不同负荷低加内部泄漏故障神经网络诊断研究

Neural Network Based Internal Leakage Fault Diagnosis forLow-Pressure Heater System under Multiple Load Points

摘要:

针对600 MW火电机组低压加热器系统内部泄漏故障,提出一种基于神经网络的以最大故障分离度为目标的寻优技术;采用征兆模糊计算方法对典型故障样本进行规整化处理,建立了低加内部泄漏故障诊断的神经网络模型;结合征兆缩放优化技术和神经网络诊断模型,对不同负荷下不同程度低加内部泄漏故障进行实时仿真实验;实验表明上述方法对不同负荷下程度迥异的低加故障均可得到具有高故障分离度的正确诊断结果,可准确诊断低加内部泄漏故障,具有较好的工程实用性。

Abstract:

By taking the internal leakage faults of the low-pressure heater system of a 600 MW supercritical boiler unit as the object studied, a new approach by combining fault symptom zoom optimization technology based on artificial neural network was proposed. The neural network fault diagnosis model is established by dealing with the typical fault samples with symptom fuzzy calculation method. Real time fault diagnosis simulation tests are further realized by combining the ANN fault diagnosis model with fault symptom zoom optimization technology. It is proved that the proposed method can accurately diagnose the low-pressure heaters’ inner leakage faults of different leakage severity degrees under multiple load points with good engineering practicability.