Sensitivity of Regional PV Production Simulations to System Characteristics and Spatial Distribution Assumptions: A Case Study over France.

Accurate simulation of regional photovoltaic (PV) electricity production is essential for energy system planning, grid operation, and climate-energy assessments. At regional and national scales, PV production is often modeled using simplified representations of PV systems due to the limited availability of system orientation and deployment information. However, system characteristics such as tilt, azimuth, and spatial distribution significantly affect daily generation patterns and seasonal energy output. Simplifying these characteristics can introduce systematic biases, not only in total energy estimates but also in the timing of generation. Despite the common use of such simplifications, the sensitivity of regional PV simulations to individual and combined assumptions remains poorly quantified, making it difficult to determine which modeling choices are acceptable and which may lead to significant errors.

We developed a regional PV modeling framework over France, combining a high-resolution inventory of PV installations with a physics-based production model driven by ERA5 reanalysis data. The framework was validated against transmission system operator (TSO) measurements for medium- and large-scale installations. Using this validated framework, we constructed a reference simulation preserving observed PV system characteristics and compared it to progressively simplified scenarios reflecting common modeling assumptions—uniform orientation, fixed tilt angles, and homogeneous spatial distribution.

The modeling framework reproduces the temporal variability of regional PV production with high correlations (0.95–0.98) relative to TSO measurements, with a moderate positive bias observed in most regions. Seasonal analysis confirms accurate capture of daily production timing (morning ramp-up, peak, evening decline). Remaining magnitude discrepancies are likely attributable to differences in installed capacity coverage between the PV inventory and TSO observations.

The sensitivity analysis demonstrates that the impact of modeling simplifications depends strongly on their combination. Individual assumptions—such as uniform south-facing orientation or fixed tilt angles—produce moderate deviations from the reference simulation and remain acceptable when broadly consistent with the underlying fleet characteristics. However, combining multiple simplifications (uniform orientation, fixed tilt, and homogeneous spatial distribution) yields substantially larger errors, particularly during winter and low-irradiance periods. These compound errors primarily affect the magnitude of the diurnal cycle, especially during morning and evening hours. 

These findings provide practical guidance for modelers who must simulate future PV production without detailed information on system characteristics. By quantifying the sensitivity to common modeling choices, this framework establishes the minimum level of system detail required for reliable scenario modeling. While individual assumptions on system orientation or spatial distribution may be acceptable for large-scale scenario analyses, combining multiple simplifications can substantially reduce reliability. This framework helps interpret the implications of modeling choices and limits uncertainty in climate change projections of regional PV production and energy transition pathways.