热力系统传感器故障检测的动态神经元网络方法

A Dynamic Neuron Network Method for Sensor Failure Detection in a Thermodynamic System

采用动态神经元网络状态观测器的故障检测和诊断方法对非线性时变热力系统状态进行估计。通过对延时神经元网络和动态递归神经元网络在热水锅炉模型上的仿真试验 ,表明 ,基于动态神经元网络状态观测器的热力系统传感器检测和诊断方法是可行的。经过仿真试验对比 ,延时RBF网络具有比较好的推广能力。

A nonlinear time dependent thermodynamic system status is analyzed with a fault detection and diagnostic method being carried out by a dynamic neuron network status observer. Through the simulation tests conducted on a hot water boiler model by the use of a time delay neuron network and dynamic recursive neuron network it is shown that the thermodynamic system sensor detection and diagnostic method based on the above mentioned observer is feasible. After a comparison of the simulation test results one can see that the time delay RBF (radial basis function) network has a relatively good potential for further widespread applications. The use of the above method in an automatic control system of Qingdao Gas Co. has confirmed its reliability.