Comparison of new anomaly detection technique for wind turbine condition monitoring using gearbox SCADA data

McKinnon, Conor and Carroll, James and McDonald, Alasdair and Koukoura, Sofia and Infield, David and Soraghan, Conaill (2020) Comparison of new anomaly detection technique for wind turbine condition monitoring using gearbox SCADA data. Energies, 13 (19). 5152. ISSN 1996-1073

[img]
Preview
Text (McKinnon-etal-Energies-2020-Comparison-of-new-anomaly-detection-technique-for-wind-turbine)
McKinnon_etal_Energies_2020_Comparison_of_new_anomaly_detection_technique_for_wind_turbine.pdf
Final Published Version
License: Creative Commons Attribution 4.0 logo

Download (1MB)| Preview

    Abstract

    Anomaly detection for Wind Turbine condition monitoring is an active area of research within the Wind Energy Operations and Maintenance (O&M) community. In this paper three models were compared for multi-megawatt operational wind turbine SCADA data. The models used for comparison were One-Class Support Vector Machine (OCSVM), Isolation Forest (IF), and Non-Linear Auto-Regression Neural Network with Exogenous Inputs (NARX). Each of these were compared for the same fault, and tested under various different data configurations. This paper presents a new method of condition monitoring that only requires two months of data per turbine. These months being separated by a year, one being healthy and the other unhealthy. The number of anomalies is compared, with a greater number in the unhealthy month being considered correct. It was found that for accuracy, run-time, and ease of implementation, IF had the best performance for both training regimes. It was also found that NARX did not perform well in the previously mentioned areas. Both the OCSVM and IF models were shown to be appropriate for the method presented. IF is shown to be more suitable due to the reduced run-time, and better accuracies.