基于Isomap+SVR的瓦斯浓度预测方法
|
摘要
对煤矿关键测点的瓦斯浓度进行科学且准确的预测,是防范瓦斯灾害的关键.为了从瓦斯监测监控系统实时采集的煤矿安全环境数据中分析和挖掘瓦斯浓度信息,从而进行关键测点的瓦斯浓度预测,本文采用等度量映射算法(Isomap)结合支持向量回归算法(SVR)来预测瓦斯浓度.该方法首先通过Isomap算法将非线性的高维煤矿井下安全环境数据进行维数约减,然后利用SVR算法进行回归预测.通过实验分析与对比,该方法行之有效,与多元线性回归(MLR)、支持向量回归(SVR)方法相比,在预测精度上有一定的优势,且在瓦斯波动异常情况下,鲁棒性更强.
The key to prevent gas disasters is to scientifically and accurately predict the gas concentration of key measuring points in coal mines. In order to analyze and mining the gas concentration information from the real-time data of coal mine safety environment collected by the gas monitoring system,and then predict the gas concentration of key measuring points,this paper proposes a method based on Isomap and SVR to predict the gas concentration. This method first reduces the non-linear high-dimensional mine safety environment data by Isomap,and then uses SVR algorithm for regression prediction. Through experimental analysis and comparison,this method has some advantages over MLR and SVR in prediction accuracy,and has stronger robustness in case of abnormal gas fluctuation.
The key to prevent gas disasters is to scientifically and accurately predict the gas concentration of key measuring points in coal mines. In order to analyze and mining the gas concentration information from the real-time data of coal mine safety environment collected by the gas monitoring system,and then predict the gas concentration of key measuring points,this paper proposes a method based on Isomap and SVR to predict the gas concentration. This method first reduces the non-linear high-dimensional mine safety environment data by Isomap,and then uses SVR algorithm for regression prediction. Through experimental analysis and comparison,this method has some advantages over MLR and SVR in prediction accuracy,and has stronger robustness in case of abnormal gas fluctuation.
引文
[1]谢东海,何爱宝.回采工作面瓦斯涌出量的聚类分析及归类预测模型的建立[J].矿业工程研究,2017(1):29-33.
[2]郭思雯,陶玉帆,李超.基于时间序列的瓦斯浓度动态预测[J].工矿自动化,2018,44(9):20-25.
[3]董丁稳,刘洁,王红刚.矿井瓦斯浓度自适应预测及其预警应用[J].中国安全科学学报,2013(5):88-93.
[4]吴海波,施式亮,念其锋.基于Spark Streaming流回归的煤矿瓦斯浓度实时预测[J].中国安全生产科学技术,2017(5):84-89.
[5]张昭昭,乔俊飞,余文.基于动态神经网络的瓦斯浓度实时预测方法[J].控制工程,2016,23(4):478-483.
[6]曹博,白刚,李辉.基于PCA-GA-BP神经网络的瓦斯含量预测分析[J].中国安全生产科学技术,2015(5):84-90.
[7]付华,丰盛成,刘晶,等.基于DE-EDA-SVM的瓦斯浓度预测建模仿真研究[J].传感技术学报,2016(2):285-289.
[8]付华,訾海,孟祥云,等.一种EKF-WLS-SVR与混沌时间序列分析的瓦斯动态预测新方法[J].传感技术学报,2015(1):126-131.
[9]杨丽,刘晖,毛善君,等.基于多元分布滞后模型的瓦斯浓度动态预测[J].中国矿业大学学报,2016(3):455-461.
[10]李香元,蔡骋,何进荣.基于密度缩放因子的ISOMAP算法[J].计算机科学,2018,45(7):207-213.
[11]赵连伟,罗四维,赵艳敞,等.高维数据流形的低维嵌入及嵌入维数研究[J].软件学报,2005(8):1423-1430.
[12]邓军,雷昌奎,曹凯,等.煤自燃预测的支持向量回归方法[J].西安科技大学学报,2018,38(2):175-180.
[2]郭思雯,陶玉帆,李超.基于时间序列的瓦斯浓度动态预测[J].工矿自动化,2018,44(9):20-25.
[3]董丁稳,刘洁,王红刚.矿井瓦斯浓度自适应预测及其预警应用[J].中国安全科学学报,2013(5):88-93.
[4]吴海波,施式亮,念其锋.基于Spark Streaming流回归的煤矿瓦斯浓度实时预测[J].中国安全生产科学技术,2017(5):84-89.
[5]张昭昭,乔俊飞,余文.基于动态神经网络的瓦斯浓度实时预测方法[J].控制工程,2016,23(4):478-483.
[6]曹博,白刚,李辉.基于PCA-GA-BP神经网络的瓦斯含量预测分析[J].中国安全生产科学技术,2015(5):84-90.
[7]付华,丰盛成,刘晶,等.基于DE-EDA-SVM的瓦斯浓度预测建模仿真研究[J].传感技术学报,2016(2):285-289.
[8]付华,訾海,孟祥云,等.一种EKF-WLS-SVR与混沌时间序列分析的瓦斯动态预测新方法[J].传感技术学报,2015(1):126-131.
[9]杨丽,刘晖,毛善君,等.基于多元分布滞后模型的瓦斯浓度动态预测[J].中国矿业大学学报,2016(3):455-461.
[10]李香元,蔡骋,何进荣.基于密度缩放因子的ISOMAP算法[J].计算机科学,2018,45(7):207-213.
[11]赵连伟,罗四维,赵艳敞,等.高维数据流形的低维嵌入及嵌入维数研究[J].软件学报,2005(8):1423-1430.
[12]邓军,雷昌奎,曹凯,等.煤自燃预测的支持向量回归方法[J].西安科技大学学报,2018,38(2):175-180.