Cornell University researchers are using a subset of artificial intelligence to predict threats to solar-energy production. WAMC's Capital Region Bureau Chief Dave Lucas spoke with project leader Max Zhang.
In June 2023, smoke from the Canadian wildfires significantly reduced power solar panels were able to produce.
Zhang, a professor of engineering at Cornell University, observed that new installations of photovoltaic panels coming online coincided with wildfires becoming more frequent.
According to Zhang, future return of wildfire smoke drifting in from Canada will adversely affect residents and businesses use of solar panels to generate electricity, and they should be warned so they can prepare for such an event.
"During the 2023 wildfire events, we saw anywhere between 10 to 20% reductions of in the PV outputs across states. That's when you had to pay attention, because that's gonna have impact," said Zhang, who led a team of researchers to create a machine learning-based model that can forecast, with greater accuracy than current methods, the impact incoming wildfire conditions will have on solar electricity generation, which will enable system operators to keep costs down by better matching supply and demand, while also alerting individual solar panel users.
It's not just wildfire smoke. Zhang says any type of concentrated smoke in the air or pollution in the air is enough to reduce the efficiency of the type of solar panels installed atop homes and businesses and on solar farms.
"We are actually seeing the impact of so called the particulates or aerosol or droplets," Zhang said. "There's different we call it on the impact of PV solar photovoltaic performances. So, for example, when you we will have, when we have a cloudy day the output from your solar panel decreases, right? So the cloud essentially consists of many small droplets, right? And also, you know, depending on the area you you are, you know, you could have pollen, you know, right now is the pollen season, right? And when those tiny particles deposit on your panels, that also is going to affect your performance. So we're seeing the impact of those aerosols particles all the time, but just not to the level we are paying attention."
Zhang says anticipating a disruptive event like wildfire smoke pollution enables system operators to prepare ahead of time to ensure a consistent flow of electricity.
"To balance supply and demand. So there's your basic action you can take from demand side, there's action you can take from the supply side, but with the goal of to balance it. So that's really the main goal for the system operators, so the better forecast, like our tool can provide potentially, right? So basically gave system operators, basically equipped them with better information to take actions that can achieve that goal," said Zhang.
Zhang's machine-learning model incorporates public domain data products from the National Oceanic and Atmospheric Administration’s new High-Resolution Rapid Refresh Smoke weather forecasting system, which includes predictions of aerosol impacts and smoke mass density during severe wildfire periods.
"Our tool, you know, to the best of our knowledge is the is the most ready to be adopted. The main reason is that, we rely on the operational forecast, weather forecast products. One of the highlights in our work is we are the first adopter of this newly available data product which predicts not just the weather, but also predicts the amount of smoke in the air. So, because there's a forecasting product, so everything we are doing is under a forecasting mode, so that we are not, you know, based on the hindsight, right? We are not based on the historical data. We are based on the forecast," he said.
Zhang believes climate change has a hand in fostering unprecedented events like wildfire smoke invading New York. His team has issued a paper extolling the benefits of government-operated weather agencies.
"I really hope that we, you know, as a country, that we can maintain and improve those weather forecasting systems, right? Because it means a lot, right? So we already show the benefits on the power prediction side, but there are many other, many other benefits to other sectors of society," said Zhang, whose hopes are threatened by federal cuts to staffing at the National Weather Service by President Trump and the Department of Government Efficiency.
