基于遗传算法的模型辨识

Identification of Steering Gear Model Based On Genetic Algorithm

关于模型辨识的问题,现在采用的识别方法有很多,每种方法都可以将模型的参数值辨识出来,但是常规的辨识方法计算所得到的参数值都不能确保所获得的解的全局性,而遗传算法就很好地弥补了易陷入局部最优的情况,介绍并分析了遗传算法的相关流程图与原理。辨识问题没有比较好的直接求解方法,都是先将辨识问题转化为优化问题。根据采集得到的舵机数据结合MATLAB遗传算法计算,对舵机系统连续采集的5组幅值相同频率不同的输出和输入信号进行优化,优化完成时最小适应度对应的参数就是模型的参数,仿真结果验证了MATLAB遗传算法在系统辨识问题上的有效性,并分析了遗传算法的精度高等优点,为后续进行精确的系统设计提供了依据,为仿真提供了坚实的基础。

There are many identification methods for model identification, however, it is difficult to guarantee the global optimality in traditional identification methods. The genetic algorithm can avoid the local optimum. This paper describes and analyzes the process and principle of genetic algorithm. Due to no direct solution for identification problem, the identification problem is always converted into optimization. The MATLAB genetic algorithm is used to optimize the parameter of a guidance and control components which has collected 5 groups of output and input signals of steering engine, the best fitness corresponding parameters of the completed optimization result are the model pa- rameters. The simulation results verify the validity of MATLAB genetic algorithm in system identification, they also prove the high precision of the genetic algorithm and provide the basis for the subsequent system design and simula- tion.