多输出支持向量机醋糟气化建模  

Modeling of multi-output support vector machine for vinegar residue gasification

作　　者：王立群[1] 白文斌[1] 许超杰[1] 王谦[1] 

机构地区：[1]江苏大学能源与动力工程学院,江苏镇江212013

出　　处：《江苏大学学报（自然科学版）》2017年第1期52-55,共4页Journal of Jiangsu University：Natural Science Edition

基　　金：国家科技支撑计划项目(2015BAD21B00);科技部农业成果转化项目2008GB2C100099);江苏省科技支撑计划项目(BE200851);江苏省机械清洁能源与应用重点实验室项目(QK09005)

摘　　要：以醋糟气化试验为基础,建立了多输出支持向量机气化模型,把醋糟气化的影响条件气化温度、醋糟含水量和生物质与水蒸气的质量比作为模型的输入,生产的燃气主要组分H_2,CH_4,CO,CO_2作为模型的输出,实现了从输入来预测输出的一个过程.以醋糟气化试验的数据作为训练样本,随机挑选其中的3组数据作为模型测试组,除去选中的3组数据后剩下的数据作为训练样本.建立醋糟气化模型,通过Matlab实现模型的仿真预测,模型最佳的惩罚参数C为9.551 2,最佳的宽度参数σ为92.723.模型建立完成后,对随机挑出的3组数据进行测试验证,预测得到各气体成分与试验数据作对比分析,H_2的平均误差为2.87%,CO为3.28%,CH_4为2.66%,CO_2为3.04%.The support vector machine gasification model was established based on the vinegar residue gasification experiment.Gasification temperature,vinegar water content and mass ratio of biomass to water vapor were input into the model with the main gas group of H2 ,CH4 ,CO and CO2 as output to realize the output process from input.To test the data from vinegar residue gasification experiment,3 sets of data were selected as model test group with the remaining data as training sample.The vinegar residue gasification model was established by Matlab.The best penalty parameter C was 9.551 2 with optimal width parameter sigma of 92.723.After the completion of model establishment,three groups of data were picked randomly and verified,and the prediction of gas composition was compared with the experimental data.The results show that the average errors of H2,CO,CH4 and CO2 are respectively 2.87%,3.28%, 2.66% and 3.04%.

关 键 词：醋糟气化 多输出支持向量机 模型 惩罚参数 宽度参数 

分 类 号：TK124[动力工程及工程热物理—工程热物理]