基于最小二乘支持向量机的动力配煤着火特性预测模型  

Ignition characteristic prediction model for blending coal based on least squares support vector machine

作　　者：常爱英[1] 吴铁军[1] 包鑫[1] 江爱朋[2] 

机构地区：[1]浙江大学工业控制技术国家重点实验室,浙江杭州310027 [2]杭州电子科技大学自动化研究所,浙江杭州310018

出　　处：《煤炭学报》2010年第8期1380-1383,共4页Journal of China Coal Society

基　　金：国家自然科学基金资助项目(60904058)

摘　　要：将最小二乘支持向量机建模方法引入到动力配煤着火特性的分析建模中,针对配煤指标中计算困难的着火温度指标建立了最小二乘支持向量机模型,一方面克服了神经网络算法的过拟合、泛化能力弱等缺点;另一方面提高了求解过程的计算速度。采用微粒群算法(PSO)对模型参数进行优化,模型留一验证得到预测均方误差为8.60,相关系数为0.93,对65个样本进行预测分析,得到较高的预测精度。因此采用最小二乘支持向量机方法可以实现较精确的配煤着火温度预测。Least squares support vector machine method was adopted in the model of analysing the igniting characteristic of powered coal.In order to solve the problem of difficult to calculate the igniting temperature,the model was established and was following optimized by particle swarm optimization（PSO） .This model overcomes the over fitting of neural network algorithm,and weak generalization shortcomings on the one hand;increases the calculation speed of the solution process on the other hand.Experimental results show least squares support vector machine method gets the mean square error of prediction of 8.60,and the correlation coefficient of 0.93,which has high precise.Therefore,using least squares support vector machine can achieve a more accurate prediction ignition temperature of coal blending.

关 键 词：最小二乘支持向量机 动力配煤 着火温度 PSO 

分 类 号：TQ533[化学工程—煤化学工程]