基于最近邻聚类支持向量机辨识的电弧炉电极逆控制-控制理论与应用.PDF

27 7 Vol. 27 No. 7 2010 7 Control Theory Applications Jul. 2010 : 1000−8152(2010)07−0909−07 , , (, 243002) : (support-vector-machines, SVM), SVM , SVM, PDSVM. MIMO , , , , . , . : α; ; ; ; ; : TP273 : A Inverse control for electrodes in electric arc furnace based on support-vector-machines identification on nearest neighbor clustering ZHANG Shao-de, MAO Xue-fei, MAO Xue-qin (School of Electrical Engineering and Information, Anhui University of Technology, Ma’anshan Anhui 243002, China) Abstract: Based on the equivalency between the support-vector-machines(SVM) with kernel functions and the three- layer feedforward neural network, we use support vector machines to build a multi-variable αth-order time-delay inverse system which realizes the pseudo-linear decoupling for the original system. A SVM model identification algorithm on nearest neighbor clustering is proposed; and the control strategy is designed which combines the feedforward self-tuning- parameter PD regulator with the inverse control based on SVM. Through the simulation research on the typical MIMO discrete nonlinear invertible system and the electrode system of the electric arc furnace, we find that the control strategy does not require the a priori knowledge of the mathematical model. Only a small number of input and output data in the sample learning process are sufficient to achieved a high-precision inverse system model. The control system has desirable characteristics of dynamic response and tracking accuracy. The model is highly robust to serious uncertainties, parameters perturbations, and outside interferences. Key words: αth-order time-delay inverse system; pseudo-linear dec