The CTC Technological Centre has presented a short-term predictive model for photovoltaic solar energy production. By incorporating advanced artificial intelligence techniques, the system makes it possible to forecast this energy generation with 92% accuracy and turn an uncertain resource into a schedulable supply. In addition to adapting companies’ consumption to the generation predicted by the model, this advance improves efficiency and reduces operating costs in industrial facilities by between 15% and 25%. It would also reduce renewable energy waste and encourage self-consumption within companies.
Marco Antonio Melgarejo, data scientist at CTC, presented this work during the 26th International Conference on Computational Science and Its Applications (ICCSA 2026), held at the University of Minho in Portugal. It is a leading global event where disruptive solutions are explored to help set new trends in the field of computational science.
The research presented is titled AI Development for Photovoltaic Installation Forecasting and corresponds to a use case within the FUTCAN project, which is co-financed by the European Regional Development Fund through Cantabria’s ERDF 2021–2027 Operational Programme, under the grant line “Aid for research projects with high industrial potential from excellent technology agents for industrial competitiveness TCNIC”. Specifically, this action falls within the work stream “Generation of methodology for the accelerated development of industrial and sustainable components”.
The model aims to introduce a predictive management approach to the generation of this type of renewable energy. This premise makes it possible to address both the uncertainty associated with weather changes and the high cost linked to the “zero-injection” restriction for facilities with capacities above 100 kW.
This regulation requires significant investment in equipment in order to inject surplus energy generated by panels directly into the distribution grid. It is a technical and legal requirement designed to protect grid stability and safety, which would be mitigated by the model proposed by CTC.
The system presented delivers highly accurate predictions, ensuring reliable decision-making. To achieve this, production data were collected from a small pilot installation located on the roof of the Cantabrian technology centre’s building. This location poses a challenge in itself due to the region’s high meteorological variability.
In addition to the small solar array, there is a meteorological station that provides data on irradiance, temperature, humidity, wind and precipitation. This makes it possible to capture production variability in relation to, and as a function of, environmental conditions and timing.
The measurement equipment is complemented by a dust sensor, which makes it possible to study and measure the direct physical impact of ambient dust on the panels. This is particularly relevant in saline environments such as the one found in Cantabria.
Beyond sharing the results with the international scientific community, comparing the proposed approach with researchers and professionals in computational science and identifying possible improvements for its application in predictive models, participating as a speaker at this event also comes with the reward of having the paper published in Lecture Notes in Computer Science (LNCS) by Springer.
This is one of the world’s most prestigious scientific book series in the field of computer science and information technologies. It is indexed by leading databases worldwide, including Scopus, EI Engineering Index and the Thomson Reuters Conference Proceedings Citation Index (included in ISI Web of Science).
