Wind farm layout optimization through multi-strategy differential evolution-based reinforcement learning algorithm

Wind energy is an inexhaustive and clean renewable energy. Optimal placement of turbines can guarantee higher energy output and lower cost. The Wind Farm Layout Optimization (WFLO) problem involves determining the best arrangement of turbines. One key factor is the wake effect, where the wind speed is reduced behind a turbine, leading to lower energy output from downstream turbines. WFLO is still a challenging combinational problem. Reinforcement learning has obtained competitive results by maximizing returns through agents' interaction with the environment. In this paper, a Reinforcement Learning-based multi-strategy Differential Evolution Algorithm (RLDEA) is developed to address WFLO problems. In the RLDEA, the mutation strategy and control parameters are dynamically adjusted by reinforcement learning so that the more appropriate ones can be applied to evolution. The effectiveness of the proposed RLDEA is evaluated across four test cases. Compared with the seven latest meta-heuristic algorithms and reinforcement learning-based algorithms, it shows the powerful optimization capacity when solving the combinational problem.