A comparative study of different kinematic wake models within metaheuristics for efficient wind farm layout optimization

Abstract

In recent years, metaheuristic algorithms have become pivotal in tackling complex non-linear optimization, including wind farm layout optimization (WFLO) problems for energy-generating wind farms. This paper addresses the optimization of wind turbine layouts in wind farms, aiming to mitigate wind speed deficits caused by wake-induced turbulence. We analyze the performance of seven analytical wake models—Jensen, Park2, Frandsen, Larsen, Bastankhah, Ishihara, and Zhang—to estimate the downstream wind speed deficits included in the objective function of metaheuristics (Genetic Algorithm, Particle Swarm Optimization, and Coral Reefs Optimization with Substrate Layers), for an optimal WFLO solution. We extensively analyze models that characterize the turbulence induced by rotating wind turbine blades, focusing on how they impact the solutions obtained by metaheuristic models. Two wind speed datasets estimate the annual energy production (AEP): a real dataset from Badajoz, Spain, and a synthetic dataset generated using a Weibull distribution. Two terrain configurations are also tested: a 20×20 grid with 20 turbines and a 50×50 grid with 50 turbines. Results show 10-15% efficiency improvements compared to random population initializations, with larger terrains favoring ensemble-based optimization, while denser layouts benefit from evolutionary approaches. Variations in terrain roughness and elevation lead to the optimization process of wind farms, albeit moderately, with energy production differences of 0.1-0.6%.

Keywords

• Wind farm layout
• Wind farm optimization
• Metaheuristics
• Wake models
• Wind energy integration