First Nations Calendars Enhance Solar Forecasting Accuracy by 26%
A research team from Charles Darwin University has discovered that a solar forecasting model incorporating First Nations seasonal knowledge can be up to 26 per cent more accurate than traditional prediction methods. This innovative approach employs machine learning to analyse seasonal calendars, enhancing the ability to predict solar panel output.
The findings, published in IEEE Explore, highlight how integrating First Nations seasonal insights into solar power generation forecasts can significantly improve accuracy by aligning predictions with natural cycles that have been observed for millennia. The study indicates that the incorporation of diverse cultural data can refine forecasting models and enhance predictive analytics in energy applications.
Enhanced Accuracy Through Local Knowledge
The research revealed that predictions using local First Nations knowledge were between 14 per cent and just over 26 per cent more accurate. The model drew on the seasonal calendars of the Tiwi, Gulumoerrgin, Kunwinjku, and Ngurrungurrudjba First Nations, alongside a modern calendar known as the Red Centre. Additionally, a new dataset called AliDKA was created, utilising information from the Desert Knowledge Australia Solar Centre in Alice Springs.
The FNS-Metrics (First Nations seasonal) dataset included variables such as temperature, irradiance, rainfall, and transitional weather patterns, while AliDKA encompassed temperature, relative humidity, two measurements for horizontal radiation, wind direction, daily rainfall, and both global and diffuse tilted radiation. The team achieved an error rate that was less than half of that found in commonly used forecasting models today.
State-of-the-Art Performance
The authors of the study reported that their framework consistently outperformed traditional methods, achieving a remarkable R2 of 0.8641 and a mean squared error (MSE) of 22.41, translating to increases of 14.60 per cent and 26.21 per cent compared to baseline models. This success suggests that the approach could serve as a valuable tool for enhancing solar power generation predictions in rural areas like Alice Springs by integrating First Nations seasonal cycles for improved accuracy.
The Challenge of Solar Predictions
Accurately predicting sunlight exposure on solar panels is complex, as it varies significantly based on location, weather, and atmospheric conditions. In 2021, researchers from South Korea developed a machine learning method capable of making predictions up to one day in advance, while a 2023 study from UNSW indicated that climate change may affect the reliability of solar power in the future.
The Darwin team asserts that incorporating long-term observational data, particularly from Indigenous populations, can enhance model accuracy. Deep learning models, often referred to as artificial intelligence (AI), process extensive historical data to generate forecasts.
Local Ecological Knowledge
The paper emphasises that different regions in Australia possess unique First Nations seasonal information, reflecting the diverse ecological knowledge and cultural practices of Indigenous communities. Unlike standard calendar systems, these seasonal insights are deeply connected to local ecological cues, such as the behaviours of plants and animals, which correlate closely with changes in sunlight and weather patterns.
By integrating this rich knowledge, predictions can be tailored to capture more nuanced shifts in environmental conditions, resulting in more precise and culturally informed forecasts for specific regions across Australia. Importantly, First Nations communities in northern and central Australia have seasonal calendars that are specific to their local areas, which could be particularly beneficial for solar farms situated in remote locations.
Challenges in Real-Time Implementation
Despite the promising results, the model faces challenges in incorporating real-time information, as the training time for the AI is described as “relatively long.” Nevertheless, the insights gained from First Nations knowledge could provide a robust framework for optimising solar resources in the future.
For those interested in supporting independent media and the dissemination of accurate information, consider making a one-off donation or becoming a regular supporter of Renew Economy. Your support is crucial.
