Neural network based digital differential relay for synchronous generators

Differential protection is the most common method used by electric utilities for generator stator winding protection. Digital relays, used in differential protection, take one cycle or more after fault inception to issue a trip signal. Minimizing the damage due to stator faults, by using high speed relays, is important. A new high speed neural network based digital differential relay for generator stator winding protection is proposed in this dissertation.; A direct 3-phase model that enables an exact study of synchronous machine performance is presented. This model is capable of simulating normal operation and various kinds of external faults. In this dissertation a new method for simulating internal faults in a synchronous generator, using the direct phase quantities, is developed. The internal faults algorithm is capable of simulating internal single phase to ground faults and internal two phase to ground faults in multi-path and single-path generators.; The relay uses two multi-layer feed-forward neural networks (FNNs). One FNN is used by the fault detector module and the other by the fault classifier module. The fault detector module is used to discriminate between three generator states, namely the normal operation, external fault and internal fault states. In the event of an internal fault the relay issues a trip signal and activates the fault classifier module, which identifies the faulted phase(s). In the case of an external fault the relay acts as a backup relay for the main protection against external faults. Simulation results showing the performance of the protection scheme are presented and indicate that it is fast, robust and reliable. The relay tripping time, for the majority of internal faults, is well within half a cycle.; The laboratory experiments are divided in two parts. The first part provides experimental verification of the developed internal faults algorithm. In the second part implementation and real-time experimental verification of the proposed relay are described.