基于改进遗传算法寻优的神经网络PID控制及应用

根据神经网络PID控制器初值的选取影响系统控制性能的特点,提出了一种基于改进遗传算法寻优的神经网络PID控制方法.即先利用遗传算法对PID控制器参数离线寻优,将求出的参数值作为控制器的比例、积分、微分系数的初值,再进行神经网络PID控制.对一类液位过程的实时控制结果表明采用本方法的控制系统具有较好的稳定性和鲁棒性.

Neural Network PID Control Method and Its Application Based on Parameters Optimization via Modified Genetic Algorithm

Considering the fact that the initial value of neural network PID controller affected the system performance, a neural network PID control method based on parameters optimization via modified GA is presented, in which parameters of the PID controller are optimized off-line using GA, the result of optimization is regarded as initial value of proportion, integral and differential. The neural network PID control system is devised according to the above method. The experimental results to level process indicate that the proposed approach possesses satisfactory stability and robustness.