Metaheuristics for the Bi-objective Offshore Wind Farm Layout Optimization Problem in Large-Scale Scenarios

With the escalating need for energy transitioning, maritime concession areas expand, and designing larger offshore wind farms becomes a highly complex computational challenge. The Offshore Wind Farm Layout Optimization Problem (OWFLOP) is a combinatorial problem that handles constrained turbine placement. This study implements four bi-objective metaheuristics (PLS, SPLS, NSGA-II, and SPEA2) to maximize Annual Energy Production (AEP) and minimize installation costs. Evaluations were performed using the Jensen-PARK wake model in large-scale synthetic instances proposed by Cazzaro and Pisinger, featuring distinct turbine models and restricted areas, with available positions ranging from 3000 to 21000. Algorithmic parameter tuning was conducted via Iterated Racing, and the hypervolume metric was chosen to analyze each metaheuristics behavior. Statistical analyzes using Hypervolume point to a better performance of population-based approaches over local searches, while hinting at a low-cost efficacy of the proposed SPLS neighborhood mechanisms, turning it into a potentially suitable hybrid component.