A NEW APPROACH OF ENHANCING FNN FAULT-TOLERANCE PERFORMANCE BASED ON PRATICAL STABILITY THEORY OF NN AND ITS APPLICATION IN POWER SYSTEMS
Sun Yaming · Proceedings of the CSEE · 2003
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.