Robust disturbance classification in power transmission systems with denoising recurrent autoencoders. (December 2022)

• Record Type: Journal Article
• Authors: Kummerow, André; Dirbas, Mohammad; Monsalve, Cristian; Nicolai, Steffen; Bretschneider, Peter
• Is Part Of: Sustainable energy, grids and networks. Volume 32(2022)
• Publication Date: 2022-12
• Subjects: Phasor measurements -- Disturbance classification -- Denoising autoencoders -- Recurrent neural networks -- Data quality; Renewable energy sources -- Periodicals; Smart power grids -- Periodicals; Electric power systems -- Periodicals; 333.794
• DOI: 10.1016/j.segan.2022.100803
• Languages: English
• ISSNs: 2352-4677

Abstract

Abstract: The automated classification of grid disturbances based on phasor measurements is a key technology for the reliable operation of power transmission systems. The predominant use of simulated training data limits the applicability of existing classification approaches due to the missing consideration of measurement errors or data quality issues. To mitigate these shortcomings, this study presents a robust disturbance classification procedure incorporating denoising recurrent autoencoders within a novel two-stage training approach. The developed disturbance classification procedure is evaluated for different noise characteristics and dataset combinations created with an optimization based error model. Experimental results based on a generic power transmission system show superior performance of the proposed two-stage design compared to a conventional, one-stage model training.