基于RST改进NN模型的高压输电线系统故障诊断

A RST-based Improved NN Model for Fault Diagnosis of High Voltage Transmission Line System

为了克服实时诊断信息在形成和传递过程中的畸变而导致故障诊断结果的错误 ,在基于粗糙集理论 (RoughSetTheory,简称RST)的高压输电线系统故障诊断模型的研究基础上 ,充分利用神经网络(NeuralNetworks,简称NN)的泛化能力和粗糙集理论强大的定性分析能力 ,构造了RST与NN相结合的故障诊断模型 .首先利用RST从诊断样本中提取领域知识 ,然后利用所提取的诊断对象知识属性形成诊断NN的初始结构 ,进而增强诊断NN的智能性和容错性 .通过高压输电线系统故障诊断的仿真结果比较 ,证明了该模型的有效性和通用性 .该模型即使在诊断信息不完整的情况下 ,也具有高的诊断容错性能 。

To overcome the mis-diagnosis of fault caused by t he distortion of real-time diagnosis information during the generation and tran sfer processes, on the basis of the research into the RST (Rough Set Theory)-ba sed fault diagnosis model of high voltage transmission line system and by utiliz ing the generalization ability of neural network (NN) and the great qualitative analysis ability of RST, the fault diagnosis model with the combination of RST a nd NN was constructed. In this approach, RST was used to extract knowledge regio n data set from diagnosis samples, and the basic structure of NN was built on th e basis of diagnosis knowledge attribute. Thus the intelligence and fault tolera nce performance of diagnosis NN system were improved. The validity and commonali ty of the proposed model were proved by the comparison of simulation results of the fault diagnosis system. The proposed model has excellent fault tolerance per formance even though the diagnosis information is not complete, so it is of impo rtant practical value in the real-time fault diagnosis of electric power system .