Volterra级数模型的一种并行辨识算法

A Parallel Algorithm of Volterra Identification

通过对Volterra级数模型全解耦辨识算法的研究,提出Volterra级数模型并行辨识算法的设计思想,设计了并行算法.用此方法可将辨识过程中并行执行的任务进行划分,并分配给不同的处理器去执行,同时设计了任务分配策略和数据交换方法.仿真结果表明,该算法有效减小了模型的在线辨识时间,能有效克服Volterra级数模型辨识中的维数灾难问题,且收敛速度快、稳态精度高,在工程应用中有很高的实用价值.

According to the study of the fully decoupled Volterra identification algorithm, this paper puts fonvard the design idea of the fully decoupled Voherra parallel identification algorithm, and designs a parallel algorithm. The tasks that can be paralleled are distributed to different processors, and the methods of task distribution and data exchange are designed. The result shows that the algorithm of the models can efficiently handle the problem of catastrophic dimension in the course of the Voherra series identification. The speed of its convergence is fast;the stability precision is very high; and then it is more utilizable.