基于RBF神经网络重构模糊PID控制器的研究(英文)

Study on remodeling of fuzzy PID controller based on RBF neural network

尽管模糊PID控制器具有良好的控制品质,但存在计算复杂和实时性差的问题,为了解决这个问题,利用RBF神经网络逼近能力重构模糊PID控制器,由于重构的RBF神经网络的并行计算能力,这简化了计算复杂性并提高实时性.通过选择不同的给定信号,比较模糊PID控制器和重构的RBF神经网络的控制性能,得到两者的控制效果是相当的.说明重构的RBF神经网络可以取代模糊PID控制器,从而减少了计算复杂性,避免维度灾难并改善控制实时性.

Though fuzzy PID controller is characterized by the excellent control quality, there still exists the problems of computation complexity and poor real-time performance. To solve the problems, a known fuzzy PID controller is accurately remodeled based on the universal approximating ability of RBF NN （radial basis function neural network）. With parallel computing ability, the remodeled RBF NN can simplify the computation complexity and enhance the real-time performance of fuzzy PID controller. Given the different reference input, the control performances of fuzzy PID controller and remodeled RBF NN are compared. Results show that the control qualities of the two controllers are extremely similar. Thus, the remodeled RBF NN can replace the fuzzy PID controller to reduce the computation complexity, avoid the curse of dimensionality and improve real-time performance.