A Review on Dynamic Security Assessment of Power System Based on Machine Learning

The massive integration of renewable energy increases the uncertainty of the system operating condition, leading to the frequent operating status of the power system at the stability boundary. Multiple dynamic security risks of power systems are prominent. The dynamic security assessment (DSA) based on machine learning is an emerging method to evaluate various dynamic security risks (e.g., power angle, frequency and voltage). It can directly obtain the mapping relationship between input features and output, which has the advantage of high assessment efficiency. With the development of artificial intelligence (AI) technology, various cutting-edge machine learning methods have been explored and applied to the DSA. This paper divides the DSA into two categories: the stability-domain DSA based on dynamic features and the security-domain DSA based on steady-state features, and reviews the current research status from the following four aspects: feature selection, model training, model updating, and online assessment. Moreover, five major challenges are summarized.