Publication type: Conference paper
Type of review: Peer review (publication)
Title: Deep learning for fault detection : the path to predictive maintenance of wind turbines
Authors: Ulmer, Markus
Jarlskog, Eskil
Pizza, Gianmarco
Goren Huber, Lilach
et. al: No
Proceedings: Sammelband zu den 6. Energieforschungsgesprächen Disentis
Pages: 24
Pages to: 26
Conference details: Energieforschungsgespräche Disentis 2021, Online, 20.-22. Januar 2021
Issue Date: 20-Jan-2021
Publisher / Ed. Institution: Stiftung Alpines Energieforschungscenter AlpEnForCe
Publisher / Ed. Institution: Disentis
Language: English
Subjects: Predictive maintenance; Smart maintenance; Deep learning; Wind turbine; Renewable energy
Subject (DDC): 006: Special computer methods
620: Engineering
Abstract: We demonstrate the deployment of a novel deep learning algorithm enabling smart maintenance of wind turbines based on 10 minute SCADA data. The newly developed algorithm has the following advantages over existing solutions: • The algorithms are based on the already available 10-minute SCADA data and do not require any additional hardware installations. • The algorithm has been proven to detect various fault types earlier and more accurately than previous methods in the scientific literature. Incipient faults would have been detected weeks or even months prior to known events of a turbine stoppage. • The method is designed to not only detect faults but also specify their localization within the main critical turbine components. • The algorithm does not require a large amount of historical data for its training. Several months of SCADA data are sufficient. This is enabled due to the possibility to adapt the trained algorithm to detect faults on turbines from different wind farms. As such, it is applicable also to newly installed wind turbines and farms. • The method has proven to be robust against parameter variations and to have short training times. As such, it is an optimal practical and scalable solution for high confidence fault detection and diagnostics for wind turbines based on already available 10-minute SCADA data.
URI: https://digitalcollection.zhaw.ch/handle/11475/21401
Fulltext version: Published version
License (according to publishing contract): Licence according to publishing contract
Departement: School of Engineering
Organisational Unit: Institute of Data Analysis and Process Design (IDP)
Published as part of the ZHAW project: Machine Learning Based Fault Detection for Wind Turbines
Appears in collections:Publikationen School of Engineering

Files in This Item:
There are no files associated with this item.


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.