Fully Automated Diagnostics of Induction Motor Drives in Offshore Wind Turbine Pitch Systems using Extended Park Vector Transform and Convolutional Neural Network

This is the authors' abstract. We don't add key points for this paper.

Due to their location and related complexities, the offshore wind farms (OWF) have higher downtimes and operation and maintenance (O&M) costs compared to their onshore counterparts. Condition monitoring could help in bringing down the O&M costs of OWFs. The pitch system is one of the components most prone to failure. This paper details an approach for enhanced diagnosis of the electric pitch systems especially focusing on the induction motor drives (IMD) in wind turbines. The proposed method uses an extended Park vector approach (EPVA) in conjunction with a convolutional neural network (CNN) to accurately classify the condition of an IMD and localize the faults. The method is validated on data collected from a laboratory setup. The advantage of the proposed approach is that the condition of the IMD can accurately be classified, and faults localized in operating conditions with varying load and frequency without any additional information on the instantaneous operating speed, frequency, or load on the motor drives. This results in a non-invasive diagnostic approach incurring least additional expenses to implement.