Solar panels are one of the most promising clean energy sources out there. To be truly optimized, though, plant operators need to be able to precisely project when the sun is shining and forecast photovoltaic power generation.
A research team from the Institute of Statistics at the Karlsruhe Institute of Technology is using AI and machine learning to improve upon current projection methods, as Tech Xplore shared.Â
The team's set of three approaches could allow future operators to better balance solar power supply and demand. The study was published in Advances in Atmospheric Sciences and pointed to ways to correct conventional weather forecast models.
"Weather forecasts aren't perfect, and those errors get carried into solar power predictions," explained the study's lead author, Nina Horat.
The three methods were correcting the weather forecasts before they enter the PV models, adjusting the power projections after, and letting AI and machine learning independently predict solar power strictly based on the weather data.
"By tweaking the forecasts at different stages, we can significantly improve how well we predict solar energy production," Horat said.
The study found that the second method of correcting power forecasts enhanced models more than adjusting the weather inputs. Another key takeaway was that the time of the day was of the utmost performance.
"We saw major improvements when we trained separate models for each hour of the day or fed time directly into the algorithms," noted the study's corresponding author, Sebastian Lerch.
The team's final approach of letting the machine learning algorithm do all the work off of weather data is one that could pay dividends in the future. It bypasses PV models and isn't reliant on having specific knowledge of a solar plant's design.
|
If you were to install home solar panels, which of these factors would be your primary motivation?
Click your choice to see results and speak your mind. |
Energy independence âš¡ Theoretically, a team could plug in the numbers trained on historical weather and power performance and get accurate predictions with those alone.
TCD Picks » Upway Spotlight
💡Upway makes it easy to find discounts of up to 60% on premium e-bike brands
The scientists' work addresses a key area to improve for solar as its growth continues to exceed industry expectations. The latency of solar energy is one of the biggest concerns that arises. Maximizing sunlight can ease some of those worries.
A team at the University of Nottingham similarly used AI to enhance weather prediction models to project changes in cloud cover. Researchers in the Philippines also unveiled an algorithm that they say can improve the accuracy of sunny day weather forecasts by up to 94%.
While the use of AI in these situations is exciting and demonstrates some of its potential for good, it's important that researchers account for its environmental impact. Generative AI models use a tremendous amount of power and water, as MIT News reported.
The Karlsruhe team said their study "only constitutes a first step," as it focused on data from just one solar plant amongst other factors. They put forward that including more solar plants, model chain approaches, and weather inputs could improve predictive power even more.
Join our free newsletter for weekly updates on the latest innovations improving our lives and shaping our future, and don't miss this cool list of easy ways to help yourself while helping the planet.
Cool Picks
