Imperial College London
Alternate

Professor Goran Strbac

Faculty of Engineering, Department of Electrical and Electronic Engineering

Chair in Electrical Energy Systems
 
 
 
//

Contact

 

+44 (0)20 7594 6169g.strbac

 
 
//

Assistant

 

Miss Guler Eroglu +44 (0)20 7594 6170

 
//

Location

 

1101Electrical EngineeringSouth Kensington Campus

//

Summary

 

Publications

Citation

BibTex format

@article{Cremer:2021:10.1016/j.ijepes.2020.106571,
author = {Cremer, JL and Strbac, G},
doi = {10.1016/j.ijepes.2020.106571},
journal = {International Journal of Electrical Power and Energy Systems},
title = {A machine-learning based probabilistic perspective on dynamic security assessment},
url = {http://dx.doi.org/10.1016/j.ijepes.2020.106571},
volume = {128},
year = {2021}
}
Download

RIS format (EndNote, RefMan)

TY  - JOUR
AB  - Probabilistic security assessment and real-time dynamic security assessments(DSA) are promising to better handle the risks of system operations. Thecurrent methodologies of security assessments may require many time-domainsimulations for some stability phenomena that are unpractical in real-time.Supervised machine learning is promising to predict DSA as their predictionsare immediately available. Classifiers are offline trained on operatingconditions and then used in real-time to identify operating conditions that areinsecure. However, the predictions of classifiers can be sometimes wrong andhazardous if an alarm is missed for instance.  A probabilistic output of the classifier is explored in more detail andproposed for probabilistic security assessment. An ensemble classifier istrained and calibrated offline by using Platt scaling to provide accurateprobability estimates of the output. Imbalances in the training database and acost-skewness addressing strategy are proposed for considering that missedalarms are significantly worse than false alarms. Subsequently, risk-minimisedpredictions can be made in real-time operation by applying cost-sensitivelearning. Through case studies on a real data-set of the French transmissiongrid and on the IEEE 6 bus system using static security metrics, it isshowcased how the proposed approach reduces inaccurate predictions and risks.The sensitivity on the likelihood of contingency is studied as well as onexpected outage costs. Finally, the scalability to several contingencies andoperating conditions are showcased.
AU  - Cremer,JL
AU  - Strbac,G
DO  - 10.1016/j.ijepes.2020.106571
PY  - 2021///
SN  - 0142-0615
TI  - A machine-learning based probabilistic perspective on dynamic security assessment
T2  - International Journal of Electrical Power and Energy Systems
UR  - http://dx.doi.org/10.1016/j.ijepes.2020.106571
UR  - http://arxiv.org/abs/1912.07477v2
UR  - http://hdl.handle.net/10044/1/85174
VL  - 128
ER  -
Download