基于粒子群优化的汽温系统神经网络自整定PID控制  

Self-tuning of neural networks PID control for main steam temperature system based on particle swarm optimization

作　　者：刘国宏[1] 倪桂杰[2] 孙明[3] 翟永杰[4] 

机构地区：[1]河南省电力公司培训中心,河南郑州450052 [2]郑州电力高等专科学校动力系,河南郑州450052 [3]华北电力大学科技学院,河北保定071003 [4]华北电力大学控制科学与工程学院,河北保定071003

出　　处：《华北电力大学学报（自然科学版）》2009年第1期44-49,共6页Journal of North China Electric Power University：Natural Science Edition

摘　　要：为了使神经网络PID取得更好的控制性能,采用改进的粒子群算法对神经网络的权值进行优化,通过对具有严重参数不确定性、多扰动以及大迟延的电厂主蒸汽温度被控对象进行的仿真研究结果表明,所提出的嵌入混沌序列的小生境粒子群算法可以避免局部极小,具有全局优化的能力,对神经网络PID的权值优化是成功和有效的,使得具有多模型特性的汽温控制系统在不同的负荷下均获得很好的调节品质。In order to get a better control performance for neural network PID controller, the improved PSO is put forward to optimize the initial weights of neural network PID controller. Simulation is proceeded for the steam temperature system in a power plant under such a control which has a severe uncertainty of parameters and multi - disturbance, as well as a large timedelay. The results show that the presented algorithm of embedded chaotic sequence niche PSO can avoid premature effectively and has powerful global optimizing ability and is also successful and effective for optimizing the weights of neural network PID controller. The simulation results also demonstrate that the multi - model main steam temperature control system has an excellent regulation performance under different steam loads.

关 键 词：比例积分微分控制 神经网络 粒子群算法 自整定 主汽温控制系统 

分 类 号：TM621.7[电气工程—电力系统及自动化]