Field Winding Short Circuit Fault Signature Analysis in Stray Flux of large Hydrogenerator

The hydroelectric power plant is expected to provide a consistent source of electricity to the grid. If a fault does occur, the maintenance team is usually quick to arrange corrective maintenance to repair the equipment, as a generator shutdown can lead to huge financial losses. This prompt response limits the amount of faulty data. Furthermore, the equipment is made for industrial use and cannot be used for testing, so it is not possible to implement faults to create faulty bench tests. As a result, there is a lack of faulty signals for large hydrogenerators that are necessary to train artificial intelligence algorithms to diagnose faults. This work presents a method to augment and complete a balanced faulty database. The proposed method consists of generating faulty synthetic signals based on in-situ stray flux measurements and fault signatures deduced from simulated signals. To compute external magnetic flux in healthy and faulty cases, a 2D finite element model of a 370M VA salient-pole synchronous generator was created, validated, and used to extract field winding short circuit fault signatures of several severities.