基于马尔科夫残差修正的瓦斯浓度预测  

Prediction of gas concentration based on residual correction of Markov chain

作　　者：韩婷婷[1] 吴世跃[1] 王鹏军 

机构地区：[1]太原理工大学矿业工程学院,山西太原030024 [2]山西亚美大宁能源有限公司,山西晋城048000

出　　处：《工矿自动化》2014年第3期28-31,共4页Journal Of Mine Automation

基　　金：国家科技支撑计划项目(2007BAK29B01)

摘　　要：针对采用灰色神经网络预测瓦斯浓度时部分预测值精度不高的问题,提出用马尔科夫模型对三阶灰色神经网络模型预测结果进行修正的方法;介绍了灰色神经网络模型的建立和马尔科夫修正残差方法,并采用该方法对某煤矿不同时间、不同地点的瓦斯浓度进行分析预测。实际应用结果表明,经马尔科夫残差修正后的瓦斯浓度预测值与实测值的最大相对误差从14%减小到6%,修正后的瓦斯浓度变化曲线更接近实际瓦斯浓度变化趋势。In view of problem of low accuracy of part of prediction values while using gray neural network for gas concentration prediction, the paper proposed a method of using Markov model to correct prediction results of three-order gray neural network model. It described establishment of gray neural network model and Markov residual correction method, and used the method to analyze and predict gas concentration in different locations of a coal mine at different times. Practical application results show that the maximum relative error of predicted gas concentration and measured value reduced from 14% to 6~ after Markov residual correction, and the corrected gas concentration curve is closer to the actual changing trend of gas concentration.

关 键 词：瓦斯浓度预测 马尔科夫链 残差修正 灰色神经网络 

分 类 号：TD712.5[矿业工程—矿井通风与安全]