Data-Driven Dynamic Security Assessment and Control of Power Systems: An Online Sequential Learning Method

Abstract

Intelligent systems (IS) have gained popularity in facilitating very fast dynamic security assessment (DSA). However, conventional IS methods are limited in their ability to be updated with current system operation conditions online due to the excessive training time and complex parameters tuning required for updates. In this paper, an online sequential extreme learning machine (ELM) based method is proposed to enable efficient real-time DSA and online model updating. To enhance the performance of ELMs, feature selection using single-feature estimation is conducted and the results are utilized to design generation shifting as a preventive control. The proposed methods are examined based on the New England 39-bus test system and compared with popular IS methods. The simulation results show that the ELM-based DSA method possesses significant superior computation speed while high, competitive accuracy is maintained. The derived generation shifting is also valid to restore system security.