煤灰熔融温度预测方法研究现状  

Research progress on prediction methods of coal ash melting temperature

作　　者：刘硕[1] 杨伏生[1] 张小艳[2] 井云环[3] 杨磊[3] 蔡会武[1] 周安宁[1] 

机构地区：[1]西安科技大学化学与化工学院,陕西西安710054 [2]西安科技大学计算机学院,陕西西安710054 [3]神华宁煤集团煤化工公司,宁夏银川750011

出　　处：《洁净煤技术》2016年第1期60-65,共6页Clean Coal Technology

基　　金：国家自然科学基金资助项目(51174279);神华宁煤集团有限责任公司科技创新项目(2014095)

摘　　要：为准确预测煤灰熔融温度,论述了国内外建立煤灰熔融温度预测模型的现状,重点分析了线性回归法、BP神经网络法、支持向量机法和Fact Sage软件法的应用情况及误差。回归分析法的应用最为广泛,其中利用最小二乘法拟合的预测公式的相关性系数较高,但适应性较差;BP神经网络法适应性较强,但必须输入大量数据对模型进行训练;支持向量机法虽然优于回归分析法与BP神经网络法,但不能阐明煤灰熔融过程中矿物演变规律,不能科学说明灰熔融特性变化机理。Fact Sage软件法不仅有较高的预测精度,还可阐明煤灰熔融过程中矿物质演变规律,优化煤灰熔融温度的评价标准,建立更可靠的预测模型。因此,Fact Sage软件法是应用前景广阔的煤灰熔融特性预测方法。In order to predict coal ash melting temperature,the present situation of ash melting temperature prediction model at home and abroad was introduced,including regression analysis method,BP neural network method,support vector machine（ SVM） method and FactSage software method. The application of regression analysis method was widely used,the correlation coefficient of predicting formula fitted by the least square method was higher,while its adaptability was poorer. The adaptability of BP neural network was stronger,but the large amounts of data must be input for training the model. The SVM method was better than the first two,but it couldn＇t clarify the ash melting process of mineral evolution law,that meant it couldn＇t scientificly indicate ash melting characteristics change mechanism. Fact Sage software method had higher prediction accuracy,it could clarify the process of mineral ash fusion conversion and optimize the ash melting temperature of evaluation criteria. Based on the properties of Fact Sage software method,a more reliable prediction model could be established.

关 键 词：灰熔融特性 预测 FactSage 回归分析法 BP神经网络 支持向量机 

分 类 号：TQ536[化学工程—煤化学工程]