A Deep Learning Algorithm for Fault Imbalance Diagnostics in Wind Turbine Rotors Using Electrical Generator Signals

This work attempts to answer the following research question: can fault imbalance diagnostics in wind turbine rotors using electrical generator signals be improved with deep learning methods? For this purpose, a framework using TurbSim/FAST/Simulink was developed to simulate electric signals generated from a 1.5 MW WT for different wind inflow scenarios and blade imbalances parameters. The simulations were used to train, validate and test a deep learning algorithm, and the fault classification metrics were
obtained.  It is possible to detect amplitude and frequency modulation from the current spectrum due to the imbalance, which spread the harmonics components in sidebands around its nominal frequency.