Feature selection method for future-state power grid operation mode based on semi-supervised learning
Jiachen Li, Baozhu Liu, Shuang Zhang, Jing Ma, Jiecong Wang, Jun Ma, Chen Shi, Bing Wang · Journal of Physics Conference Series · 2024
Abstract With the rapid development of grid interconnection and the large-scale integration of renewable energy, the number of key features in the power system is very large. To proactively monitor the status of future power grids, it is necessary to extract key features that comprehensively reflect their operation modes from the operational data. This paper presents a feature selection method for power grid operation modes based on semi-supervised learning. The method is designed to consider the semi-supervised dataset of the future-state power grid. The efficient semi-supervised feature selection method based on the eigenspace model and manifold regularization can be used to quickly and accurately obtain the key features reflecting the way the grid is operated, and its validity is verified at the IEEE39-bus system.