A simulation-based assessment of water-supply reliability in a standalone photovoltaic water-pumping system under dust soiling in semi-arid Fez, Morocco

International audience Standalone photovoltaic water-pumping systems (PVWPSs) are often evaluated through energy yield or pumped volume, although drinking-water applications require daily service reliability. This simulation-based study quantifies how dust soiling propagates from maximum-power-point (MPP) energy to pumped water and then to reliability for a standalone PVWPS in semi-arid Fez, Morocco. The model combines PVGIS-SARAH3 hourly data for 2020, pvlib irradiance and thermal modelling, a 2.01 kWp fixed PV array, an MPP-equivalent power model with 0.98 tracking efficiency, a centrifugal pump model, and a daily tank balance. Soiling is represented by a linear transmittance-loss rate with cleaning resets and compared with the pvlib Kimber model as an internal consistency check. For the 30-day dry-season window, the clean system delivered 1248.7 m3 of water. With no cleaning, a 1.5 %/day rate reduced delivered water by 20.49% and MPP energy by 20.16%, indicating a near-linear energy-to-water deficit relation. Seasonal divergence remained minor: at 1.5 %/day, the water-energy gap was 0.335 percentage points in summer, 0.384 in winter, and 1.522 over the full year. The reliability response was much sharper. With a 50 m3 tank and 25 m3/day demand, clean annual reliability was 0.997, but no-cleaning reliability fell to 0.2186 at 0.5 %/day and 0.1503 at 1.0 %/day. Loss of Water Supply Probability (LWSP), used as the water analogue of LPSP, reveals a reliability cliff that is not visible from energy deficit alone. At 1.0 %/day, the 50 m3 tank allowed a 41-day annual reliability-compliant cleaning interval, whereas the 30-day dry-season 5% water-deficit rule gave 13 days. Because the annual no-cleaning cases neglect rainfall cleaning, they represent conservative upper-bound soiling stress for service-oriented storage and cleaning design.