动力配煤线性预测与非线性预测方法研究  

Research on Linear and Nonlinear Prediction in Coal Blending

作　　者：陈栋楠[1] 欧俭平[1] 马爱纯[1] 周成建[1] 廖文高[1] 

机构地区：[1]中南大学能源科学与工程学院,湖南长沙410083

出　　处：《金属材料与冶金工程》2012年第2期44-48,共5页Metal Materials and Metallurgy Engineering

摘　　要：分别采用线性加权平均、线性拟合、BP神经网络和Elman神经网络方法建立煤质特性预测模型,对混煤煤质特性进行预测。将四种方法的预测结果进行分析比较,辨别混煤煤质特性与各单煤煤质特性间的线性与非线性关系。分析表明:混煤的发热量、水分、灰分和硫分与各单煤相应煤质特性具有简单的线性关系。其中,发热量、水分和硫分的线性拟合预测结果优于线性加权平均预测结果;混煤挥发分与各单煤挥发分之间呈非线性关系,Elman神经网络预测效果最佳。Prediction models of blended coal properties are established based on linear weighting method、linear fitting method、BP neural network and Elman neural network.Compared the results of these four methods to understand the relationship of coal properties between blended coal and single coal.Qnet.ar、Mad、AAD and Sad from the results show linear.Among them,the results of linear fitting method is better than that of linear weighting method when predicting Qnet.ar、Mad and Sad;The prediction result of Vad shows nonlinear and Elman neural network is the best method to predict it.

关 键 词：煤质预测 线性加权平均 线性拟合 BP神经网络 ELMAN神经网络 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]