基于元素分析的煤粉工业分析GA-SVM预测模型

摘要

运用遗传算法(genetic algorithm,GA)和支持向量机(support vector machine,SVM)方法,建立了基于元素分析的煤质工业分析快速预测模型.该模型基于以干燥基为基准的6029组美国煤质数据,以煤质元素分析(C、H、O、N、S)为输入,工业分析(挥发分、固定碳)为输出.另外通过实验得到74组中国煤质数据,用于模型验证.结果表明:以美国煤质数据构建的模型,挥发分、固定碳预测平均相对误差分别为4.60%、3.22%;用中国煤质数据验证模型时,挥发分、固定碳预测平均相对误差分别为9.16%、3.55%.该模型预测误差较小,能较好地利用元素分析数据预测固定碳、挥发分.%Genetic algorithm (GA)and support vector machine (SVM)methods were used to build a rapid prediction method for pulverized coal proximate analysis based on ultimate analysis.According to 6 029 groups of the U.S.pulverized coal da-ta,this model took ultimate analysis on pulverized coal including C,H,O,N and S as inputs and proximate analysis inclu-ding volatile and fixed carbon as outputs.Another 74 groups of Chinese pulverized coal data were obtained through standard experiments for model verification.Corresponding results indicate average relative errors of prediction on volatile and fixed carbon of the U.S.pulverized coal are 4.60% and 3.22%,and average relative errors of prediction on volatile and fixed carbon of the Chinese pulverized coal are 9.16% and 3.55%.It is proved this model can well make use of ultimate analysis data to predict fixed carbon and volatile and has small prediction errors.