Sparse Representation-based Classification of Geomagnetically Induced Currents

Geomagnetically Induced Currents (GIC) are the significant effects of the Geomagnetic Disturbances (GMDs) on power systems. GICs typically appear in the form of DC components in the current wave forms of high voltage transmission lines and may lead to transformer saturation, so-called DC saturation. Such saturation scenarios, if experienced, can result in severe damages to the transformer core and significantly increase the system-wide risk of major blackouts. It calls for developing detection and classification mechanisms for GICs in power systems so they can be prevented or interrupted before emerging as a threat. A major challenge associated with GIC detection is the presence of similar events, resulted from faults and other distortions, such as AC saturation caused by harmonics. The current signals recorded by current transformers can be analyzed to classify these events. In this paper, the time-frequency S-Transform is integrated with a sparsely-enhanced version of the collaborative representation-based classification to implement a fast, reliable, and adaptive GIC events classification approach. Unlike usual techniques, the proposed mechanism does not need any training procedure while, due to its linear formulation, acts inherently fast and is adaptable to recognize the challenging scenarios of combined events.