---
title: "Validating PV simulation uncertainty against field measurements"
slug: "validating-pv-simulation-uncertainty-against-field-measurements"
updated: 2026-08-11T08:51:11Z
published: 2026-08-11T08:51:11Z
canonical: "kb.solargis.com/validating-pv-simulation-uncertainty-against-field-measurements"
---

> ## Documentation Index
> Fetch the complete documentation index at: https://kb.solargis.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Validating PV simulation uncertainty against field measurements

| **Authors**: Branislav Schnierer, Jozef Rusnak, Tomas Cebecauer, Lubos Helienek (Solargis) Jürgen Sutterlüti (Gantner Instruments) | Presented at PVPMC 2026 (Albuquerque) |
| --- | --- |
| 12-14 May 2026 |

[Read full document](https://pvpmc.sandia.gov/app/uploads/sites/243/dlm_uploads/2026/05/Sesion9_Cebecauer_Solargis_PVPMC2026_ValidationOfPVsimulation.pdf)

## Overview

Every PV energy yield simulation carries [uncertainty](https://kb.solargis.com/docs/en/uncertainty-estimates), and that uncertainty determines how much confidence lenders, investors, and operators can place in the result. Solargis quantifies uncertainty for each stage of the [Argus PV simulation chain](https://kb.solargis.com/docs/en/argus-pv-simulation-chain) used in [Solargis Evaluate](https://kb.solargis.com/docs/en/evaluate-overview), but a calculated uncertainty is only useful if reality stays inside it. This study compares simulated PV power output and its monthly uncertainty against measured production at five operating sites across four climates.

## Methodology

The PV simulation chain consists of a sequence of blocks, each contributing its own uncertainty depending on the model used, the uncertainty of its configuration, and the operational conditions. Partial uncertainties are combined by the root-sum-square method into a monthly uncertainty of PV power output (PVOUT), following the framework published by Helienek et al. (2023):

$UNC_{PVOUT} = \sqrt{UNC_{GTI}^{2} + \sum UNC_{BLOCKS}^{2}}$

![](https://cdn.document360.io/ae2d502f-6c0d-4865-a68e-43ad8da61149/Images/Documentation/image-UFYZLN82.png)

***Figure 1****: Example of monthly PV simulation uncertainty for a PV power plant in East Devon, United Kingdom. Each component is rescaled so that the bar height equals the actual combined uncertainty.*

The largest contributions come from solar radiation (measured or satellite-based Global Tilted Irradiation - GTI), environmental factors such as shading, [soiling](https://kb.solargis.com/docs/en/soiling-losses), and [snow](https://kb.solargis.com/docs/en/snow-losses), angular and spectral losses, and the conversion of solar energy to electrical energy. Electrical simulation contributes the least.

| Site | Location and conditions | PV system | Measurement period |
| --- | --- | --- | --- |
| 1 | Gantner Instruments test facility, Arizona, USA. Dry, dusty, high temperatures | 2 monofacial cSi modules, fixed tilt 33°, azimuth 180° | Year 2022, 15-minute data |
| 2 | Sedrun Solar test facility, Switzerland. Alpine, high elevation, snow | Bifacial cSi module, fixed tilt 65°, azimuth 155° | Year 2024, 1-minute data |
| 3 | Vertical bifacial installation, France. Temperate climate, LiDAR surface model | Vertical bifacial cSi modules | 14 months, 10-minute data |
| 4 and 5 | Industrial ground-mounted plants, Slovakia. Continental climate | Fixed-tilt ground-mounted arrays | 14+ years, 15-minute data |

***Table 1****: The five validation sites. All uncertainty bands represent an 80% occurrence interval.*

## Key research findings

**Measured output stays within the predicted band.** At the Arizona site, simulated GTI shows a bias of 1.73% against measurement, and PVOUT bias reaches 0.59% and 1.25% for the two modules, against a monthly uncertainty of roughly 7% to 8%. The alpine site is a far harder test, with high albedo, reflections, and snow cover: GTI bias reaches 4.59% and PVOUT bias 2.13%, against a monthly uncertainty of 7.4% to 9.1%. Measured production tracks the simulated profile at both sites.

![](https://cdn.document360.io/ae2d502f-6c0d-4865-a68e-43ad8da61149/Images/Documentation/image-CLA09ON8.png)

*Figure 2: Measured and simulated monthly specific PVOUT with the uncertainty band, test site 1 (Arizona, USA).*

**Remaining discrepancies come from the measurement setup.** The largest residual deviations trace back to differences between what the module experiences and what the sensor records: shading seen by the module but not by the pyranometer, snow on the modules but not on the pyranometer, and a temperature sensor insulated by snow. At the Arizona site, restricting the October comparison to 07:00-16:00 (removing shading seen by PV module but not by the pyranometer) reduces the daily bias from 1.83% to 0.12%.

**The uncertainty band detects operational problems.** At the French site, the comparison exposed an outage of one inverter branch. At the Slovak plants, the long record surfaced snow representation in satellite data, production outages, and a conservative [degradation](/v1/docs/energy-system-statistics-losses-and-long-term-degradation) estimate that opens a visible gap by the fourteenth year of operation.

## Conclusions and implications for the PV industry

- The **Solargis Evaluate simulation chain and its accuracy are confirmed against real PV power production** across four climates and system types, and the uncertainty concept is verified at monthly scale.
- Results are **sensitive to how well shading, albedo, snow, system specification, and degradation are quantified**. These inputs deserve as much attention as the simulation model.
- **Simulation uncertainty is usable as a monitoring instrument**: production falling outside the uncertainty band indicates a probable operational issue.
- **Work continues on the optical part of the chain**, where shading, snow, and soiling dominate, and on daily uncertainties and portfolio-level analysis.

## Further reading

- ["Uncertainties in PV Power Simulation Chain"](https://ieeexplore.ieee.org/document/10359583): Helienek, L., Rusnak, J., Schnierer, B., Opatovsky, M., Dvonc, L., Fanego, VL, Skoczek, A., Cebecauer, T.
- ["Feature review of photovoltaic modeling software utilizing blind performance assessment"](https://www.sciencedirect.com/science/article/pii/S0038092X25009703): Deville et al.
- ["Uncertainty estimates"](/v1/docs/uncertainty-estimates): Solargis
- ["Impact of Quality Control on Uncertainty in PV Performance Evaluation"](/v1/docs/impact-of-qc-on-uncertainty-in-pv-performance-evaluation): Solargis