A Microgrid Protection Scheme with Conventional Relay Measurements

The integration of renewable energy and emergence of microgrid are reshaping the power industry unprecedentedly. Active distribution systems interconnecting distributed energy resources (DERs), including generation and storage, along their feeders create new challenges in terms of power system protection. Among other issues, most DERs are interfaced with the grid by means of power electronic converters, which have low short-circuit contributions. This makes it difficult to detect certain types of faults and to selectively isolate the faulty sections. The proposed solution utilizes data for applied intelligence in microgrid protection system design and implementation to ensure reliable protection under different microgrid configurations and operating conditions. It shows that detection features acquired from conventional protective relay measurements are sufficient for machine-learning-based prediction models in microgrid fault detection. A detection feature subset in each microgrid operation mode is suggested, meanwhile, the performance of six types of learning algorithms are evaluated in this paper.