多变量非线性系统超平面神经网络辨识算法设计  

Design of hyperplane neural network identification algorithm for multivariable nonlinear system

作　　者：余世明[1] 孙云坤 岑江晖 何德峰[1] YU Shiming;SUN Yunkun;CEN Jianghui;HE Defeng(College of Information Engineering,Zhejiang University of Technology,Hangzhou 310023,China)

机构地区：[1]浙江工业大学信息工程学院,浙江杭州310023

出　　处：《浙江工业大学学报》2022年第2期119-127,共9页Journal of Zhejiang University of Technology

基　　金：国家自然科学基金资助项目(61773345)。

摘　　要：针对常规高效链接超平面(EHH)神经网络无法辨识多变量系统的问题,将多目标规划方法运用到该神经网络的训练过程中,提出了多变量非线性系统超平面神经网络辨识算法。与常规训练方法相比,该算法考虑到EHH神经网络在辨识多个输出变量时存在的冲突性,将各变量的辨识作为不同的优化目标,以此设计了线性加权训练方法和理想点训练方法,实现对多变量非线性系统的辨识。以循环流化床锅炉(CFBB)燃烧过程为例,利用大量的实验数据,使用该算法建立了CFBB燃烧过程模型,并验证了该算法的准确性。For the identification of multivariable nonlinear system,the traditional efficient hinging hyperplanes(EHH)neural network identification method is difficult to get results.In this paper,the hyperplane neural network identification algorithm for multivariable nonlinear system is proposed,where the multi-objective programming method is applied to the training process of the EHH neural network.Compared with the traditional EHH neural network identification method,the proposed algorithm takes into account the conflict in identifying multiple output variables,and takes the identification of various variablesas different optimization objectives.Basedonthis,the linear weighted training method and ideal point training method are designed to realize the identification of multivariable nonlinear system.Taking the combustion process of circulating fluidized bed boiler(CFBB)as an example,using a large number of experimental data,the identification algorithm proposed is used to establish the CFBB combustion process model,and the accuracy of the model is verified.

关 键 词：高效链接超平面神经网络 系统辨识 多目标规划方法 循环流化床锅炉 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]