基于偏最小二乘支持向量机的燃煤电站锅炉效率模型

Coal-fired Boiler Efficiency Modeling Based on Partial Least Squares Support Vector Machines

针对电站锅炉燃烧系统非线性强、变量间强耦舍及信号噪声大等特点,提出了基于电站历史运行数据的锅炉效率建模方法。根据锅炉燃烧的机理选取关键输入变量,利用偏最小二乘原理(PLS)对其进行特征提取,建立锅炉效率与所提取特征之间的最小二乘支持向量机(LSSVM)关系模型,组成一个PLS-LSSVM混合模型,并利用电站实际数据对模型的准确性进行验证。结果表明:PLS-LSSVM模型相比于PLS模型具有更强的泛化能力,相比于LSSVM模型有更好的运行效率。

Considering the nonlinearity,multivariable coupling and strong noise of power plant boilers,a modeling method based on historical data was proposed to choose key variables which based on boiler combustion mechanism and to extract the feature by making use of partial least squares(PLS) so that a PLS-SVM (support vector machine) model between extracted feature and boiler efficiency can be established.Verifying this hybrid PLS-SVM model with actual operation data shows that the hybrid PLS-SVM model outperforms the PLS model in accuracy and generalization ability and the SVM model in operating efficiency.