基于粒子群优化的支持向量机在瓦斯浓度预测中的应用研究  

Application Research of Support Vector Machine Based on Particle Swarm Optimization in Gas Concentration Forecasting

作　　者：张剑英[1] 许徽[1] 陈娟[2] 曹新德[1] 

机构地区：[1]中国矿业大学信电学院 [2]中国矿业大学材料学院,江苏徐州221008

出　　处：《工矿自动化》2010年第10期32-35,共4页Journal Of Mine Automation

摘　　要：为了准确预测煤矿瓦斯浓度,基于从芦岭煤矿KJ98监控系统中提取的生产现场瓦斯浓度时间序列数据,对基于粒子群优化的支持向量机理论在瓦斯浓度短期预测中的应用进行了研究。首先对瓦斯浓度时间序列进行小波软阈值去噪和相空间重构等预处理,然后采用粒子群优化算法对支持向量机的惩罚因子、损失函数、核函数参数进行了优化,并基于最优参数建立了瓦斯浓度预测的支持向量机模型。仿真结果表明,采用粒子群优化的支持向量机理论进行煤矿瓦斯浓度预测,极大地提高了预测的准确性和精确度;误差分析结果表明,该方法预测结果的误差很小,且测试样本越小,误差越小。In order to forecast gas concentration of coal mine,the paper studied application of support vector machine based on particle swarm optimization in gas concentration short-term forecasting according to original gas concentration time series data in KJ98 monitoring and control system of Luling coal mine.In the application,gas concentration time series are pre-processed firstly by use of methods of soft-threshold de-noising of wave-let and phase-space reconstructing,then parameters of penalty factor,loss function and kernel function of support vector machine are optimized by use of particle swarm optimization algorithm,and a model of support vector machine of gas concentration forecasting is built based on the optimal parameters.The simulation result showed that the forecasting method of support vector machine based on particle swarm optimization greatly improves correctness and accuracy of gas concentration forecasting.Error analysis result showed that forecasting errors are small by use of the method and the less forecasting sample is,the smaller error is.

关 键 词：煤矿 瓦斯浓度 预测 支持向量机 粒子群优化 建模 

分 类 号：TD712.5[矿业工程—矿井通风与安全]