Real- Time Detection of Critical Generators in Power Systems: A Deep Learning HCP Approach

Abstract

Assessment of power system transient stability is critical for a reliable continuous operation and to ensure none of the working generating units in the system go out of synchronism. Therefore, a fast and accurate surveillance of transient stability in power systems is necessary. This paper proposes a deep learning neural network framework that captures the phasor measurement unit (PMU) measurements and monitor the system transient stability in real-time. The proposed framework utilizes the convolutional neural network (CNN) with hypotheses CNN pooling (HCP) to identify the state of the system and detect the set of critical generators. The suggested CNN module for stability estimation and the robust HCP module for detecting critical generators through multi-label classification are tested on the IEEE lIS-bus test system, where different types of faults at different locations and under varying system load conditions are simulated. The test results verified that our proposed framework is fast and accurate, thereby a viable approach for online system monitoring applications.