基于PCA-PSO-ELM的瓦斯涌出量预测  

Prediction of Gas Emission Based on PCA-PSO-ELM

作　　者：王彦彬[1] Wang Yanbin(College of Business Administration,Liaoning Technical University,Huludao 125105,China)

机构地区：[1]辽宁工程技术大学工商管理学院,辽宁葫芦岛125105

出　　处：《湖南科技大学学报（自然科学版）》2020年第4期1-9,共9页Journal of Hunan University of Science And Technology：Natural Science Edition

基　　金：国家自然科学基金资助项目(71771111)。

摘　　要：为了更加准确有效地预测瓦斯涌出量,提出采用主成分分析结合粒子群算法、极限学习机的瓦斯涌出量预测方法,其中极限学习机中隐含层节点数量及激活函数的类型由粒子群算法进行组合优化.实验综合考虑影响回采工作面瓦斯涌出量的13个因素对沈阳某煤矿历史数据进行分析,首先采用主成分分析对数据进行降维,消除指标数据之间的相关性,将降维后的数据划分为训练集和测试集2部分,设计了粒子群算法的惯性权重,并由粒子群算法结合十折交叉验证对极限学习机的2个参数进行优化,选择最优参数组合建立预测模型,通过对测试集瓦斯涌出量进行预测,其均方误差为0.1083,优于采用极限学习机及随机森林的预测结果.In order to predict gas emission accurately and effectively,a method of using principal component analysis(PCA),particle swarm optimization(PSO)and extreme learning machine(ELM)was presented,in which the number of hidden layer neurons and the type of excitation functions in ELM were optimized by PSO.In the experiment,13 main influencing factors were considered and gas emission data of a coal mine in Shenyang were analyzed.First,PCA was used to reduce the data dimensions,eliminate the correlations of the data.Then the result was divided into training set and testing set,and the inertia weight of PSO was designed,meanwhile the two parameters of ELM were optimized by PSO with 10-fold cross-validation.After that,the prediction model was built with the optimized parameters.Finally,the gas emission of testing set samples were predicted,and the mean square error(MSE)was 0.1083,which was better than the prediction result by ELM and random forest.

关 键 词：瓦斯涌出量预测 主成分分析 粒子群算法 极限学习机 十折交叉验证 

分 类 号：TD712.5[矿业工程—矿井通风与安全]