Non-differential protection of generator using fuzzy neural network

This paper presents a new technique for the protection of a generator stator using a fuzzy logic controlled neural network. The technique utilises a specially designed transient detection unit to detect the fault generated transient current signals. A multi-channel filter unit is then applied to the captured signal to extract desired bands of signals. The filter outputs are then fed to a fuzzy neural network for internal and external fault identification. Since the scheme only requires one set of CTs installed at the output terminal of the generator, it leads to a new concept in generator protection, the nondifferential protection principle.