An Intelligent and Time Adaptive Model for Symmetrical and Asymmetrical Faults during Power Swing

Abstract

The unwanted operation of the distance relay during the power swing conditions can lead to increased disturbances and exacerbate of the power grid. Therefore, the rapid and accurate detection of the power swing and blocking of the distance relay after a power swing is necessary to maintain the security and reliability of the power grid. On the other hand, for a fault condition during the power swing, in order to maintain the dependability index of the protective system, it is necessary to identify the fault. This paper presents an intelligent and time adaptive algorithm for detecting symmetric and asymmetric faults in series compensated transmission lines through the long short-term memory (LSTM) recurrent neural network. This method uses three-phase currents in the distance relay point as input. In order to investigate the proposed algorithm, the reference power system for transmission-line relay testing introduced by the IEEE Power System Relaying Committee (PSRC), was considered. Different fault types in different conditions such as fault location, fault resistance, load angle and fault inception time were modeled and simulated in PSCAD software. The results show that the proposed method has an average response time (ART) and an average accuracy (AA) of 0.1004 ms and 99.04%, respectively.

Keywords

• distance relay
• recurrent neural network
• long short term memory
• fault identification
• power swing