An Improved AdaBoost-based Ensemble Learning Method for Data-Driven Dynamic Security Assessment of Power Systems

Abstract

In modern power systems, power system dynamic security assessment is a critical task against the risk of blackout. This paper aims to develop reliable recognition models for system real-time dynamic security assessment, where ensemble learning models consisting of extreme learning machine, stochastic configuration networks and random vector functional link have been constructed. The principle for decision making is carried out based on the optimized tradeoff between credibility and accuracy. Moreover, the AdaBoost.RA strategy is afterwards introduced into the modelling process, which allows these critical (instances to be assigned with larger weights and verifies that this proposed methodology could provide more convincing models. This results in a more reliable recognition system for dynamic security assessment.