基于整体辨识策略的非线性自适应控制方法

Nonlinear adaptive control method based on global identification strategy

中文关键词  随机向量函数链接网络  非线性  自适应控制  未建模动态补偿  输出权值偏差惩罚

英文关键词  random vector function link network  nonlinear  adaptive control  unmodeled dynamic  output weight deviation penalty

中文摘要

      针对一类离散时间下的未知动态非线性系统, 为解决传统自适应控制方法在交替辨识非线性系统时由于辨识精度低而导致的控制性能差的问题, 本文提出了一种基于整体辨识策略的未建模动态补偿的自适应控制方法.利用随机向量函数链接(RVFL)网络的直链与增强结构特性挖掘其与低阶线性模型和高阶未建模动态项的等价对应关系, 并融入权值偏差惩罚项, 设计了网络模型参数在线更新算法辨识非线性系统参数. 根据在线辨识的线性模型参数和未建模动态估计量, 采用一步超前最优控制策略设计线性控制器和未建模动态补偿器. 数值仿真表明, 所提方法优于交替辨识下的非线性自适应控制方法, 并通过工业应用的仿真研究验证所提方法在工业上的可用性.最后, 对本文控制方法在实际应用中的潜在问题及理论受限条件的放松进行分析和展望.

英文摘要

      For a class of unknown dynamic nonlinear systems composed in discrete time , the traditional adaptive control method has the problem of poor control performance caused by low identification accuracy. To solve this problem, a new adaptive control method with unmodeled dynamic compensation based on global identification strategy is proposed. First, the equivalent correspondence between random vector function link (RVFL) network and low-order linear model and high-order unmodeled dynamic terms is mined by using the linear and enhanced structure characteristics. Then, the weight deviation penalty term is integrated to design the online updating algorithm of network model parameters to identify nonlinear system parameters. In addition, the one-step ahead optimal control strategy is used to design the linear controller and unmodeled dynamic compensator based on online identification of linear model parameters and unmodeled dynamic estimatorsl. Numerical experiments show that the proposed method is superior to the nonlinear adaptive control method based on alternating identification, and the industrial example verifies the industrial applicability of the proposed method. The potential problems of this control method in practical application and the relaxation of theoretical constraints are analyzed and prospected.