Centralized System for Automatically Restoring Normal Operation of a Power District with Distributed Energy Resources
Abstract
The article deals with the development of principles for organizing centralized control for automatically restoring and optimizing the normal operation of power districts with distributed energy resources (DER). The aim of the study is to achieve more reliable operation of distribution power grids through making a shift to intelligent control without human intervention. The research methods are based on artificial intelligence technologies that combine the mathematical apparatus of graph neural networks and deep reinforcement learning algorithms. The article proposes a method for post-emergency restoration of the normal operation of a power district with DER, which allows overcoming the computational limitations of conventional combinatorial approaches. The proposed automatic control was tested using the example of a power district with a medium-voltage distribution network using the pandapower software environment. It is shown that the combined use of graph neural networks and reinforcement learning ensures high speed and efficiency of control action synthesis. The advantages of and potential for implementing the proposed intelligent automatic control to ensure fault tolerance in power districts with DER are substantiated.
