AI-Based Internal Power Grid Topology Change System for Optimal Operation of Offshore Wind Farms

Offshore wind power is gaining attraction as a sustainable energy solution, but optimizing topologies for changing environments remains a significant challenge. Existing algorithms design static topologies based on specific environmental conditions, which limits the flexibility of real-time adaptation.<br/>In this study, we propose a dynamic topology optimization technique using deep Q-networks (DQN) to address this problem. We model offshore wind farm topology optimization as a Markov decision process (MDP) and apply DQNs to solve it in real-time. Experiments are conducted through simulations using an offshore wind farm model with 40 wind turbines (5MW). DQN-based optimization achieved an annual energy production of 894.7 GWh and an average transmission loss rate of 4.80%, outperforming the fixed topology and random breaker switching methods. DQN showed high adaptability to seasonal wind direction changes and power demand fluctuations, maintaining stable performance throughout the year.