PCA-GA-ELM煤矿瓦斯涌出量预测  

Prediction of mine gas emission based on PCA-GA-ELM

作　　者：洪林[1,2] 赫祥林 董晓雷[1] 杨志博 

机构地区：[1]辽宁工程技术大学安全科学与工程学院,辽宁阜新123000 [2]矿山热动力灾害与防治教育部重点实验室,辽宁葫芦岛125105 [3]东北煤田地质局一〇七勘探队,辽宁阜新123000

出　　处：《辽宁工程技术大学学报（自然科学版）》2015年第7期779-784,共6页Journal of Liaoning Technical University (Natural Science)

基　　金：国家自然科学基金项目(51004062);辽宁省教育厅科学技术研究项目(L2012122)

摘　　要：为对井下瓦斯涌出量进行预测,采用主成分分析与改进极限学习机相结合的方法,在样本数据的筛选上吸取主成分分析数据降维的优点;充分利用极限学习机训练速度快、能够获得全局最优解且具有良好的泛华性能的特点,将遗传算法与其相结合,选择最优的输入权值矩阵和隐含层偏差,避免随机产生所造成的误差.利用编写程序确定隐含层神经元个数,比依靠经验更为准确,在实际中得到成功应用.研究结果表明:运用PCA-GA-ELM预测模型最大相对误差为19.58%,最小相对误差为0.8%,平均相对误差为6.0551%.从预测模拟结果可以看出,利用主成分分析与改进极限学习机相结合模型进行预测,结果准确可靠,克服了以往模型的不足.In order to forecast the coal mine gas emission, this paper used the method of combining the principal component analysis and the improved extreme learning machine, and absorbed the advantages of principal component analysis data dimension reduction in the screening of the sample data. The choice of data samples is concise and more representative. Making full use of the extreme learning machine training speed can obtain the global optimal solution and has the characteristics of good performance of shi, and combining with genetic algorithm （GA）, and choosing the optimal input weight matrix and the hidden layer deviation, can avoid the error caused by random. Program is used to determine the number of hidden layer neurons, more accurate than relying on experience. Finally get successful application in practice. The results show that by using of the principal component analysis and improvement combined model to predict extreme learning machine, the results are accurate and reliable which overcomes the deficiency of the previous model. The model of mine gas emission prediction has a certain reference value.

关 键 词：矿业安全 涌出量 主成分分析 极限学习机 遗传算法 数据降维 

分 类 号：TD712[矿业工程—矿井通风与安全]