A hybrid geospatial–machine learning framework for evaluating floating PV technical potential using satellite imagery

Floating photovoltaic (FPV) systems present a strategic alternative to land-based PV installations that faces land scarcity, land-use conflicts and rapid electrification needs. This study develops an integrated geospatial framework that combines machine learning-based water body extraction and fuzzy GIS-based multi-criteria decision-making (MCDM) to perform a first-order techno-spatial suitability screening of floating PV locations and estimates the theoretical installed capacity of floating PV systems over verified water bodies in Amhara Region. Furthermore, FPV potential estimates were constrained using FAHP suitability analysis and bathymetric screening to provide more realistic deployment scenarios for the Amhara Region of Ethiopia. Water bodies were delineated automatically using a Random Forest (RF) classifier applied to Sentinel-2 satellite imagery on Google Earth Engine. Subsequently, a Fuzzy Analytical Hierarchy Process (FAHP) was implemented to rank candidate reservoirs using eight techno-economic and socio-environmental criteria such as Global Horizontal Irradiance (GHI), elevation, slope, proximity to transmission lines, proximity to roads, settlement distance, wind speed, and temperature. Sensitivity analysis across 10 perturbation scenarios confirmed model robustness, maintaining CR < 0.1 with 9/10 scenarios showing 100% rank stability. Results for water detection demonstrated that the random forest classifier identified 21 water bodies with OA of 0.904, recall of 0.917, F1-score of 0.906, precision of 0.894 and a Cohen’s kappa coefficient of 0.78. The framework also produced suitability maps for identified priority reservoirs for FPV development based on detected water bodies. Further this work also quantified the installed capacity potential of suitable sites using temperature-adjusted capacity estimates. The combined results showed that the total usable FPV potential in Amhara’s water bodies lies almost entirely within locations classified as suitable, with approximately 822.5 GW (96.88%) and an additional 7 GW (0.82%) in the “Most Suitable” patches, while only about 2.3% of the water-surface pixels are excluded as unsuitable. Under bathymetric constraints, the bathymetry-screened installed capacity potential of suitable pixels reaches 304.33 GW, with the most suitable patches estimated at 2.59 GW. The paper demonstrated that integrating machine learning with GIS-FAHP delivers a high-resolution FPV planning tool capable of supporting strategic renewable-energy expansion in Ethiopia’s grid-level planning decisions.

Authors

Institutions

Publication Details

Journal
Discover Applied Sciences
Published
2026-11-17
DOI
https://doi.org/10.1007/s42452-026-09892-4
Primary Topic
Solar Radiation and Photovoltaics
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

A hybrid geospatial–machine learning framework for evaluating floating PV technical potential using satellite imagery

Fekadu Shewarega, Getachew Biru Worku, Tewodros Gera Workineh, Tassew Tadiwos Zewdie et al.
Discover Applied Sciences
Solar Radiation and Photovoltaics
article

A hybrid geospatial–machine learning framework for evaluating floating PV technical potential using satellite imagery

Fekadu Shewarega, Getachew Biru Worku, Tewodros Gera Workineh, Tassew Tadiwos Zewdie, Tefera Terefe Yitayew
article en

Abstract

Floating photovoltaic (FPV) systems present a strategic alternative to land-based PV installations that faces land scarcity, land-use conflicts and rapid electrification needs. This study develops an integrated geospatial framework that combines machine learning-based water body extraction and fuzzy GIS-based multi-criteria decision-making (MCDM) to perform a first-order techno-spatial suitability screening of floating PV locations and estimates the theoretical installed capacity of floating PV systems over verified water bodies in Amhara Region. Furthermore, FPV potential estimates were constrained using FAHP suitability analysis and bathymetric screening to provide more realistic deployment scenarios for the Amhara Region of Ethiopia. Water bodies were delineated automatically using a Random Forest (RF) classifier applied to Sentinel-2 satellite imagery on Google Earth Engine. Subsequently, a Fuzzy Analytical Hierarchy Process (FAHP) was implemented to rank candidate reservoirs using eight techno-economic and socio-environmental criteria such as Global Horizontal Irradiance (GHI), elevation, slope, proximity to transmission lines, proximity to roads, settlement distance, wind speed, and temperature. Sensitivity analysis across 10 perturbation scenarios confirmed model robustness, maintaining CR < 0.1 with 9/10 scenarios showing 100% rank stability. Results for water detection demonstrated that the random forest classifier identified 21 water bodies with OA of 0.904, recall of 0.917, F1-score of 0.906, precision of 0.894 and a Cohen’s kappa coefficient of 0.78. The framework also produced suitability maps for identified priority reservoirs for FPV development based on detected water bodies. Further this work also quantified the installed capacity potential of suitable sites using temperature-adjusted capacity estimates. The combined results showed that the total usable FPV potential in Amhara’s water bodies lies almost entirely within locations classified as suitable, with approximately 822.5 GW (96.88%) and an additional 7 GW (0.82%) in the “Most Suitable” patches, while only about 2.3% of the water-surface pixels are excluded as unsuitable. Under bathymetric constraints, the bathymetry-screened installed capacity potential of suitable pixels reaches 304.33 GW, with the most suitable patches estimated at 2.59 GW. The paper demonstrated that integrating machine learning with GIS-FAHP delivers a high-resolution FPV planning tool capable of supporting strategic renewable-energy expansion in Ethiopia’s grid-level planning decisions.

Discover Applied Sciences
University of Rwanda (RW), Bahir Dar University (ET), University of Duisburg-Essen (DE), Adama Science and Technology University (ET)
Solar Radiation and Photovoltaics
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon