基于布谷鸟算法与bp 神经网络的煤灰变形温度预测 - 应用概率统计.pdf

第42卷 第12期 燃摇 料摇 化摇 学摇 学摇 报 Vol.42 No.12 2014年12月 Journal of Fuel Chemistry and Technology Dec. 2014 文章编号:0253鄄2409(2014)12鄄1423鄄08 基于布谷鸟算法与BP神经网络的煤灰变形温度预测 沈铭科,黄镇宇,王智化,周俊虎 (浙江大学 能源清洁利用国家重点实验室,浙江 杭州摇 310027) 摘摇 要:以120种煤样为数据基础,采用布谷鸟算法(CS)优化BP(Back Propagation)神经网络,建立了CSBP模型对单煤、煤 掺添加剂和配煤等3类样本的煤灰变形温度(DT)样本进行预测。 模型以煤灰化学成分及其组合参数等13个变量作为输入 量,以变形温度(DT)作为输出量。 CSBP模型预测结果与BP神经网络模型预测结果进行对比发现,无论是单煤、煤掺添加剂 还是配煤,CSBP模型较 BP 模型对煤灰变形温度(DT)的预测都更加精准,平均相对误差分别达到了3. 11%、4.08%和 4.22%。 另外,对比3类样本预测结果发现,无论是CSBP模型还是BP模型,相比单煤预测而言,煤掺添加剂及配煤的预测误 差都有明显的增加。 关键词:煤灰;灰变形温度;BP神经网络;布谷鸟算法;添加剂;配煤 中图分类号:TK223摇 摇 文献标识码:A Prediction of coal ash deformation temperature based on Cuckoo Search and BP Neural Network SHEN Ming鄄ke,HUANG Zhen鄄yu,WANG Zhi鄄hua,ZHOU Jun鄄hu (State Key Laboratory of Clean Energy Utilization,Zhejiang University,Hangzhou摇 310027,China) Abstract:On the basis of 120 coal ash samples,a CSBP model based on BP(Back Propagation) Neural Network optimized by Cuckoo Search (CS) was proposed for predicting the ash deformation temperature of single coals,coals mixed with addictives and mixed coals. The thirteen chemical composition parameters and combined parameterswere employed as inputs,and the ash deformation temperature was used as output of the CSBP model. Theresultsshowthat whether singlecoal,coalmixedwith additivesor mixedcoals,CSBPmodel has abetterperformancecomparedwith BPmodelandtheaveragerelativeerrorsarereducedto3.11%,4.08% and4.22%,respectively. In addition,comparing the prediction results of three kinds of samples,both the CSBP model and BP model have higher prediction errors for coals mixed with addictives and mixed coals more than that for single coals. Key words:coal ash;ash deformation temperature;BP Neural Network;Cu