基于BP神经网络的PID控制器在气体混合配比系统中的应用

摘要

目的 设计一种应用在气体混合配比系统上的基于BP神经网络的PID控制器.方法 将BP神经网络和PID控制相结合构成一种新型控制器.建立3层神经网络,设定性能指标函数,按照梯度下降法修正网络的权值,实现对PID参数的在线整定.另外进行模型识别构建系统模型,完成控制器的设计.结果 将设计的新型控制器和传统增量式PID控制器分别应用到系统中进行仿真实验,新型控制器的控制超调量和达到稳定状态的时间均优于传统增量式PID控制器.结论 设计的新型控制器具有更好的控制性能,对气体混合配比系统具有重要意义.%Objective To design a new PID controller for gas blenders based on BP neural network.Methods The new controller was developed by combining BP neural network with PID control method.Three neural network layers and a performance index function were formulated to tune PID parameters,which could modify the weight coefficients of the neural network using gradient descent method.Results The simulation results showed that the new controller was superior to the traditional PID controller as evidenced by quicker adjustment and less overshoot.Conclusion The new controller has better performance,and is of greater importance for gas blenders.