研究了基于 BP 神经网络的连续搅拌反应釜 PID 自校正控制,采用梯度下降法调整 PID 参数, BP 神经网络的逼近特性和自适应能力改善了控制效果。通过仿真实例对基于神经网络的 PID 控制器和经典 PID 控制器性能进行比较,结果表明:在相同的暂态响应时间下,前者的超调量更小,而且控制器具有较小的输出量。%The BP neural network-based PID self-tuning control over the continuous stirred tank reactor was investigated,which has gradient descent adopted to regulate PID parameters and the BP neural network’s properties of approximation and adaptive ability employed to improve the control effect.Comparing dynamic performances of neural network PID controller and classic PID controller in simulation examples shows that, under the same transient responding time,the BP neural network-based PID controller has smaller overshoot and output than classic PID controller.
展开▼