Historical Data Based Monitoring of Hydro Generator Using Machine Learning

This paper discusses the health monitoring of synchronous generators used in hydropower plants. In recent years, maintenance of generating stations has shifted its focus from preventive maintenance to predictive maintenance. Machine prognosis is a significant part of condition-based maintenance. It intends to monitor and track the time evolution of a fault, so that maintenance can be performed, or the task can be terminated to avoid a catastrophic failure. This paper focuses on the machine learning model for health detection of stator winding of synchronous generator by using stator terminal voltage and stator winding current as input and stator winding temperature as output. More than five years of real-time data of a synchronous generator of Sardikhola hydropower plant in Nepal are collected to predict and present the Adaptive Neuro-Fuzzy Interference System (ANFIS) model. This model predicts faulty data range and healthy data range of stator winding temperature corresponding to stator terminal voltage and current.