Sequential feature selection for power system event classification utilizing wide-area PMU data

• Published in: Frontiers in energy research 2022-08, Vol.10
• Main Authors: Rafferty, Mark, Liu, Xueqin, Rafferty, John, Xie, Lei, Laverty, David, McLoone, Seán
• Format: Article
• Language: English
• Subjects: dimensionality reduction; event classification; machine learning; PMU data; power system monitoring
• ISSN: 2296-598X; 2296-598X
• DOI: 10.3389/fenrg.2022.957955

The increasing penetration of intermittent, non-synchronous generation has led to a reduction in total power system inertia. Low inertia systems are more sensitive to sudden changes and more susceptible to secondary issues that can result in large-scale events. Due to the short time frames involved, automatic methods for power system event detection and diagnosis are required. Wide-area monitoring systems (WAMS) can provide the data required to detect and diagnose events. However, due to the increasing quantity of data, it is almost impossible for power system operators to manually process raw data. The important information is required to be extracted and presented to system operators for real/near-time decision-making and control. This study demonstrates an approach for the wide-area classification of many power system events. A mixture of sequential feature selection and linear discriminant analysis (LAD) is adopted to reduce the dimensionality of PMU data. Successful event classification is obtained by employing quadratic discriminant analysis (QDA) on wide-area synchronized frequency, phase angle, and voltage measurements. The reliability of the proposed method is evaluated using simulated case studies and benchmarked against other classification methods.