基于LS-SVM的单元机组协调控制系统模型辨识  

Coordinated Control System Identification Based on LS-SVM

作　　者：袁一丁[1] 董泽[1] 姚峻 

机构地区：[1]河北省发电过程仿真与优化控制工程技术研究中心(华北电力大学),河北保定071003

出　　处：《山东电力技术》2017年第11期38-41,51,共5页Shandong Electric Power

基　　金：国家自然科学基金项目(71471060);山西省煤基重点科技攻关项目(MD2014-03-06-02)

摘　　要：预测模型的准确性在非线性预测控制中起着重要的作用。采用最小二乘支持向量机(LS-SVM)回归算法辨识单元机组协调控制系统模型。LS-SVM具有良好的非线性逼近能力和较快的计算速度,适用于非线性系统的实时建模。考虑到系统的耦合问题,提出将MIMO对象分解为多个MISO对象,分析历史输入和输出数据对每个输出分量进行建模。通过基于实际采样数据的仿真实验,证明了该模型辨识方法是有效的,且具有很好地泛化能力。The accuracy of the predictive model plays an important role in the nonlinear predictive control.Least squares support vector machine（LS-SVM）,which has good nonlinear approximation capability and fast calculation speed,is suitable for real-time identification of nonlinear model.In this paper,LS-SVM regression algorithm is used to identify the model of coordinated control system.Considering the coupling problem,a decoupled modeling method is proposed in this paper by which,the Multi-Input-Multi-Output（MIMO） object is decomposed into multiple Multi-Input-Single-Output（MISO） objects.Each output component is modeled by analyzing the historical input and output data.Experiments based on the actual field sampling data demonstrate that the prediction model proposed in this paper is effective and has good generalization ability.

关 键 词：协调控制系统 最小二乘支持向量机 多输入多输出对象 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]