Power System Frequency Dynamic Prediction Based on Physical-Information Driven

Accurately predicting the frequency dynamics of the system after the most probable faults, and thus deploying auxiliary services such as rotating standby in advance, is a key measure to guarantee the stable and continuous operation of the modern power system in the future. Traditional frequency response models generally suffer from problems such as insufficient computational accuracy due to their poor heterogeneity and other characteristics. Additionally, artificial intelligence-based frequency prediction methods may exhibit limited generalization capabilities across various power grid scenarios. Therefore, this paper proposes a physical-information driven power system frequency prediction method, which combines the frequency response physical model and the xgboost frequency prediction information model within the gcforest framework, and constructs a strong learner to improve the frequency prediction accuracy while guaranteeing its generalization ability and interpretability. The effectiveness of the proposed method is verified using the IEEE 39-bus benchmark test system.