Physics-informed experimental analysis of dust soiling, partial shading, and electrical configuration in desert photovoltaics

The energy yield of photovoltaic (PV) plants in hyper-arid climates is governed by the combined effects of ageing, dust soiling, partial shading, and electrical configuration, yet these stressors have rarely been evaluated together on the same field-aged modules. This study addresses this gap through a factorial outdoor comparison of three crystalline-silicon module technologies, including monocrystalline ISOFOTON I-75 and UDTS-50, and polycrystalline Photowatt PWX-500, after approximately twelve years of exposure at the Research Unit for Renewable Energies in the Saharan Medium (URERMS, CDER), Adrar, Algeria. Current–voltage characteristics were measured under clear-sky, naturally soiled, and partially shaded conditions for individual modules and two-module series and parallel configurations. Uniform-condition quantities were compared at Standard Test Conditions (STC), whereas shaded-condition voltage, fill-factor, and shape descriptors were retained at measured outdoor test conditions (OTC), because the analyzer’s STC translation is not physically reliable for distorted multi-peak curves. The shaded-condition maximum power was additionally reported in the analyzer-translated form as a comparative index. Under clear sky, aged modules retained 75.5–81.2% of their rated maximum power. Soiling reduced short-circuit current and power by 4.4–22.5% and 7.2–29.8%, respectively, with limited effects on voltage and fill factor. Partial shading reduced the module-type mean measured fill factor to 0.180–0.400. Using analyzer-translated maximum power as a comparative index, the reductions relative to the clear-sky STC baseline were 46.8–68.6%. The translated shaded parallel-to-series power ratio reached 1.94, with a mean of 1.48, whereas the corresponding measured OTC ratios ranged from approximately 0.987 to 1.260. Differences in acquisition conditions and shaded-curve translation prevent interpreting these ratios as controlled topology gains. An illustrative mismatch map relates the measured series–parallel behavior to current- and voltage-dominated mismatch without fitting an electrical model to the data. In addition, 183 measured OTC sweeps, grouped into 36 cases, were analyzed using four physics-guided dimensionless descriptors. The module-type-held-out Random Forest classified 25 of 36 cases correctly, corresponding to 69.4% case accuracy and a macro-F1 score of 0.705. Logistic regression provided an 80.6% baseline. These results support a proof-of-concept diagnostic separation of clear, soiled, and shaded states rather than a deployment-ready classifier. The observations show condition-dependent performance differences among the tested specimens and support further controlled evaluation of stressor mitigation, electrical configuration, and physics-guided monitoring in Saharan environments.