분산전원이 있는 배전계통에서 딥러닝을 통한 고저항 지락사고 판별

Abstract

This paper proposes a method to detect a high-resistance ground fault in a distribution system with complicated configuration such as installation of a distributed generation. A method to detect high-resistance ground fault accidents by converting the fault current into visual data and applying the CNN technique to this is presented and verified. The data for learning the CNN technique was generated through simulation of the model system. Simulations were performed for data generation by changing the fault resistance, the size, location of faults and amount of distributed power generation, in the case of a high-resistance ground fault and an increase in load in the model system. The generated data was transformed into graphic data by applying Morlett wavelet transform, and then learning was performed by applying CNN. As a result of the learning, high-resistance ground faults were identified with 98.29% accuracy, and a protective algorithm including this result that can respond to high-resistance ground faults occurring in the distribution system was proposed.