High-Resolution 1-Minute Photovoltaic Power, Temperature and Irradiance Time Series from GEDERlab (UPV)

The experimental infrastructure of GEDERlab, an energy hub located at the Polytechnic University of Valencia (UPV), was used to develop and validate photovoltaic power prediction models. This environment was specifically designed to integrate, operate, and monitor different renewable energy sources, including solar photovoltaic, wind, hydroelectric, and biomass. Thanks to its real-time energy and meteorological data acquisition system, it provides an ideal setting for evaluating the performance of solar systems under real-world, controlled conditions.The laboratory's photovoltaic installation is composed of nine SERAPHIM SRP-310-E11B solar panels, each with a power of 310 W (STC) and 230 W (NOCT), providing a total peak power of approximately 2.1 kW. These panels are installed on the roof of the building, in a strategic position to efficiently capture incident solar radiation. The system is complemented by a Schneider Conext MPPT 60 150 charge regulator, responsible for optimally managing the energy flowing to the batteries, thus prolonging their useful life. In addition, a Schneider Conext XW+ inverter is included, which transforms the direct current generated by the panels into alternating current usable by the system or for grid injection.To correlate energy production with weather conditions, the lab uses a Davis Vantage Pro2 weather station, also located on the rooftop. This station continuously records key meteorological variables such as solar radiation, ambient temperature, relative humidity, wind speed and direction, and precipitation. Data is collected automatically at one-minute intervals, ensuring adequate temporal resolution for detailed modeling and analysis.All the information generated by both the production systems and the weather station is stored in a structured database. This continuous collection system provides an extensive and accurate historical record, essential for identifying patterns of behavior, establishing relationships between variables, and training reliable prediction models. Furthermore, this database constitutes a strategic tool for decision-making, optimizing energy efficiency, and extrapolating results to real-life scenarios, facilitating the study's replicability in other similar facilities.The dataset presented comes directly from the photovoltaic installation described above. The dataset covers 26 days, from 11\/01\/2024 to 11\/26\/2024, inclusive. In this case, the separation is arbitrary, so that 21 days are used to train the models and 5 days to validate them. A high-resolution time series, with a frequency of one minute, is then available, including three key variables: the power generated by the system, the ambient temperature, and the global irradiance.These records reflect in great detail the plant's operating behaviour under real-world conditions, including rapid variations due to meteorological phenomena such as variable cloud cover or temperature changes.