Hybrid Deep Reinforcement Learning for UAV Inspection in Large-Scale Wind Farms: Deployment and Routing Optimization

The growth of global wind capacity necessitates efficient maintenance and inspection, particularly in offshore wind farms. However, existing manual inspection method is time consuming and lacks scalability. This article introduces an autonomous unmanned aerial vehicle (UAV) inspection method for offshore wind farms by combing heuristic algorithm with deep reinforcement learning (DRL). Our approach integrates UAVs deployment and routing optimization, factoring in environmental conditions, such as wind and weather. The inspection task is divided into UAVs deployment and routing optimization, with models constructed for UAVs, offshore wind farms, and wind conditions. A heuristic algorithm is used to optimize UAVs deployment, ensuring efficient coverage despite the UAVs’ limited range. For routing optimization, we develop a DRL model for effective path planning of UAVs simultaneously. Experimental results demonstrate the superior efficiency of our approach in terms of coverage, energy consumption, and adaptability to environmental changes. By introducing the hybrid DRL algorithm, a scalable and robust solution for offshore wind farm inspections is developed, enhancing both operational efficiency and reliability.