Development and implementation knowledge-based system for on-line fault diagnosis of power systems

Abstract

This paper demonstrates a novel knowledge-based system for on-line fault diagnosis of power systems. During operation, the software first identifies the blackout islands from the post-fault network topology. It then uses genetic-algorithm (GA)-based optimization for selecting the actual fault section(s) from the blackout islands. The bulk of the knowledge required by the software describes the current status of the power network. To speed up the on-line computation and to keep track of the current network status, two tree-search algorithms have been developed for automatic formation of blackout islands and the corresponding GA fitness function only for these blackout islands. Apart from being fully tested with sample power systems, the software has been put on field tests at the dispatching center of Zhejiang Provincial Electric Network in P.R. China since September 1997. Experience from the field tests has confirmed the effectiveness of the system for on-line fault diagnosis of large-scale power systems.