Deep Reinforcement Learning for Multi-Objective Location Optimization of Onshore Wind Power Stations: A Case Study of Guangdong Province, China

Yanna Gao¹Hong Dong¹Liujun Hu¹Fanhong Zeng¹Yuqun Gao¹Zhuonan Huang²Shaohua Wang^3*

• ¹Guangzhou Power Supply Bureau, Guangdong Power Grid Co., LTD., Guangdong, China
• ²School of Information Engineering, China University of Geosciences (Beijing), Beijing, China
• ³Aerospace Information Research Institute, Chinese Academy of Sciences (CAS), Beijing, China

As the development of wind power progresses, the optimization of wind farm locations and allocations has become a critical challenge. This study proposes a multi-objective maximal covering location problem (MO-MCLP) to optimize the siting of onshore wind power stations (OWPS), addressing issues such as low wind energy utilization, insufficient development of suitable areas, and unbalanced electricity demand coverage. To solve this model, a deep reinforcement learning (DRL) algorithm was developed and compared with the optimization method implemented by the Gurobi solver. Experimental analysis focused on wind-rich coastal regions of Guangdong Province, where the computational efficiency and solution quality of both approaches were rigorously evaluated. The results demonstrate that the DRL algorithm not only achieves competitive optimization performance but also significantly outperforms Gurobi in terms of computational speed and scalability for largescale problems. This highlights DRL's potential as an efficient alternative to traditional optimization methods in complex spatial planning scenarios. The study provides actionable insights for wind farm planning and advances the application of artificial intelligence in sustainable energy infrastructure development.