Field failures of wind turbine main bearings cause unwanted downtime and significant maintenance costs. Currently, this industry seeks to increase its reliability, for which condition monitoring and predictive maintenance systems have been adopted. In most industrial wind farms, the integrated Supervisory Control and Data Acquisition (SCADA) system provides data that is stored averaged every 10 minutes that can be used to quantify the health of a wind turbine (WT). This research presents a framework for the analysis of data collected from the SCADA system of an operating wind farm, aiming to early detect the main bearing failure using a Long-Short-Term Memory (LSTM) neural network. For prediction, SCADA variables of the temperature of turbine components near the main bearing, rotor speed, ambient temperature, and generated power are taken into account. The results show that the proposed methodology can detect the target failure up to 4 months in advance of the fatal breakdown. The results obtained confirm the applicability of the proposed model in real scenarios that can help the operator with enough time to make more informed maintenance decisions.

Predictive maintenance of wind turbine's main bearing using wind farm SCADA data and LSTM neural networks

Castellani F.;
2023

Abstract

Field failures of wind turbine main bearings cause unwanted downtime and significant maintenance costs. Currently, this industry seeks to increase its reliability, for which condition monitoring and predictive maintenance systems have been adopted. In most industrial wind farms, the integrated Supervisory Control and Data Acquisition (SCADA) system provides data that is stored averaged every 10 minutes that can be used to quantify the health of a wind turbine (WT). This research presents a framework for the analysis of data collected from the SCADA system of an operating wind farm, aiming to early detect the main bearing failure using a Long-Short-Term Memory (LSTM) neural network. For prediction, SCADA variables of the temperature of turbine components near the main bearing, rotor speed, ambient temperature, and generated power are taken into account. The results show that the proposed methodology can detect the target failure up to 4 months in advance of the fatal breakdown. The results obtained confirm the applicability of the proposed model in real scenarios that can help the operator with enough time to make more informed maintenance decisions.
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11391/1569557
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 2
  • ???jsp.display-item.citation.isi??? 1
social impact