Researchers have assessed the reliability of models used to estimate direct and diffuse irradiance from global horizontal irradiance (GHI) data at two sites in southern Peru. The study identifies a systematic, altitude-dependent divergence between the model results and Solargis satellite data, with potential implications for estimating photovoltaic (PV) yields in Andean regions.
The study, “GHI-based PV yield assessment in high-altitude Andean sites: workflow performance and systematic limitations of irradiance decomposition models,” compares two sites with contrasting climates: Cota Cota, located 4,240 meters above sea level on the Peruvian Altiplano, and Pampas de Majes, at 1,498 meters above sea level in an arid coastal valley in southern Peru.
The researchers used GHI and ambient temperature measurements recorded between 2021 and 2023. They used the data to adapt ERA5 meteorological reanalysis data with multilayer perceptron models, a type of neural network, to reconstruct a 25-year hourly climatology for the 2000-24 period.
Based on these time series, they generated representative meteorological years and estimated diffuse horizontal irradiance (DHI) and direct normal irradiance (DNI).
For the irradiance decomposition, the researchers used the Engerer2 model with two clear-sky schemes. The first was Threlkeld-Jordan, which was used in the original calibration of the Engerer2 model. The second was ARGPv2, an empirical model developed specifically for high-altitude sites in the Andes of northwestern Argentina, in the provinces of Salta and Jujuy, and recalibrated using the McClear clear-sky model.
Engerer2 is an empirical decomposition model that separates GHI into its direct and diffuse components. It uses a clear-sky model as a reference to describe expected irradiance under cloud-free conditions. Threlkeld-Jordan and ARGPv2 provide this clear-sky reference, with ARGPv2 designed to better represent the atmospheric conditions found at high-altitude Andean sites.
The researchers then fed the resulting irradiance components into a Liu-Jordan transposition model to calculate irradiance on the plane of the PV modules. This model is a widely used solar-energy model for estimating how much solar irradiance reaches a tilted surface, such as a PV module, from irradiance measured or estimated on a horizontal surface.
They subsequently applied a PV production model based on standard test conditions (STC) to calculate annual generation across a full matrix of module tilt and orientation combinations.
The procedure allowed the researchers to assess how differences introduced during irradiance decomposition ultimately affect energy yield estimates for PV installations.
The limited availability of direct DNI and DHI measurements is one of the main challenges facing solar resource assessments in many parts of the Andes. Where only GHI measurements are available, both components must be estimated using models, introducing an additional source of uncertainty.
Because direct DNI and DHI measurements were unavailable at the two study sites, the researchers compared their decomposition results with Solargis data, an independent dataset based on satellite observations and atmospheric models.
The comparison revealed a systematic divergence between the two approaches that, according to the researchers, varies with altitude.
The finding is particularly relevant to PV projects on the Altiplano and in other high-altitude areas of the Andes, where atmospheric conditions differ from those at the locations used to calibrate many conventional irradiance decomposition models.
The study highlights a limitation of solar resource assessment methods based exclusively on GHI. Although these approaches can reconstruct long-term time series and estimate PV production at sites without comprehensive solar radiation measurements, uncertainty associated with separating direct and diffuse irradiance can become systematic at high altitudes.
The researchers said the findings also point to the need for local validation of models used in PV resource and yield assessments in the Andes, particularly when the results are used for project sizing or long-term generation forecasts.
This content is protected by copyright and may not be reused. If you want to cooperate with us and would like to reuse some of our content, please contact: [email protected].
This content is protected by copyright and may not be reused. If you want to cooperate with us and would like to reuse some of our content, please contact: [email protected].
Comments
Please login to comment
Martedì, 22 Settembre 2026
11:00 – 12:00 CEST, Roma
Monday, October 26, 2026
10:30 am – 11:30 am CEST, Berlin, Paris, Madrid
Thursday, September 10, 2026
2:00 pm – 3:00 pm CEST, Berlin, Paris, Madrid
Tuesday, September 15, 2026
5:00 pm – 6:00 pm CEST, Berlin, Paris, Madrid
Our special edition for Intersolar South America 2026 is here!
Discover the latest insights into the Brazilian solar market – in Portuguese.
A two-day conference in Austin, Texas, bringing together leaders in US solar manufacturing, equipment specification, and factory execution.
Saudi Arabia is accelerating its clean energy transition—join the SunRise Arabia Clean Energy Conference 2026 in Riyadh to explore how solar PV and energy storage are powering its digital economy.
pv magazine USA hosts its multi-day virtual event on U.S. solar and energy storage, covering domestic manufacturing, distributed energy and the growing role of solar-plus-storage in meeting AI-driven power demand.
Thursday, October 7, 2026
11:00 am – 12:30 pm CEST, Berlin, Paris, Madrid
You have no items in your basket.
