Abstract:With the complex grid structure and ever-changing operation mode of modern power systems, the requirement for on-line transient stability assessment (TSA) is increasing. The artificial intelligence machine learning methods are prominent to deal with this TSA problem. However, with the random fluctuation of load and renewable energy, network topology switching, new scenarios always appear during the system operation, thus the past training samples usually cannot cover all possible operation modes of the system. The pre-established machine learning TSA model may be lack of generalization ability. So in this paper, Support vector machine (SVM) is used to build the pre-training learning machine model, and then an incremental learning (IL) method based on Karush-Kuhn-Tucker (KKT) condition for on-line TSA of power system is proposed. The simulation results in IEEE-39 bus test system show that the proposed incremental SVM method based on KKT condition can effectively update the TSA learning model, and can greatly improve the training efficiency while maintaining the prediction accuracy, which is expected to promote the online application of artificial intelligence and machine learning in TSA.

Incremental Learning Method for Online Transient Stability Assessment of Power System Based on KKT Condition