基于神经网络实用稳定性理论提高FNN容错性的方法及其在电力系统中的应用  

A NEW APPROACH OF ENHANCING FNN FAULT-TOLERANCE PERFORMANCE BASED ON PRATICAL STABILITY THEORY OF NN AND ITS APPLICATION IN POWER SYSTEMS

作　　者：姜惠兰[1] 孙雅明[1] 

机构地区：[1]天津大学电气自动化与能源工程学院,天津300072

出　　处：《中国电机工程学报》2003年第5期29-34,共6页Proceedings of the CSEE

基　　金：国家自然科学基金项目(59877016)~~

摘　　要：在神经网络(NN)实用稳定性理论和对前馈神经网络(feed-forward neural network, FNN)容错性能(fault- tolerance performance, FTP)理论分析的基础上,提出了一种提高FNN容错性的实用方法,该法建立了基于FNN学习与模糊处理相结合的FNN模型重构机理,以改变FNN“伪吸引子”和“伪吸引域”来地提高FNN的FTP。文中以电力输电线路故障诊断为例,通过仿真测试证明了所提方法的可行性,并能有效地提高FNN的FTP,所研究的方法为基于FNN实时信息处理系统的实际应用提供了重要的保证。On the basis of practical stability theory of neural networks (NN) and theoretic analysis for the fault-tolerance performance (FTP) of feed-forward neural networks (FNN), this paper presented a practical method for enhancing FTP of FNN, built the mechanism of reconstructing FNN model based on combining NN learning and fuzzy processing to change the 'fake attractor' and 'fake attraction domain' of FNN, so that FTP of FNN is enhanced greatly. Power system transmission line fault diagnosis problem is used as an example in this paper, simulation results are shown that the practical method presented is feasible and can enhance the FTP of FNN effectively. The researched method provides important guarantee for FNN based practical application in the real time information process system.

关 键 词：电力系统 神经网络 实用稳定性理论 FNN 容错性 输电线路 故障诊断 

分 类 号：TM712[电气工程—电力系统及自动化] TM726