基于模糊神经网络的动态非线性系统辨识研究

Research on Identification of Dynamic Nonlinear System Based on Fuzzy Neural Network

针对静态模糊神经网络对动态系统辨识精度低的特点,在5层静态模糊神经网络基础上进行了优化和改进,形成了可将暂态信息记忆于网络的动态回归层的动态模糊神经网络,来提高对动态系统的辨识能力。同时给出了参数的动态自适应学习算法。通过仿真实验,证明提出的动态模糊神经网络对动态非线性系统的辨识,可以取得较好的辨识精度,较快的网络收敛速度,为动态非线性系统的辨识提供新的思路。

Aiming at improving the identification ability of dynamic system, a new dynamic fuzzy-neural network was proposed based on the traditional static fuzzy neural networks （TFNN）. The new fuzzy-neural network adopted a recurrent layer to memory the temporal information. The inference algorithm and dynamic adaptive learning algorithm were deduced to demonstration the system. It is demonstrated that proposed DFNN is better than TFNN used in dynamic nonlinear system identification by simulation.