模型降阶和参数估计的一种快速遗传算法  

A fast GA for model reduction and parameter estimation

作　　者：王凌[1] 李彬彬[1] 郑大钟[1] 金以慧[1] 

机构地区：[1]清华大学自动化系,北京100084

出　　处：《控制与决策》2005年第4期426-429,433,共5页Control and Decision

基　　金：国家自然科学基金项目(60204008;60374060);国家973计划项目(2002CB312200).

摘　　要：针对传统遗传算法求解模型降阶和参数估计时,适配值评价既费时又效率较低的缺点,对实数编码GA引入相似度和可信度的概念,采用插值方法进行适配值评价,大大减少了评价环节的计算量,提高了整个算法的效率和实时性.通过对典型模型降阶和参数估计问题的仿真,验证了所提出方法的可行性和有效性.Due to the time-consuming evaluation of fitness value, the efficiency of classical genetic algorithm is very low when solving model reduction and parameter estimation problems. By introducing similarity degree and (reliability) degree in real-coded genetic algorithm, fitness value can be evaluated by interpolation so as to reduce the computational effort of evaluation process and improve the efficiency and real-time property of the whole algorithm. Simulation results based on typical model reduction and parameter estimation problems demonstrate the feasibility and effectiveness of the proposed algorithm.

关 键 词：遗传算法 插值 模型降阶 参数估计 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]