AI and GIS Modelling for Agrivoltaic Deployment in Ukraine

Agrivoltaics can support Ukraine's post-war energy reconstruction by combining distributed solar generation with continued agricultural land use. This is particularly relevant for communities affected by war damage, where energy resilience, rural recovery, and preservation of productive soils must be addressed simultaneously. This study develops a reproducible AI-GIS screening workflow for preliminary agrivoltaic suitability modelling in the Borodianka Territorial Community, Kyiv Oblast, Ukraine. The workflow integrates open geospatial datasets, QGIS-compatible preprocessing, GIS multicriteria decision analysis, and Random Forest classification. A baseline suitability model was produced using COD-AB administrative boundaries and OpenStreetMap layers, including roads, power infrastructure, water bodies, buildings, protected areas, and land-use polygons. The baseline GIS-MCDA model identified 52.19 km2 as highly suitable and 155.87 km2 as moderately suitable for further assessment. A Random Forest classifier trained on expert-rule MCDA labels identified 54.18 km2 as highly suitable and 152.61 km2 as moderately suitable. The model is interpreted as a screening-level decision-support prototype rather than an engineering design. The study demonstrates that open AI-GIS workflows can support early-stage spatial planning of agrivoltaics, while also revealing major barriers: incomplete land-use and cadastral data, limited power-grid information, absence of grid-capacity data, lack of soil and crop-rotation layers, DEM and satellite-resolution limitations, war-related land risks, and the absence of dedicated Ukrainian agrivoltaic standards.