ConvLSTM-based fault detection and classification in active distribution grids

Abstract

The high penetration rate of distributed energy resources even at the edge of the distribution grids drastically changes the operational status quo, mainly caused due to their high intermittent nature. In order to ensure the uninterrupted electricity supply of the end-consumers, the fast and accurate response to fault events is of critical importance for the operators. This paper proposes a data-driven fault location identification and type classification application based on the ConvLSTM models, which leverages the proliferation of advanced measurement devices in the distribution networks and can locate the exact position of the fault and classify it in eleven different types. These models grasp the spatiotemporal characteristics of the three-phase voltage and current timeseries measurements stemming from the field devices, increasing the visibility of the operators for their networks in real-time conditions. The results conducted through the use of synthetic data showcase the efficacy of this application with accuracy in faulty feeder detection reaching 96% and in the fault type exceeding 88%.