Single-tank liquid level system is a complex system which is nonlinear, multi-constraint, and delays. It is difficult of traditional PID control algorithm to adaptively adopt optimal control on single-tank liquid level system. An improved neural dynamic programming ( NDP) algorithm which the model network is wavelet neural network is introduced to solve the problem adaptively adopt optimal control on liquid level of single-tank liquid level system. The results show that NDP algorithm has the better robustness, control accuracy and control effects than traditional PID control algorithm.%单容液位控制系统是一个强非线性、多约束、时滞的复杂系统,传统的PID控制算法很难对其进行精确自适应优化控制.介绍了一种改进型的神经动态规划(NDP)算法,其中模型网络用小波神经网络来替代,并针对单容液位控制系统的液位进行自适应优化控制.仿真结果表明,基于神经网络的NDP算法在鲁棒性、控制精度和控制效果都优于传统的PID算法.
展开▼