Optimizing power system topology for congestion management : a reinforcement learning approach

Details

Supervisors

Jungers, Raphaël M.

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] : ingénieur civil en mathématiques appliquées, à finalité spécialisée

Abstract

Given the ever-changing landscape of power networks, characterized by the growing presence of intermittent and decentralized energy sources like renewables, the transmission of electricity through the grid presents a significant challenge. Conventional power systems were not originally designed to accommodate these emerging energy sources. Due to limitations imposed by physical constraints on power lines, congestion arises as a bottleneck in the electricity grid's transmission capacity. In this context, designing low-cost solutions such as topology management for power networks becomes a promising approach. The congestion problems encountered in power systems represent significant optimization challenges, particularly when aiming to address them through cost-effective topological actions. Recognizing the growing importance of data in real-world applications, this work explores the application of reinforcement learning methods to effectively manage congestion in power networks by leveraging topological actions. By utilizing these methods, the aim is to enhance the efficiency and reliability of power transmission while minimizing costs and mitigating congestion-related issues.