To plan a renewable energy project means knowing how much wind and sun an area gets — today and 50 years from now. Models of future climate change scenarios help with that, but processing the information into a visual format has resulted in a pixelated image like the one below.
"It's really not high enough resolution for us to be able to say anything about renewable energy resources and assessing how much power we might generate from a wind or solar plant in the future," said Ryan King, a senior computational scientist at the National Renewable Energy Laboratory in Golden.
King and his colleagues tried enhancing these images through "super resolution," a process where a computer uses machine learning to add specific pixels to the fuzzy picture.
Their experiment resulted in accurate, high-resolution climate forecasts decades into the future.
"It lets us preserve the right characteristics of atmospheric turbulence and solar radiation at a really low [computational] costs," King said. "We can do these enhancements basically in real time."
The goal of this new tool is to help determine the technological and economic viability of renewable energy resources, as power from wind and solar plants change in different climate scenarios. That includes anticipating where some resources might increase.
The code is open-source and available to others who might benefit from these higher-resolution images.