A Machine Learning Algorithm to Minimize Distribution Lines Overloads

To address the issue of line overloads arising from the increasing integration of Renewable Energy Sources (RES) in distribution networks, advanced grid management strategies are needed to dynamically optimize network configurations. Within this context, grid reconfiguration methods allow to define the best configuration of the network to minimize line overloads, improve voltage values within the network and reduce power losses. Hence, reconfiguration methods allow to enhance the operations and management of distribution networks hosting a high share of RES. In this framework, this paper proposes a grid reconfiguration tool based on a machine learning algorithm, aimed at minimizing line overloads and reducing the number of reclosures in a distribution system. The proposed methodology is validated on a 10-node test network with significant RES penetration. The outcomes obtained show that, by applying the reconfigurations proposed by the tool, 56% reduction in total overload occurrences is obtained. The computational time needed by the machine learning-based algorithm to output the best configuration among all the possible ones is less than 1 second, demonstrating the usefulness of the proposed tool to cope with (near)real-time network issues. This result demonstrates the effectiveness of the proposed algorithm in reducing line overloads and improving the system performance.