基于神经网络逆模型的过热汽温补偿控制研究  

Study on boiler superheated steam temperature compensation control with neural network inverse process models

作　　者：马良玉[1] 史振兴[1] 

机构地区：[1]华北电力大学控制与计算机工程学院,河北保定071003

出　　处：《华北电力大学学报（自然科学版）》2011年第5期70-75,共6页Journal of North China Electric Power University：Natural Science Edition

基　　金：中央高校基本科研业务费专项资金资助项目(09MG21)

摘　　要：为改善大型电站锅炉的过热汽温控制效果,提出一种基于人工神经网络扩展逆系统模型的过热汽温补偿控制方案。首先结合锅炉的结构和运行特性,分析影响过热汽温的主要因素,确定神经网络逆系统扩展模型的输入和输出参数。借助MATLAB神经网络工具箱建立过热汽温系统的动态逆过程模型(IDPM),运用历史运行数据完成神经网络模型的训练。以训练好的逆过程模型为基础,构建2个神经网络逆控制器,分别作为过热器一、二级喷水减温系统原串级PID控制器的输出补偿环节,提供附加控制信号以改善汽温控制效果。上述方案在300 MW燃煤机组全范围仿真系统进行详细的控制仿真试验,结果表明与原串级PID控制系统相比,增加神经网络逆控制器补偿环节后,过热汽温的控制品质得到显著改善。In order to improve the control effect of the superheated steam temperature for a large-scale boiler unit, this paper presents a compensation control scheme based on artificial neural network （ANN） extended inverse process models. With deep understanding of the boiler structure and operating characteristics, the main influence factors of the superheated steam temperature are analyzed, and the input and output variables of the neural network extended inverse process models for the superheater system are determined. Two inverse process models are established with MATLAB neural network toolbox and trained with enough historical operating data. To improve the control effect of the superheated steam temperature, two NN inverse controllers are constructed based on the trained models and used as on-line out-put compensators for the original cascade P1D controllers by providing supplementary signals. Detailed simulation tests are carried out on the full-scope simulator for the given 300 MW power generating unit. It is shown by the tests that the control effect of the superheated steam temperature with NN inverse compensation controllers are significantly improved compared with original cascade PID control scheme.

关 键 词：人工神经网络 逆过程模型 过热汽温控制 PID补偿 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]