Fault detection & recognition in generator-transformer units by neural network based adaptive protection

Large turbogenerator units are nowadays in majority of cases protected by dedicated relaying systems based upon once for all accomplished relay settings. Such as approach is justified for many of protective relays e.g. overvoltage, generator differential, restricted stator-earth protection. However there are some of protective devices which should change their settings or operating characteristics according to the present power system and/or generating unit conditions. A new concept of a microprocessor-based protective system for large generator-transformer units in which a coordinated coexistence of adaptive and non-adaptive protective functions is ensured, is presented. The system structure, protective strategy and architecture, based on the EBP neural network, are also described.