Fault Diagnosis in Wind Energy Management System using Extreme Learning Machine: A Systematic Review

Abstract

Fault diagnosis is increasingly important given the worldwide demand on wind energy as one of the promising renewable energy sources. This systematic review aimed to summarize the fault diagnosis using Extreme Learning Machine (ELM) on wind energy. Firstly, two databases (i.e. Engineering Village (EV) and IEEE Explore were searched to identify relevant articles, using three important keywords, including Extreme Learning Machine/ELM, fault and wind. Of the 14 included studies, only eight studies mentioned the use of sensor to collect vibration signals as the fault data. Sensors were commonly installed at four places (gearbox, generator, bearing, or rotor) in the included studies. Only nine studies used either single or fusion feature extractions for the fault data. Two types of ELM (i.e. single/multi-layered or hybrid-ELM) were identified to diagnose fault. In general, studies showed the superiority of the application of ELM in producing accuracy results in fault diagnosis of WT, compared to other algorithms. Future studies should incorporate the use of real-world data, and improve on the reporting on the methodological components of the study, to better inform on the usefulness of ELM for fault diagnosis in real-world wind energy settings.