A deep long Short-Term memory based scheme for Auto-Reclosing of power transmission lines

Abstract

The operating speed of adaptive single-phase auto-reclosure (ASPAR) is of great importance to maintain power systems stability in high-voltage power transmission lines (PTLs). This paper proposes a two-step protection scheme using the long short-term memory (LSTM) network to enhance the ASPAR performance. The discrimination between transient and permanent faults is made by an LSTM with high accuracy in the first step. If transient faults are detected in the second step, another LSTM is applied to predict the secondary arc extinction time (SAET). To this end, the second LSTM accurately foresees the voltage waveform of the faulty phase a half-power cycle earlier, and predicts the SAET in order to get a successful reclosing. The results obtained from extensive simulations using EMTP-RV and MATLAB software environments indicate that the presented protection scheme outperforms other ones in terms of fault type classification, achieving an F-measure value of 98.90%. Moreover, the results verify that the LSTM can accurately estimate the voltage waveform and the SAET, that ensures a successful reclosing of the faulty phase.