Application of hierarchical neural networks to fault diagnosis of power systems

Abstract

The application of a hierarchical neural network to fault diagnosis of power systems is presented. In this paper, the neural network for fault diagnosis is hierarchically formed of three neural network classes. The whole power system is divided into subsystems. Class II neural networks control each subsystem, and Class III neural networks connect the subsystems. Every section of the power system is classified into one of the typical sections where the same diagnosis rules can be applied, such as line-section, bus-section and transformer-section. Only one neural network (Class I) is made for each typical section. The error back-propagation algorithm is used for training. The simulation results show that the proposed method reduces the number of training patterns, the necessary memory for storing the states of the neural networks, and the number of multiplications during training, compared with the conventional methods which use one neural network for the whole power system. This paper shows the possibilities of applying the proposed method to real large-scale power systems.