DOI: 10.62520/fujece.2004826 ISSN: 2822-2881

Reconciling Food and Energy on the Same Land: An Explainable Artificial Intelligence Framework for Agrivoltaic Suitability Mapping

Batuhan Kılıç
Show PDF Cite
The swift growth of utility-scale solar farms has created fierce competition for agricultural land, especially in countries where prime solar energy sites and productive croplands coincide. Agrivoltaic structures, which incorporate agricultural food and electricity generation on one parcel, are a structural solution to this kind of land-use conflict; yet the question of where such systems should be sited has, to date, been addressed in the literature predominantly through subjective multi-criteria decision making methods. This study develops a fully data-driven and explainable machine learning framework for agrivoltaic site selection, using the province of Konya—Türkiye's grain basket and its largest agricultural region—as a case study. The site selection problem is formulated as a supervised binary classification task in which the geographic inventory of existing solar installations provides the ground-truth labels. Thirteen selected feature spanning climatic, topographic, economic, and socio-environmental dimensions were examined through Pearson correlation, variance inflation factor, and mutual information gain analyses. Logistic Regression, Random Forest, and Extreme Gradient Boosting machine learning algorithms were trained with hyperparameter optimization, and observed performance variations were verified using omnibus Cochran's Q and pairwise McNemar statistical tests. Extreme Gradient Boosting achieved the highest performance (total accuracy: 0.9549, F1 score: 0.9533), followed by Random Forest (0.9390) and Logistic Regression (0.8661). Explainable artificial intelligence analysis, which renders the black-box nature of the models transparent, revealed that the most decisive factors determining agrivoltaic suitability were, respectively, distance to grid lines and global horizontal irradiance (GHI). The findings suggest that the proposed methodology may help minimize conflicts between energy and food security, thereby providing a strategic planning tool that can support progress toward the Sustainable Development Goals (particularly SDGs 7 and 13).

More from our Archive