一种改进的极限学习机煤与瓦斯突出预测模型  

A Prediction Model of Coal and Gas Outburst Based on Improved Extreme Learning Machine

作　　者：付华[1] 李海霞[1] 卢万杰[2] 徐耀松[1] 王雨虹[1] 

机构地区：[1]辽宁工程技术大学电气与控制工程学院,辽宁葫芦岛125105 [2]辽宁工程技术大学机械工程学院,辽宁阜新123000

出　　处：《传感技术学报》2016年第1期69-74,共6页Chinese Journal of Sensors and Actuators

基　　金：国家自然科学基金项目(51274118;70971059);辽宁省科技攻关项目(2011229011);辽宁省教育厅基金项目(L2012119)

摘　　要：较高精度的煤与瓦斯突出预测是煤矿安全生产的必要前提和保证。为了提高煤与瓦斯突出预测模型的预测精度,提出了一种改进的极限学习机煤与瓦斯突出预测模型。首先利用核主成分分析法对煤与瓦斯突出的影响指标进行降维简化处理,提取指标数据的主成分序列;把主成分序列分为训练样本和验证样本,然后在训练阶段,使用训练样本通过结合了全局搜索和局部搜索的文化基因算法对极限学习机的输入权值和隐含层偏差进行优化,得到最佳预测模型;最后,在最佳预测模型中,用验证样本对煤与瓦斯突出强度进行预测。通过实例验证,该模型能够有效预测煤与瓦斯突出强度。与BP、SVM、ELM、KPCA-ELM等预测模型相比,该模型具有更高的预测精度。Higher accuracy of coal and gas outburst prediction is the necessary prerequisite and guarantee for thecoal mine safety production. In order to improve the prediction accuracy of coal and gas outburst prediction model,aprediction model of coal and gas outburst based on improved extreme learning machine was proposed. Firstly,thismethod used Kernel Principal Component Analysis（KPCA）for coal and gas outburst index to dimension reduction,and extracted the key feature of the principal components. This paper lists example of training coal and gas outburstprediction model by key principal components,which are divided into two groups of training samples and test sam-ples. Then,to obtain the best prediction model,the weights of the input and hidden layer deviation in extreme learn-ing machine were optimized by training of training samples and utilizing Memetic algorithm which adopts the advan-tage of global search and local search. Finally,in the best prediction model,the intensity of coal and gas outburstwas predicted by using test samples. The analysis results show that：the model can effectively predict the intensity ofcoal and gas outburst. Compared with the prediction model of BP,SVM,ELM,KPCA-ELM,the model has higherprediction precision.

关 键 词：煤与瓦斯突出 预测模型 极限学习机 核主成分分析法 文化基因算法 

分 类 号：TP39[自动化与计算机技术—计算机应用技术] TP183[自动化与计算机技术—计算机科学与技术]