A deep learning based multiple signals fusion architecture for power system fault diagnosis

Abstract

Deep leaning (DL) methods that have emerged in recent years have shown a number of advantages and great potentials in the field of image recognition and feature extraction. However, these methods cannot adapt to multiple signals and do not work well in fault diagnosis for power system, due to increasingly complicated network topology and coupling disturbances. In this paper, a novel fault diagnosis method called multi-dimensional aggregation and decoupling network (MADN) is proposed to tackle these technology and application problems. This novel DL architecture is designed fundamentally with three sequential stages: a multi-dimensional image building (MIB) stage, a feature decoupling mapping (FDM) stage and a system fault state classification (SSC) stage. By imposing these stages, the significant information of multiple signals will be incorporated automatically and the implicit features will be decoupled and mapped in the more advanced space. In addition, the proposed MADN can make a precise analysis for power system even with the transmission loss or unsynchronized multiple signals. The analysis of power system can be accomplished more quickly and precisely using than those conventional approaches. The experimental results acquired by experimental platform confirm the effectiveness and superiority of the proposed method.