An Unsupervised Learning Approach to Condition Assessment on a Wound-Rotor Induction Generator

Accurate online diagnosis of incipient faults and condition assessment on generators is especially challenging to automate through supervised learning techniques, because data imbalance. Fault-condition training test are either not available or experimentally emulated, therefore do precisely account for all the eventualities nuances practical operating conditions. Thus, it would be more convenient harness ability unsupervised in these applications. An investigation into use as a means recognizing fault patterns assessing wound-rotor induction generator presented. High-dimension clustering performed using stator rotor current voltage signatures measured under healthy varying conditions an experimental generator. analysis validation results carried out determine performance suitability technique. Results indicate that presented technique can accurately distinguish different investigated manner. This research will contribute ongoing development frameworks data-driven diagnostic systems WRIGs similar electrical machines.