摘要
针对具有多指标的被控对象,提出一种基于神经网络的控制器自设计方法。算法利用并行遗传算法按照被控对象各项性能指标进化神经网络控制器,在遗传算法每代结束时利用适应性权重法根据各项指标数据计算综合适应度值,选择综合适应度最佳个体进行遗传操作,从而获得综合性能指标最佳的控制器。将算法应用于异步电机矢量控制系统的速度控制器自设计中,通过仿真实验验证了本方法的有效性。
This paper proposes a method of designing controller for muhi-performance-index system based on neural networks reinforcement learning. It uses parallel genetic algorithm to evolve the neuroeontroller according to the different performance index. At the end of each generation during the evolution, it uses adaptive weight approach to calculate the synthetical fitness of the system and selects elitists to execute evolutionary operation. With the synthetical fitness function, it will design the optimal neurocontroller. Through applying the method to design a speed controller for an asynchronous drive system, the simulation results validate the feasibility of the proposed method.
出处
《电子技术应用》
北大核心
2010年第10期76-79,共4页
Application of Electronic Technique
关键词
遗传算法
神经网络
矢量控制
控制器自设计
genetic algorithm
neural network
vector control
controller self-design
