Feature selection for intelligent stability assessment of power systems

Abstract

Intelligent System (IS) strategy have been widely adopted in facilitating real-time stability assessment of modern power systems. Typically, the development of such an IS comprises four steps: database generation, input/output specification, knowledge extraction, and validation. While current research efforts are mainly focusing on the third step, much less attention is paid to the second step, which actually is very important to the performance of the IS. It is also noted that only a limited number of feature selection methods were used in the literature. In this paper, systematical study, review and comparisons of feature selection technique are conducted, through which better understanding and decision-support for incorporating a feature selection technique in an IS can be obtained. Additionally, this paper also introduces two alternative feature selection methods, which can be applied in this area. Numerical simulations are also conducted to verify the studied approaches. The immediate value of this paper is that it provides straightforward guideline and reference to researchers who intend to integrate effective feature selection in developing an IS-based stability assessment model.