Brain-inspired AI shakes up clean energy forecasts 

When the weather shifts, so does the power supply in grids that rely on renewable energy. Clouds gather, winds fade and sunshine disappears; the change can happen in minutes. This unpredictability makes balancing energy supply and demand a challenge.  

Forecasting how much solar and wind energy can be produced is crucial, but traditional methods often fall short, forcing grid operators to fall back on fossil fuels. Researchers at Khalifa University, in collaboration with the Malaviya National Institute of Technology Jaipur in India and the University of Johannesburg in South Africa, have developed an artificial intelligence framework that improves the accuracy of renewable energy forecasting.  

Their new model, called Rate-Encoding Spiking Neural Network (RSNN), mimics the way brain neurons communicate, offering a more robust solution for predicting solar and wind energy generation. “Inspired by how the human brain learns and adapts, the work harnessed brain-inspired AI methods to forecast changes in solar and wind power generation,” says Ameena Al-Sumaiti from Khalifa University’s Department of Electrical Engineering.  

“It’s a double win. We’re improving renewable energy forecasting and reducing the energy the AI tool needs to do it.” 

Ehab El Saadany 

Here’s how it works. The RSNN converts environmental data such as solar irradiation, temperature and wind speed into a sequence of spikes or short bursts of signals, similar to the electrical impulses transmitted by neurons. These spikes represent time-based data, such as hourly changes in solar and wind energy patterns. Their timing and frequency carry information, enabling the RSNN to detect subtle patterns in data that other AI systems can miss. 

The team trained the model on solar and wind data from 2019 to 2022, then tested it using data from May, July and December 2023. The tests spanned seven locations—namely the United Arab Emirates, India, China, Australia, Brazil, Germany and the United States—covering diverse climates from desert to coastal regions.  

The results were unprecedented. RSNN outperformed several advanced deep learning algorithms. It achieved an improvement of more than 120% in solar forecasting accuracy and more than 50% in wind forecasting compared with other AI models.  

Beyond energy forecasting, the RSNN represents progress toward energy-efficient AI. Since spiking neural networks only fire when necessary, as do neurons in the brain, they consume far less power than conventional deep learning systems.  So, the same technology that is helping to manage clean energy can also make AI itself more sustainable.  

“It’s a double win,” says co-author Ehab El Saadany, Dean of Engineering and Physical Sciences at Khalifa University. “We’re improving renewable energy forecasting and reducing the energy the AI tool needs to do it.” 

Reference

Saini, V.K., Al-Sumaiti, A.S., Kumar, A., Kumar, R., Zeineldin, H., & El-Saadany, E.F. (2025). RSNN: Rate encoding mechanism-based spiking neural network for renewable energy forecasting. Energy, 137099. | Article