Researchers Improve Solar Power Forecasting With AI
International research team from the Australian University of Sydney, National science agency CSIRO, and Spanish University Pablo de Olavide has proposed a breakthrough approach for using Artificial Intelligence for solar power forecasting.
The researchers became the first to combine neural networks and pattern sequences in their Pattern Sequence Neural Network (PSNN), described in the research paper published by Springer on December 5.
The proposed PSNN neural network was tested on data from photovoltaic solar power system built at the University of Queensland, on weather data and forecasts, collected for a period over 2 years.
As a result, PSNN system has outperformed existing technologies, and “the differences were statistically significant”, claimed researchers.
The researchers plan to further examine the ways to improve solar power forecasting, as well as explore ways to apply new approach to other forecasting tasks.
Artificial intelligence technology is already widely used in the energy sector, making it a perfect case for application in solar energy industry.