The Basic Tasks in the Development of the Smart Protection Device

The Smart Protection Devices are intelligent relay protection devices based on machine learning methods. They can delimit complex unconnected parameters areas of various electrical system modes in a multidimensional space. Despite the complexity of the functional load, their implementation does not require an avant-garde computing environment, since all the laborious work on the development of the cognitive abilities of the device's intelligence is carried out even at the development stage. The report notes the importance of thoughtful localization of machine learning methods in relay protection algorithms. It is noted that the solution of the well-known problem of intersection of areas of controlled parameters of various modes of the electrical network is possible if the dimension of the precedent space and the power of the training dataset is sufficient. The success of the Smart Protection Device training largely depends on the appropriate feature engineering, the provision of the necessary capacity and power of the training dataset, and the formation of a compressed training dataset while preserving its information basis. The paper formulates the listed tasks and presents the ways to solve them on the example of the development of the Intelligent Mode Discriminator.