Wind energy is a key option for reducing dependence on fossil fuels, yet wind-farm site selection remains challenging because it is a high-dimensional spatial decision problem involving heterogeneous criteria, nonlinear interactions, and decision-making uncertainty. To address these challenges, this study develops a nationwide hybrid Machine Learning (ML) and Multi-Criteria Decision-Making (MCDM) framework for wind-site assessment in Iran. A dataset of 55 candidate regions described by ten technical, environmental, climatic, social, and economic criteria is first preprocessed and embedded using Uniform Manifold Approximation and Projection (UMAP) to capture nonlinear structures. The low-dimensional features are then clustered with K-Means++, with the optimal number of clusters determined via Silhouette Coefficient (SC), Davies–Bouldin Index (DBI), and Enhanced Gap Statistic (EGS) (SC = 0.594, DBI = 0.459, EGS = 1.2). The resulting cluster-level decision matrix is evaluated with two objective weighting methods (Entropy and MEREC) and four methodologically diverse MCDM techniques (TOPSIS, WASPAS, ARAS, EDAS). Rankings are aggregated via the BORDA rule to obtain a consensus suitability map. Results show that population density is the most influential criterion, whereas wind speed at 10 m has the lowest weight. Eastern regions such as Zabol and Zahak, central counties including Moalleman and Semnan, and northwestern sites such as Ahar and Sarab emerge as the most suitable sites, all exhibiting wind power densities above 150 W/m 2 . High agreement among methods and consistency with operational wind farms demonstrate the robustness of the proposed model as a scalable tool for wind-energy planning.
Aghapour et al. (Tue,) studied this question.