Anomaly detection for parabolic trough power plants with density-based outlierness
Josua Braun, Alex Brenner, Gabriele Gühring · AIP conference proceedings · 2023
Defects and faults in parabolic trough power plants often lead to lower energy production.The automated detection of such anomalies could reduce downtimes and increase efficiency.A machine learning anomaly detection approach which is based on processing data recorded during regular plant operation is examined.The measured data are high-dimensional data in a spatio-temporal context.Existing data is preprocessed and useful features are selected and extracted.These features build the input for the anomaly detection.The basic idea of the method is the consideration of multivariant time series on loop level, which are segmented and further quantified in its outlierness.For the segmentation of the time series two different segmentation methods, a window-sliding and a periodic method, are used and compared.For the quantification of the outlierness the local outlier factor is used in both cases.The results of both variants of anomaly detection show differences in the time points of detection, but globally they show a similar behavior and detection.