基于局部神经网络模型的过热汽温多模型预测控制的研究

针对锅炉过热汽温的特点，设计串级过热汽温控制系统。主调节器采用基于局部神经网络模型的多模型预测控制，即由主蒸汽流量确定的不同工作点，用神经网络建立局部动态模型，进而建立一组局部神经网络预测控制器，然后通过加权合成的方式获得最终的控制信号；副调节器采用基于多个简单比例控制器的加权合成比例控制。该控制方案融合了神经网络、多模型控制和预测控制的特点。仿真结果表明，在负荷大范围变化的工况下，控制系统仍保持了良好的控制性能，具有较强的鲁棒性。

A STUDY ON MULTIPLE MODELS PREDICTIVE CONTROL FOR SUPERHEATED STEAM TEMPERATURE BASED ON LOCAL NEURAL NETWORK MODELS

Based on the characteristics of superheated steam temperature of a boiler, a new cascade control system is designed. The master regulator adopts multiple models predictive control based on local neural network models, namely, at the different operating points specified according to the flux of main steam, local dynamic models are built with neural networks, based on which a set of local neural network predictive controllers are constructed, then the final control signal is synthesized by weighting. The secondary regulator adopts weighting synthesized proportional control based on several proportional controllers. This control strategy combines the characteristics of neural network, multiple models control, and predictive control. The simulation shows that the control system can maintain better performance when the load changes within a large range and has a strong robustness.