汽轮机热耗率多模型建模方法研究  

Investigation on Multi-model Modeling Method of Steam Turbine Heat Rate

作　　者：牛培峰[1,2] 刘超[1] 李国强[1] 马云飞[1] 陈贵林[1,2] 张先臣[1,2] 

机构地区：[1]燕山大学工业计算机控制工程河北省重点实验室,河北秦皇岛066004 [2]国家冷轧板带装备及工艺工程技术研究中心,河北秦皇岛066004

出　　处：《计量学报》2015年第3期251-255,共5页Acta Metrologica Sinica

基　　金：国家自然科学基金(60774028);河北省自然科学基金(F2010001318)

摘　　要：针对汽轮机热耗率难以准确计算的问题，提出了核模糊C均值与混合蛙跳算法优化最小二乘支持向量机（LS—SVM）的汽轮机热耗率多模型建模方法，用来计算不同工况下的热耗率。该方法利用核模糊C均值算法对热耗率数据聚类，采用5折交叉验证平均误差作为LS—SVM参数选择的适应度值，利用混合蛙跳算法优化参数并建立局部模型，采用开关切换得到模型输出，以此实现热耗率的多模型建模。与单一的LS—SVM模型和BP网络热耗率预测模型比较，结果表明该多模型方法有更高的预测精确和更好的泛化能力，能更准确地计算汽轮机热耗率。Taking into account the problem that the heat rate of steam turbine is difficult to accurately calculate, a novel heat rate multi-model soft measurement methodology based on kernel fuzzy c-means and shuffled frog-leaping algorithm optimized least squares support vector machine （LS-SVM is proposed）, which is employed to calculate the heat rate under different working conditions. This method applies kernel fuzzy c-means algorithm clustering heat rate data. Taking the mean error of 5-fold cross-validation as fitness value of parameters selection for LS-SVM, LS-SVM based on SFLA is trained and established local model for each cluster, and then the model output is obtained by the switch way, so as to realize the heat rate multi-model method. Compared with the single LS-SVM model and BP network heat rate prediction model, the multimodel has a higher prediction accuracy and better generalization ability.

关 键 词：计量学 汽轮机热耗率 混合蛙跳算法 多模型建模 最小二乘支持向量机 核模糊c均值 

分 类 号：TP941[自动化与计算机技术]