摘要
为了克服单独应用遗传算法或BP算法调节模糊神经网络控制器参数时存在的缺陷,提出了一种将二者复合的方法,即在传统的GA中加入一个BP算子,以充分利用二者各自的优点。而在GA中引进了动态的交叉和变异率,在BP中引进动态的学习率,从而更有效地提高其收敛速度和执行效率。计算机仿真结果表明,用该复合算法调节的模糊神经网络控制器的性能明显优于传统的算法调节的控制器。
In order to overcome the deficiency when GA or BP is singly used to tune the parameters of fuzzy-neural controller, the hybrid tuning of GA and BP was suggested, which adds an BP operator into GA, making full use of respective merits. For fast convergence and good efficiency, the GA implementation incorporate dynamic crossover and mutation variation rates, the BP uses dynamic learning rate. Simulation results expatiate that the proposed hybrid tuning of fuzzy-neural controller offers more encouraging advantages than that of traditional algorithm controller, and has better performance.
出处
《辽宁工学院学报》
2005年第1期7-9,16,共4页
Journal of Liaoning Institute of Technology(Natural Science Edition)
基金
国家自然科学基金资助项目(60274024)
辽宁省教育厅资助项目(202163346)
