Abstract
This paper discusses the methodology for fast prediction of power system dynamic behavior. A combination of features that can be obtained from PMU data is proposed, that can improve the prediction time while keeping high accuracy of prediction. Several combinations of features including generator rotor angles, kinetic energy, acceleration and energy margin are used to train and test decision trees for the online identification of unstable generator groups. The predictor importance for trained decision trees is also calculated to highlight in more detail the effect of using different predictors.
Original language | English |
---|---|
Title of host publication | 2017 IEEE Power & Energy Society General Meeting |
Subtitle of host publication | 16-20 July 2017 |
Publisher | IEEE |
Pages | 1-5 |
Number of pages | 5 |
ISBN (Electronic) | 9781538622117, 9781538622131 |
ISBN (Print) | 9781538622124 |
DOIs | |
Publication status | Published - 1 Feb 2018 |
Event | 2017 IEEE PES Society General Meeting: Energizing a More Secure, Resilient & Adaptable Grid - Chicago, United States Duration: 16 Jul 2017 → 20 Jul 2017 http://www.pes-gm.org/2017/ |
Conference
Conference | 2017 IEEE PES Society General Meeting |
---|---|
Abbreviated title | IEEE PES GM 2017 |
Country/Territory | United States |
City | Chicago |
Period | 16/07/17 → 20/07/17 |
Internet address |
Keywords
- decision trees
- dynamic security assessment
- online transient stability