A Preference-Driven NSGA-III Using Fuzzy AHP for Multiobjective Wind Farm Layout Optimization

Abstract

Wind farm layout optimization (WFLO) requires balancing energy production against infrastructure requirements while providing decision-makers with a defensible method for selecting Pareto-optimal alternatives. This study proposes a preference-guided NSGA-III framework integrated with the Fuzzy Analytic Hierarchy Process (FAHP) for the multiobjective optimization of annual energy production (AEP) and internal road length. Wake interactions are modeled using the Jensen wake model with sum-of-squares superposition, while the internal road network is approximated using a Euclidean minimum spanning tree. Decision-maker preferences are represented through triangular fuzzy pairwise judgments between AEP and road length. The fuzzy priorities are defuzzified and normalized to obtain objective weights, which are used conditionally during partial-front environmental selection and subsequently to rank the final nondominated solutions. The framework was evaluated on a 20-turbine, 2000 m × 2000 m test case using a population of 50, 300 generations, and 30 independent runs per algorithm. Compared with conventional NSGA-III, the proposed NSGA-III FAHP method increased mean hypervolume from 16.5012 to 28.1578. The improvement remained statistically significant after Holm correction (p=4.617×10−7) and showed a large effect size (Cliff’s (δ=0.809)). Differences in the ratio of nondominated individuals, uniformity degree, execution time, best AEP, and best road length were not statistically significant. Sensitivity analysis across seven AEP-road preference scenarios showed that balanced and moderately biased weights produced the highest hypervolume, while AEP remained stable between 84.904 and 84.948 GWh. These results indicate that the proposed framework primarily improves objective-space coverage and provides an explicit mechanism for incorporating uncertain preference information into WFLO.