Time series anomaly detection in power electronics signals with recurrent and ConvLSTM autoencoders

Majdi I. Radaideh; Chris Pappas; Jared Walden; Dan Lu; Lasitha Vidyaratne; Thomas Britton; Kishansingh Rajput; Malachi Schram; Sarah Cousineau

doi:10.1016/j.dsp.2022.103704

Time series anomaly detection in power electronics signals with recurrent and ConvLSTM autoencoders

Majdi I. Radaideh, Chris Pappas, Jared Walden, Dan Lu, Lasitha Vidyaratne, Thomas Britton, Kishansingh Rajput, Malachi Schram, Sarah Cousineau

Research output: Contribution to journal › Article › peer-review

24 Scopus citations

Abstract

The anomalies in the high voltage converter modulator (HVCM) remain a major down time for the spallation neutron source facility, that delivers the most intense neutron beam in the world for scientific materials research. In this work, we propose neural network architectures based on Recurrent AutoEncoders (RAE) to detect anomalies ahead of time in the power signals coming from the HVCM. Bi-directional gated recurrent unit, bi-directional long-short term memory (LSTM), and convolutional LSTM (ConvLSTM) are developed, trained, and tested using real experimental signals from the HVCM module. The results show a good performance of the proposed RAE models, achieving precision up to 91%, recall up to 88%, false omission rate as low as 20% (i.e. 80% of the anomalies were detected), and area under the ROC curve up to 0.9. The three RAE models provide very comparable performance, with LSTM showing slightly better performance than GRU and ConvLSTM. The RAE models are benchmarked against other anomaly detection methods, including isolation forest, support vector machine, local outlier factor, feedforward and convolutional autoencoders, and others; showing a better performance. The results of this study demonstrate the promising potential of RAE in anomaly detection for real-world power systems, and for increasing the reliability of the HVCM modules in the spallation neutron source.

Original language	English
Article number	103704
Journal	Digital Signal Processing: A Review Journal
Volume	130
DOIs	https://doi.org/10.1016/j.dsp.2022.103704
State	Published - Oct 2022
Externally published	Yes

Funding

The authors are grateful for support from the Neutron Sciences Directorate at ORNL in the investigation of this work. This work was supported by the DOE Office of Science under grant DE-SC0009915 (Office of Basic Energy Sciences, Scientific User Facilities program). This research used resources of the Compute and Data Environment for Science (CADES) at the Oak Ridge National Laboratory, which is supported by the Office of Science of the U.S. Department of Energy under Contract No. DE-AC05-00OR22725 . The authors would like to thank Willem Blokland from the spallation neutron source for his useful comments and feedback on this work. The authors are grateful for support from the Neutron Sciences Directorate at ORNL in the investigation of this work. This work was supported by the DOE Office of Science under grant DE-SC0009915 (Office of Basic Energy Sciences, Scientific User Facilities program). This research used resources of the Compute and Data Environment for Science (CADES) at the Oak Ridge National Laboratory, which is supported by the Office of Science of the U.S. Department of Energy under Contract No. DE-AC05-00OR22725. The authors would like to thank Willem Blokland from the spallation neutron source for his useful comments and feedback on this work. Notice: This manuscript has been authored by UT-Battelle, LLC, under contract DE-AC05-00OR22725 with the US Department of Energy (DOE). The US government retains and the publisher, by accepting the article for publication, acknowledges that the US government retains a nonexclusive, paid-up, irrevocable, worldwide license to publish or reproduce the published form of this manuscript, or allow others to do so, for US government purposes. DOE will provide public access to these results of federally sponsored research in accordance with the DOE Public Access Plan (http://energy.gov/downloads/doe-public-access-plan).

Funders	Funder number
CADES
DOE Public Access Plan
Data Environment for Science
U.S. Department of Energy	DE-AC05-00OR22725
Office of Science	DE-SC0009915
Basic Energy Sciences
Oak Ridge National Laboratory

Keywords

Anomaly detection
Autoencoders
ConvLSTM
High voltage converter modulator
Signal processing
Spallation neutron source

Access to Document

10.1016/j.dsp.2022.103704

Cite this

@article{e15cbd83343e4677b3ccce0d5f673b19,

title = "Time series anomaly detection in power electronics signals with recurrent and ConvLSTM autoencoders",

abstract = "The anomalies in the high voltage converter modulator (HVCM) remain a major down time for the spallation neutron source facility, that delivers the most intense neutron beam in the world for scientific materials research. In this work, we propose neural network architectures based on Recurrent AutoEncoders (RAE) to detect anomalies ahead of time in the power signals coming from the HVCM. Bi-directional gated recurrent unit, bi-directional long-short term memory (LSTM), and convolutional LSTM (ConvLSTM) are developed, trained, and tested using real experimental signals from the HVCM module. The results show a good performance of the proposed RAE models, achieving precision up to 91%, recall up to 88%, false omission rate as low as 20% (i.e. 80% of the anomalies were detected), and area under the ROC curve up to 0.9. The three RAE models provide very comparable performance, with LSTM showing slightly better performance than GRU and ConvLSTM. The RAE models are benchmarked against other anomaly detection methods, including isolation forest, support vector machine, local outlier factor, feedforward and convolutional autoencoders, and others; showing a better performance. The results of this study demonstrate the promising potential of RAE in anomaly detection for real-world power systems, and for increasing the reliability of the HVCM modules in the spallation neutron source.",

keywords = "Anomaly detection, Autoencoders, ConvLSTM, High voltage converter modulator, Signal processing, Spallation neutron source",

author = "Radaideh, {Majdi I.} and Chris Pappas and Jared Walden and Dan Lu and Lasitha Vidyaratne and Thomas Britton and Kishansingh Rajput and Malachi Schram and Sarah Cousineau",

note = "Publisher Copyright: {\textcopyright} 2022",

year = "2022",

month = oct,

doi = "10.1016/j.dsp.2022.103704",

language = "English",

volume = "130",

journal = "Digital Signal Processing: A Review Journal",

issn = "1051-2004",

publisher = "Elsevier",

}

TY - JOUR

T1 - Time series anomaly detection in power electronics signals with recurrent and ConvLSTM autoencoders

AU - Radaideh, Majdi I.

AU - Pappas, Chris

AU - Walden, Jared

AU - Lu, Dan

AU - Vidyaratne, Lasitha

AU - Britton, Thomas

AU - Rajput, Kishansingh

AU - Schram, Malachi

AU - Cousineau, Sarah

PY - 2022/10

Y1 - 2022/10

N2 - The anomalies in the high voltage converter modulator (HVCM) remain a major down time for the spallation neutron source facility, that delivers the most intense neutron beam in the world for scientific materials research. In this work, we propose neural network architectures based on Recurrent AutoEncoders (RAE) to detect anomalies ahead of time in the power signals coming from the HVCM. Bi-directional gated recurrent unit, bi-directional long-short term memory (LSTM), and convolutional LSTM (ConvLSTM) are developed, trained, and tested using real experimental signals from the HVCM module. The results show a good performance of the proposed RAE models, achieving precision up to 91%, recall up to 88%, false omission rate as low as 20% (i.e. 80% of the anomalies were detected), and area under the ROC curve up to 0.9. The three RAE models provide very comparable performance, with LSTM showing slightly better performance than GRU and ConvLSTM. The RAE models are benchmarked against other anomaly detection methods, including isolation forest, support vector machine, local outlier factor, feedforward and convolutional autoencoders, and others; showing a better performance. The results of this study demonstrate the promising potential of RAE in anomaly detection for real-world power systems, and for increasing the reliability of the HVCM modules in the spallation neutron source.

AB - The anomalies in the high voltage converter modulator (HVCM) remain a major down time for the spallation neutron source facility, that delivers the most intense neutron beam in the world for scientific materials research. In this work, we propose neural network architectures based on Recurrent AutoEncoders (RAE) to detect anomalies ahead of time in the power signals coming from the HVCM. Bi-directional gated recurrent unit, bi-directional long-short term memory (LSTM), and convolutional LSTM (ConvLSTM) are developed, trained, and tested using real experimental signals from the HVCM module. The results show a good performance of the proposed RAE models, achieving precision up to 91%, recall up to 88%, false omission rate as low as 20% (i.e. 80% of the anomalies were detected), and area under the ROC curve up to 0.9. The three RAE models provide very comparable performance, with LSTM showing slightly better performance than GRU and ConvLSTM. The RAE models are benchmarked against other anomaly detection methods, including isolation forest, support vector machine, local outlier factor, feedforward and convolutional autoencoders, and others; showing a better performance. The results of this study demonstrate the promising potential of RAE in anomaly detection for real-world power systems, and for increasing the reliability of the HVCM modules in the spallation neutron source.

KW - Anomaly detection

KW - Autoencoders

KW - ConvLSTM

KW - High voltage converter modulator

KW - Signal processing

KW - Spallation neutron source

UR - http://www.scopus.com/inward/record.url?scp=85138821181&partnerID=8YFLogxK

U2 - 10.1016/j.dsp.2022.103704

DO - 10.1016/j.dsp.2022.103704

M3 - Article

AN - SCOPUS:85138821181

SN - 1051-2004

VL - 130

JO - Digital Signal Processing: A Review Journal

JF - Digital Signal Processing: A Review Journal

M1 - 103704

ER -

Time series anomaly detection in power electronics signals with recurrent and ConvLSTM autoencoders

Abstract

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this