| Abstract: | This paper presents a label propagation controlled islanding (LPCI) algorithm for large interconnected power systems. The proposed algorithm is suitable for finding an islanding solution quickly when a power system with large disturbances need to avoid uncontrolled system separation and to prevent wide‐area blackouts. By referring to the graph theory for the power system, a similar graph is first introduced according to the branch power flows the power system. Then, the buses of coherent generators are viewed as labeled data, while the left buses are viewed as unlabeled data. As a result, the controlled islanding problem is converted into a semi‐supervised learning problem. At last, a classical semi‐supervised learning algorithm, i.e. label propagation, is applied to solve the controlled islanding problem. In addition, a small‐scale strategy space for power system splitting is provided according to the proposed entropy criterion. Characteristics of the proposed algorithm are presented, as well as feasibility to the active splitting control of power systems. The results on four test cases show its correctness and computationally efficient. © 2015 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc. |
