Switch State Identification in Distribution Network Based on Edge Computing

With the increasing penetration of renewable energy in the distribution network, changes on its line switch states take place more frequently. To monitor these changes, the data-driven switch state identification algorithms usually use large quantity of historical measurements and historical switch states to discover the relationship between them, and identify the real-time switch states using the real-time measurements according to the relationship. Large quantity of data are involved in the relationship discovery process, which impose heavy computational burden to the server at the central operator of the distribution automation system. To tackle this problem, an edge computing framework for data-driven switch state identification algorithm is developed in this paper. The proposed framework utilizes the computational resources of the smart terminals so that some of the computational tasks are accomplished in the terminals, and others are sent to the computational resource pool with an even distribution over time. Using the proposed framework, the computational and communication pressure of the distribution automation system is relieved, and the performance of the data-driven identification algorithm is improved. Case studies verify its effectiveness.