神经网络自学习PID控制器的研究

从兼顾人工神经网络控制系统自学习能力和实时性角度出发，通过构建新的人工神经网络自学习PID控制器的结构，控制系统的样本拾取，更新与优化均采用在线学习方式，使控制系统具有较强的自学习能力。在分析了目前广为应用的多层前向神经网络误差反向传播算法(BP)的局限性及原因的基础上，提高了改进的BP算法-自适应动量项BP算法。从而提高了神经网络的收敛速度和收敛精度，并通过实例验证了自学习PID控制器的可行性及改进算法的合理性。图6参4

Research on Neural Network Self-Learning PID Controller

For giving attantion to both real time performance and self-learning ability of control system, a new structure of artificial neural network PID controller that has neural network self-learning ability was put forward. The samples of control system are picked up, updaded and optimized by online learning, which strengthens the self-learning compacity of PID controller. On the basis of analysis the disadwantages and reasons leading from multi layer feed forward neural network and error back propagation BP algorithm, which is widely used now, an advanced algorithm-self-adaptable momentum BP algorithm was put forward, which improved the accuracy and speed of convergence of artificial neural network. At the same time the feasibility of self-learning PID controller and the rationality of advanced algorithm algorithm are validated through examples. Figs 6 and refs 4.