ПРИМЕНЕНИЕ СВЕРТОЧНЫХ РЕКУРРЕНТНЫХ НЕЙРОННЫХ СЕТЕЙ С МЕХАНИЗМОМ ВНИМАНИЯ ДЛЯ ОБНАРУЖЕНИЯ АНОМАЛИЙ В РАБОТЕ ОБОРУДОВАНИЯ АЭС

Abstract

We present the results of numerical experiments with the MSCRED algorithm, based on a convolutional recurrent autoencoder with an attention mechanism, to detect anomalies in the operation of NPP equipment. The results are superior to some of the well-known algo-rithms used to monitor equipment health.