Researchers improve day-ahead solar forecasting by 13%

Researchers at North Carolina State University have found a way to make next-day solar power forecasts more accurate, which matters more for the grid than for any single home, but it's worth knowing about if you have solar panels or are considering them.

The team tested several forecasting models, including statistical models and artificial neural networks (computer models that learn patterns in data), using weather and power data from two California utilities. No single model worked best everywhere. Instead, combining forecasts from multiple models improved accuracy by as much as 13% compared to the best individual model. One method, which gives more weight to better-performing forecasts, worked best in one utility's territory; a different method, which pulls in weather data from multiple locations, worked best in the other.

The main lesson is that forecasting tools need to be tuned to the specific region where they're used, since local weather patterns and grid conditions vary. Better forecasting helps grid operators plan for how much solar power will be available on a given day, which supports reliability as more homes and businesses add solar. There's no direct action for homeowners here. But if you already have rooftop solar, or use a program that pays you for exporting power to the grid, improvements like this could eventually mean utilities get better at predicting and integrating solar output, which supports the broader push toward more renewable energy on the grid.