时滞系统传感器故障检测的神经网络方法  

Sensor Failure Detection of Time-lag System Based on Neural Network

作　　者：李明[1] 徐向东[1] 

机构地区：[1]清华大学

出　　处：《电站系统工程》2001年第6期371-374,共4页Power System Engineering

基　　金：国家攀登B项目(85-35)

摘　　要：大型热力系统的控制系统必须能够检测传感器故障,并采取相应的措施,保证控制过程的顺利进行。针对热力系统这类时滞系统,提出一种基于时间序列神经网络的故障检测新方法,神经网络的训练采用Powell方法,其收敛速度快、过程稳定。本方法具有在线学习、可诊断多个传感器故障等优点,对锅炉实际试验结果表明本方法行之有效。To keep the capability of control process, the control system of thermodynamic system must be able to detect sensor failure and take related measures. A sensor failure detecting method of time-lag system based on time series neural network is proposed. Neural network is trained with Powell method, which is rapid and stable. This method has abilities of on-line working and identifying multiple sensor failure. The practical experimental results of boiler system show that this method is very useful.

关 键 词：神经网络 故障检测 传感器 时滞系统 热力系统 单层感知器 

分 类 号：TK284.1[动力工程及工程热物理—动力机械及工程] TP183[自动化与计算机技术—控制理论与控制工程]