As wildfires grow in frequency and intensity across the United States, more Americans are becoming familiar with smoke-filled days. These hazy, yellow-tinted days not only impact air quality and human health; they can also hinder energy production as smoke blocks the sun from shining onto solar panels. Now, a new tool may help us better predict how wildfires impact the nation’s solar production. 

Using artificial intelligence (AI), researchers at Cornell University have developed a machine learning model that can forecast the impact of wildfire smoke on solar energy generation. The team believes the model is more accurate than current prediction methods and could benefit solar system operators in areas impacted by smoke. “If you don’t have a good forecast, then you have to rely on your so-called reserve generators, which are very costly,” stated Max Zhang, an engineering professor at Cornell and the project lead.

Zhang recognized the threat wildfire smoke poses to solar output during the summer of 2023, when the northeastern United States was blanketed in smoke from Canadian fires. Solar production in the area dipped, leaving some areas unprepared. New York is one example. The New York Independent System Operator (NYISO) monitors output and coordinates how the state’s power grid system will work; they significantly overpredicted what solar output would be during the 2023 fires. 

“There are day-ahead markets and real-time markets. They need a forecast of the energy production to balance supply and demand,” explained Zhang. “Either overprediction or underprediction is not good, especially overprediction.”

>>> READ: Wildfire-Proof Homes Can Be the Next American Dream

The Cornell team built their model using public domain products from the new High-Resolution Rapid Refresh Smoke (HRRR-Smoke) weather forecasting system, developed by the National Oceanic and Atmospheric Administration. HRRR-Smoke analyzes severe wildfire periods and makes predictions on factors like smoke density. After building the model, the team tested it on hourly solar data collected by the New York State Energy Research and Development Authority during wildfire periods. Cornell’s model outperformed NYISO’s predictions. 

The model is now operational and distinguishes itself from other prediction tools by providing forecasts on an hourly basis, instead of using daily averages. Unlike models that focus solely on real-time assessments or post-event analysis, Cornell’s model also offers forecasts, which are the data that solar operators need to make adjustments during wildfire periods. 

“This is just the start. We are improving the model while creating pathways for adoption by system operators,” says Zhang. “The better the forecast, the more reliable the power system.”

Solar power capacity is increasing across the United States, but so are North American wildfires. More accurate predictions of how smoke will impact solar production can help grid operators make more informed decisions, reduce costs, and maintain power system reliability. 

The views and opinions expressed are those of the author’s and do not necessarily reflect the official policy or position of C3.