基于聚类分析与粒子群算法优化神经网络的灰熔点预测  

Prediction of Ash Melting Point Based on Cluster Analysis and Particle Swarm Optimization Neural Network

作　　者：樊宏桂 武成利[1,2] 李寒旭 沈澍昊[1,2] 陈和荆 Fan Honggui;Wu Chengli;Li Hanxu;Shen Shuhao;Chen Hejing(College of Chemical Engineering,Anhui University of Science and Technology,Huainan 232001,China;Anhui Province Coal Resources Comprehensive Utilization Engineering Technology Research Center,Huainan 232001,China)

机构地区：[1]安徽理工大学化学工程学院,安徽淮南232001 [2]安徽省煤炭资源综合利用工程技术研究中心,安徽淮南232001

出　　处：《山东化工》2023年第19期190-194,198,共6页Shandong Chemical Industry

摘　　要：针对解决燃煤锅炉或气化炉的结渣现象,影响锅炉安全性问题,以灰成分金属氧化物为自变量,灰熔点流动温度为因变量,建立了K-Means-PSO-BPNN的灰熔点预测模型,误差分析结果表明,经过粒子群算法优化,BP神经网络模型在聚类分析后的预测效果得到了显著提高,表现出更好的相关性,相关系数为0.967,高于未优化的0.917,平均绝对误差为5.81,小于未聚类的26.98,并且模型的准确性提高到98.89%。因此,聚类分析以及粒子群算法优化后的神经网络模型能够更准确预测煤灰的流动温度(FT)。In order to solve the problem of slagging in coal-fired boilers or gasifiers,which affects boiler safety,taking the ash composition metal oxide as the independent variable and the ash melting point flow temperature as the dependent variable,the ash melting point prediction model of K-Means-PSO-BPNN is established.The results of the error analysis showed that after the optimization of the particle swarm algorithm,the prediction of the BP neural network model after the clustering analysis was significantly improved,showing better correlation,the correlation coefficient is 0.967,higher than the unoptimized 0.917,the average absolute error is 5.81,less than the non clustered 26.98,and the accuracy of the model improved to 98.89%.Therefore,the neural network model optimized by clustering analysis and particle swarm optimization can more accurately predict the flow temperature of coal ash.

关 键 词：煤灰成分 灰熔点预测 均值聚类分析 粒子群算法 BP神经网络 

分 类 号：TQ533.5[化学工程—煤化学工程]