不同负荷下高加给水系统程度迥异故障诊断的新方法  

A New Approach to Diagnose Variable-degree Faults Under Different Operating Points for High-pressure Feedwater Heater System

作　　者：马良玉[1] 马永光[1] 王兵树[1] 

机构地区：[1]华北电力大学控制科学与工程学院,河北省保定市071003

出　　处：《中国电机工程学报》2010年第2期115-121,共7页Proceedings of the CSEE

摘　　要：在深入研究高加系统故障特征规律的基础上,提出一种结合最大故障分离度为目标的征兆寻优技术和神经网络,适用于不同负荷下高加给水系统程度迥异故障诊断的新方法。仿真诊断结果表明:在系统拓扑结构基本不变的前提下,该方法对不同负荷下程度迥异的高加故障均可得到具有高故障分离度的正确诊断结果,并可通过最优缩放倍率大致判断故障程度。该方法应用于高加故障诊断可大大减少神经网络训练样本的数量,增强故障诊断系统对不同负荷下程度迥异故障的诊断能力,改善系统现场应用效果。With deep investigation on the changing rules of different faults in the feedwater heater system, a new approach by combining fault symptom zoom optimization technology with artificial neural network was proposed for diagnosing faults of variable-degree under different loads for the feedwater heater system. Detailed fault diagnosis tests were carried out on a 300 MW full-scope simulator. The results show that this new method can diagnose faults of variable degree under different operating points accurately if the system topological structure keeps little change. The severity of an actual fault compared with its typical sample can also be approximately estimated through the optimal symptom zoom factor. The method greatly reduces the fault samples used for neural network training and improves the adaptability of a fault diagnosis system to different loads and variable fault degree.

关 键 词：故障诊断 高压给水加热器 神经网络 征兆优化 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]