Charles Darwin University (CDU) has recently conducted a world-first study combining Indigenous seasonal calendars with deep learning models, providing a new solution for solar power generation forecasting. The research, titled "Convolutional Ensembles for Solar Power Forecasting Using Indigenous Seasonal Information," has been published in the IEEE Computer Society Open Journal.

Solar energy, as one of the world's leading renewable sources, has long faced challenges in prediction accuracy, which limits its reliability. Forecasting solar power output is difficult due to influences from weather, atmospheric conditions, and energy absorption on solar panel surfaces. The CDU research team innovatively utilized Tiwi, Gulumoerrgin (Larrakia), Kunwinjku, and Ngurrungurrudjba Indigenous calendars, along with the modern "Red Centre" calendar, combined with a novel deep learning model to develop a new method for predicting future solar panel electricity output.

The team validated the model using data from the Desert Knowledge Australia Solar Centre in Alice Springs, showing that the model's error rate in predicting solar power generation was significantly lower than current industry-leading models—less than half.

Co-author and CDU PhD student from the Bungarlung people, Luke Hamlin, stated that the environmental knowledge embedded in Indigenous calendars is invaluable. He noted: "Incorporating Indigenous seasonal knowledge into solar power forecasting aligns predictions with natural cycles observed over thousands of years, greatly improving accuracy."

Hamlin further explained that, unlike traditional calendar systems, Indigenous seasonal insights are deeply rooted in local ecological cues, such as changes in plant and animal behavior, which are closely linked to sunlight and weather patterns. By integrating this knowledge, the prediction model can more accurately reflect subtle environmental changes, providing more precise and culturally meaningful forecasts for specific Australian regions.

Co-authors Associate Professor Bharanidharan Shanmugam and Lecturer Dr. Sucheethan Selvarajah from Information Technology stated that combining advanced AI technology with ancient Indigenous wisdom has the potential to revolutionize forecasting techniques. Associate Professor Shanmugam pointed out: "Accurate solar forecasting is full of challenges that hinder the development of universal prediction models." Dr. Selvarajah added that the success of this method indicates its potential as a powerful tool for promoting solar power forecasting in rural areas, with future research exploring the model's application in other regions and renewable energy fields.