Time Series Anomaly Detection in Power Electronics Signals with Recurrent and Convlstm Autoencoders

24 Pages Posted: 29 Mar 2022

See all articles by Majdi I. Radaideh

Majdi I. Radaideh

Government of the United States of America - Oak Ridge National Laboratory

Chris Pappas

Government of the United States of America - Oak Ridge National Laboratory

Jared Walden

Government of the United States of America - Oak Ridge National Laboratory

Dan Lu

Government of the United States of America - Oak Ridge National Laboratory

Lasitha Vidyaratne

Thomas Jefferson National Accelerator Facility, Newport News

Thomas Britton

affiliation not provided to SSRN

Kishansingh Rajput

Thomas Jefferson National Accelerator Facility, Newport News

Malachi Schram

Thomas Jefferson National Accelerator Facility, Newport News

Sarah Cousineau

Government of the United States of America - Oak Ridge National Laboratory

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, while each model seems to slightly excel over the other two depending on the selected false positive threshold. The RAE models are benchmarked against other anomaly detection methods, including isolation forests, support vector machine, local outlier factor, and feedforward autoencoders; 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 for the spallation neutron source.

Keywords: anomaly detection, Autoencoders, ConvLSTM, High Voltage Converter Modulator, Spallation Neutron Source

Suggested Citation

Radaideh, Majdi I. and Pappas, Chris and Walden, Jared and Lu, Dan and Vidyaratne, Lasitha and Britton, Thomas and Rajput, Kishansingh and Schram, Malachi and Cousineau, Sarah, Time Series Anomaly Detection in Power Electronics Signals with Recurrent and Convlstm Autoencoders. Available at SSRN: https://ssrn.com/abstract=4069225 or http://dx.doi.org/10.2139/ssrn.4069225

Majdi I. Radaideh (Contact Author)

Government of the United States of America - Oak Ridge National Laboratory ( email )

1 Bethel Valley Road, P.O. Box 2008, Mail Stop 608
Room B-106, Building 5700
Oak Ridge, TN 37831
United States

Chris Pappas

Government of the United States of America - Oak Ridge National Laboratory ( email )

1 Bethel Valley Road, P.O. Box 2008, Mail Stop 608
Room B-106, Building 5700
Oak Ridge, TN 37831
United States

Jared Walden

Government of the United States of America - Oak Ridge National Laboratory ( email )

1 Bethel Valley Road, P.O. Box 2008, Mail Stop 608
Room B-106, Building 5700
Oak Ridge, TN 37831
United States

Dan Lu

Government of the United States of America - Oak Ridge National Laboratory ( email )

1 Bethel Valley Road, P.O. Box 2008, Mail Stop 608
Room B-106, Building 5700
Oak Ridge, TN 37831
United States

Lasitha Vidyaratne

Thomas Jefferson National Accelerator Facility, Newport News ( email )

Thomas Britton

affiliation not provided to SSRN ( email )

No Address Available

Kishansingh Rajput

Thomas Jefferson National Accelerator Facility, Newport News ( email )

Malachi Schram

Thomas Jefferson National Accelerator Facility, Newport News ( email )

Sarah Cousineau

Government of the United States of America - Oak Ridge National Laboratory ( email )

1 Bethel Valley Road, P.O. Box 2008, Mail Stop 608
Room B-106, Building 5700
Oak Ridge, TN 37831
United States

Do you have negative results from your research you’d like to share?

Paper statistics

Downloads
115
Abstract Views
553
Rank
407,904
PlumX Metrics