基于神经网络的主汽温控制系统

针对火电厂主汽温被控对象的大迟延、模型不确定性，设计了基于神经网络的主汽温控制系统。系统结构为串级系统。内回路采用常规比例调节器，外回路采用带辨识器的单神经元ＰＩＤ控制器。辨识器为３层ＢＰ网络结构，以广义δ规则为学习规则。控制器学习算法为有监督的Ｈｅｂｂ算法，教师信号由系统定值和辨识器输出构成。对系统在多种工况下的仿真结果表明，所设计的系统在控制品质、鲁棒性方面明显优于主汽温常规ＰＩＤ控制系统。

Main Stream Temperature Control System Based on Neural Network

In order to overcome the large delay and the uncertainty of the main-stream temperature object in fossil-fired power station, a control system based on neural network is proposed. The inner loop uses a general proportion adjuster and the outer loop uses a single neuron PID controller with identification implement. The identification implement is a three layer BP network based on δ-rule. Hebb study arithmetic is adopted. The system desired value and the controller output constitute the teacher＇s signals. Through simulation in various situations, it is validated that the control quality and the robustness of this control system apparently are superiors to the general PID system.