The solar energy available in Andalusia will increase through the end of the century
A team from the University of Córdoba is using artificial intelligence to forecast the annual solar energy available in Andalusia through 2100 based on temperature data.
Peak solar hour (PSH) is a key concept in the field of photovoltaic energy. This standardized measurement allows us to determine the annual energy available and to plan more efficiently by ascertaining how much energy a solar panel can receive at a given location and time. Each peak solar hour is equivalent to 1,000 watts per square meter over the course of one hour. Having a medium- and long-term forecast of this variable can facilitate planning for clean energy generation.
This was the context for the latest work by Javier Estévez, Juan Antonio Bellido, and Amanda García, who conduct research in the Hydrology and Agricultural Hydraulics group at the University of Cordoba. In their work, they developed an artificial intelligence–based model that generates a map of peak solar hours up to 2100 using only temperature data . "Using a simple, easily measurable variable for which we have large datasets and future forecasts - namely, temperature - we estimate solar radiation and then peak solar hours," explained Professor Javier Estévez.
After evaluating four machine learning models and different model configurations (incorporating a greater or lesser number of temperature-derived variables, such as daily maximums and minimums, thermal range, etc.), the researchers found that the multilayer perceptron model, with the configuration that included the most variables, performed best.
To estimate solar radiation and project peak sun hour figures across Andalusia based on temperature, the researchers used future daily temperature projections under different climate change scenarios. Across all the future scenarios, there is one constant: peak solar hours increase. "In the most moderate emissions scenario, the annual figures rose from 1,850 kWh/m ² in the 2024–2030 period to 1,950 kWh/m² per year in 2100. In the worst-case scenario, with higher emissions, we see an even more pronounced increase, exceeding 2,000 kWh/m ² per year," the researcher noted.
Thus, overall, a significant positive trend in this variable was found in Andalusia, which is good news for clean energy, although some coastal areas and parts of the Sierra Nevada showed a decrease in it. It is important to remember that increases in solar radiation and available energy are also linked to rises in temperatures.
How is a model validated?
To ensure that the model's temperature-based solar radiation estimates are reliable, it is validated using existing solar radiation data. For this the team used data from 122 weather stations that had measured solar radiation between 2000 and 2022. As a result, the team verified that the model was working correctly and could predict future data in a similar manner.
The model itself is an outcome. "The model is open-access and available to other research groups and managers. And, although the computational effort required to develop it is significant, it can then be run on a standard computer," explained researcher Juan Antonio Bellido. By using temperature, a widely available and inexpensive-to-measure variable, this model can be adapted to locations with limited resources and still deliver good results.
Therefore, from a planning perspective, the study highlights the growing potential of solar energy in southern Spain over the course of this century. These results could be used to analyze how energy performance changes across different locations, providing a basis to select the best options to generate clean energy.
Reference
Bellido-Jiménez, Juan Antonio and Estévez Gualda, Javier and García-Marín, Amanda P., Projecting Solar Peak Hours (Until 2100) in Southern Spain Using Temperature-Based Machine Learning Models http://dx.doi.org/10.2139/ssrn.5189938
Applied Energy
10.1016/j.apenergy.2026.127395
Projecting solar peak hours in southern Spain using temperature-based machine learning models until 2100
17-Jan-2026