PCA-GA-SVM的回采工作面瓦斯涌出量预测  

Working face gas emission prediction based on PCA-GA-SVM

作　　者：张强[1] 贾宝山[1,2] 董晓雷[1] 李宗翔[1,2] 

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,辽宁阜新123000

出　　处：《辽宁工程技术大学学报（自然科学版）》2015年第5期572-577,共6页Journal of Liaoning Technical University (Natural Science)

基　　金：国家自然科学基金资助项目(51174109)

摘　　要：为预测回采工作面瓦斯涌出量,采用主成分分析(PCA)与遗传算法(GA)优化支持向量机(SVM)相耦合的方法,在样本数据的筛选上汲取主成分分析数据降维的优点,使选择的数据样本简洁且更具代表性;充分利用支持向量机训练速度快、能够获得全局最优解且具有良好泛华性能的特点,将遗传算法与其相结合,寻找最优的惩罚参数c和核函数参数g;建立基于PCA-GA-SVM的回采工作面瓦斯涌出量预测模型,并在实际中得到成功应用.研究结果表明:该预测模型预测的最大相对误差为16.15%,最小相对误差为2.43%,平均相对误差为13.25%,相比其他预测模型有更强的泛化能力和更高的预测精度.In order to forecast the gas emission of mining working face, this paper utilized principal component analysis and genetic algorithm （GA） to optimize the coupling of the method of support vector machine （SVM）, and took the advantages of absorbing the principal component analysis data dimension reduction in the sample data screening, As the result, the choice of data samples is concise and more representative. Making full use of support vector machine training speed can obtain the global optimal solution with characteristics of good performance of Shi, and by combining with genetic algorithm （GA）, the optimal penalty parameter c and the kernel function parameter g are searched. The SVM prediction model of the mining working face gas emission was established based on PCA - GA, and was successfully applied in practice use. Research results show that the prediction model has the maximum relative error of 30.15%, the minimum relative error of 5.13%, and the average relative error of 12.8603%. Compared with other prediction model, this model has better generalization ability and higher prediction precision.

关 键 词：主成分分析 支持向量机 瓦斯涌出量 遗传算法 数据降维 回采工作面 

分 类 号：TD712[矿业工程—矿井通风与安全]