
[21] Chandra B,Varghese P P. Fuzzy SLIQ decision tree algorithm[J].IEEE Transaction on System,Man,and Cybernetics,Part2,2008,38(5):1294-1301.
[22]刘友军,汪林林.SPRINT算法的改进[J].计算机工程,2006,32(16):55-57.
[23] Ovanesova A V,Suarez L E,Application of wavelet transforms to damage detection in frame structures[J]. Engineering Structures.2004,26(1):39-49.
[24]付华.煤矿瓦斯灾害特征提取与信息融合技术研究[D].阜新:辽宁工程技术大学图书馆,2006,6.
[25] Ghasemi H,Canizares C.On-line damping torque estimation and oscillatory stability margin prediction[J]. IEEE Transaction on Power Systems,2007,22(2):667-674.
[26] Bruzzone L,Persello C.A novel context-sensitive semisupervised SVM classifier robust to mislabeled training samples[J]. IEEE Transaction on Geoscience and Remote Sensing,2009,47(7):2142-2154.
[27]张登峰,王执铨,张卫.一种改进的基于VC维的非平稳信号小波消噪方法[J].南京理工大学学报(自然科学版),2009,33(5):648-652.
[28]郑建华.基于支持向量机的数据挖掘[D].天津:天津大学图书馆,2004,6.
[29]莫宏伟,左兴权.人工免疫系统[M].北京:科学出版社,2009.
[30] Dasgupta D.Advances in artificial immune system[J].Computational Intelligence Magazine.2006,1(4):40-49.
[31]韩军.张宏伟.朱志敏等.阜新盆地构造应力场演化对煤与瓦斯突出的控制[J].煤炭学报,2007,9(32):934-938.
[32]赵明鹏,王宇林,梁冰等.煤(岩)与瓦斯突出的地质条件研究—以阜新王营矿为例[J].中国地质灾害与防治学报,1999,10(1):14-19. .
[2]白新华,贾天让,张子敏等.新密煤田瓦斯动力现象分析[J].煤田地质与勘探,2009, 37(4):19-21.
[3]韩军,张宏伟,霍丙杰.向斜构造煤与瓦斯突出机理探讨[J].煤炭学报,2008,32(8):908-912.
[4]陈清华,张国枢,秦汝祥等.基于瓦斯涌出异常的煤与瓦斯突出预报软件开发[J].煤田地质与勘探,2007,35(3):18-21.
[5]陈祖云.煤与瓦斯突出前兆的非线性预测及支持向量机识别研究[D].北京:中国矿业大学图书馆,2009,5.
[6]程远平,俞启香.中国煤矿区域性瓦斯治理技术的发展[J].采矿与安全工程学报,2007,24(4):383-390.
[7]蔡寒宇,魏国营,辛新平.石门短导硐快速揭煤防突技术研究与应用[J].河南理工大学学报(自然科学版),2007,26(5):489-492.
[8]张天军.富含瓦斯煤岩体采掘失稳非线性力学机理研究[D].西安:西安科技大学图书馆,2009,05,30.
[9]许文全,赵恩来,马衍坤等.钻屑量采样技术分析与改进[J].煤田地质与勘探,2008, 36(1):78-80
[10] Ahmed,M.,and J.W,Smith,Biogenic methane generation in the degradation of eastern Australian Permain coals[J].Organic Geochemistry,2001,32(1):809-816.
[11]周丕昌,刘万伦,李伟.大河边向斜地质构造对煤与瓦斯突出的影响[J].采矿与安全工程学报.2009,26(1):55-59.
[12]孙忠强.苏昭贵.张金锋.煤与瓦斯突出预测预报技术研究现状及发展趋势[J].能源技术与管理,2008(2):56-65.
[13]何秋学,王恩元,魏建平.煤岩电磁辐射的力-电耦合模型[J].科技导报,2007,25(17):46-49.
[14] Kovalenlo V.,Yarovoy A.G.,Ligthart L P.A novel cluter suppression algorithm for landmine detection with GPR[J].IEEE Transaction on Geoscience and Remote Sensing,2007,45(11):3740-3751.
[15]肖长亮,刘有哲.矿用变频器对监测系统影响的解决方案[J].煤矿安全,2008,39(4):89-91.
[16]刘章平.地震检波器串并组合性能分析[J].江汉石油科技,2007,3(9):16-18.
[17] Ian H Witten,Eibe Frank.Data Mining: Practical machine Learning Tools and Techniques[M].San Francisco: Morgan Kaufmann,2005.
[18]孙莹,胡学钢.基于频繁概念格的序列模式发现研究[J].计算机科学,2004,1(31):169-170.
[19] Qian Wan,Aijun An.Discovering Transitional Patterns and Their Significant Milestones in Transaction Databases[J].IEEE Transaction on Knowledge and Data Engineering,2009,21(12):1692-1707.
[20]陈宗海.基于复杂过程简化模型的DHP学习控制[J].控制与决策,2006,21(10):1087-1091.
[21] Chandra B,Varghese P P. Fuzzy SLIQ decision tree algorithm[J].IEEE Transaction on System,Man,and Cybernetics,Part2,2008,38(5):1294-1301.
[22]刘友军,汪林林.SPRINT算法的改进[J].计算机工程,2006,32(16):55-57.
[23] Ovanesova A V,Suarez L E,Application of wavelet transforms to damage detection in frame structures[J]. Engineering Structures.2004,26(1):39-49.
[24]付华.煤矿瓦斯灾害特征提取与信息融合技术研究[D].阜新:辽宁工程技术大学图书馆,2006,6.
[25] Ghasemi H,Canizares C.On-line damping torque estimation and oscillatory stability margin prediction[J]. IEEE Transaction on Power Systems,2007,22(2):667-674.
[26] Bruzzone L,Persello C.A novel context-sensitive semisupervised SVM classifier robust to mislabeled training samples[J]. IEEE Transaction on Geoscience and Remote Sensing,2009,47(7):2142-2154.
[27]张登峰,王执铨,张卫.一种改进的基于VC维的非平稳信号小波消噪方法[J].南京理工大学学报(自然科学版),2009,33(5):648-652.
[28]郑建华.基于支持向量机的数据挖掘[D].天津:天津大学图书馆,2004,6.
[29]莫宏伟,左兴权.人工免疫系统[M].北京:科学出版社,2009.
[30] Dasgupta D.Advances in artificial immune system[J].Computational Intelligence Magazine.2006,1(4):40-49.
[31]韩军.张宏伟.朱志敏等.阜新盆地构造应力场演化对煤与瓦斯突出的控制[J].煤炭学报,2007,9(32):934-938.
[32]赵明鹏,王宇林,梁冰等.煤(岩)与瓦斯突出的地质条件研究—以阜新王营矿为例[J].中国地质灾害与防治学报,1999,10(1):14-19. .