一种自适应性神经元控制器研究

Study on the Self-adapting Neural Network Controller

经典的控制器参数的整定过于依赖控制对象的数学模型,不便在线调整。为了解决这一问题,提出一种自适应性PID神经元控制器。在PID控制原理的基础上,充分利用神经网络用于非线性控制的特性,对神经元赋予了比例、积分和微分的功能。采用可微可导的激励函数,通过反向传播算法,实现了控制器参数的在线调整。仿真实例证明PID神经网络控制器具有较好的自适应性和抗干扰能力。

The parameter of classical controller depends on its mathematics model strongly. It＇s not convenient to ad- just parameters especially online. In order to solve the problem, it brings forward a kind of improved PID neural network self--adapting controller. Based on the PID control principle, one takes full advantage of the neural network characteris- tic, and defines the function of Proportional--Integral--Derivative for the neural cell. Moreover,one introduces the func- tion which can derivative, uses back--propagation algorithm, and realizes the controller parameters adjusting online. At last, the way of simulation experimentation is given, which indicates that the PID neural network controller is of self--a- dapting and anti--jamming.