Current forecasts are problematic, explained Larochelle at ARAP-E's recent technology summit in Washington, DC, and errors can be anywhere from 3 to 30 percent.
"Bad forecasting leads to firing up a gas plant to pick up the slack," Larochelle said. "When the wind does come, you can't ramp down the coal, gas or nuclear fast enough. This is what we are seeing in Texas and in Europe."
Improving renewable energy forecasting by 10 to 20 percent could result in between $140 million to $975 million in utility savings annually. But ARPA-e officials want to challenge inventors, software engineers and scientists to figure out a way to boost existing forecasting by up to 40 percent.
They want better sensors to monitor solar radiation and atmospheric conditions that generate wind. They also want better computer models to crunch the tons of observational data once it's collected. These models then have to be turned into customized, local forecasts for power producers.
"Right now there's a growing demand for solar forecasting," said Manajit Sengupta, senior scientist at the National Renewable Energy Laboratory in Golden, Colo. Current weather forecasts are being used to generate solar forecasts, but the errors are high. "We can bring them down," Sengupta said.