The University of Connecticut has partnered with Eversource, the state’s leading energy provider, on a research effort to predict extreme weather events and their effects on energy systems and the aquatic environment.
Dr. Israt Jahan is an environmental engineering student and one of the researchers working on this project with Dr. Marina Astisa, an associate professor of civil and environmental engineering at UConn. Their research focuses on improving the predictability of wind gusts using deep learning with machine learning and artificial intelligence (AI) models.
“We developed a hybrid model that integrates ML. [machine learning] And DL [deep learning] Physics-based numerical weather prediction approach [NWP] “We also applied various explainable AI techniques to demonstrate the factors behind AI model predictions and uncertainties, improving interpretability and transparency,” Jahan said. Currently, I am analyzing high wind events across the continental United States under both historical climate and simulated global warming conditions to assess how extreme wind hazards will evolve in a warming climate. ”
The use of AI was not first proposed in research programs until 2016. Astitha’s research was based on physical modeling. However, with the introduction of AI, many of the challenges of capturing nonlinear relationships and uncertainties in physically-based models have disappeared, according to Astitha.
She finds AI integrated into all areas of research on extreme weather prediction, renewable energy climate assessment, and aquatic ecosystem health.
“We use AI to reduce prediction bias, quantify uncertainty, and fuse multiple data streams to build hybrid physics and ML models that outperform either approach alone,” Astitha said. “And we are increasingly developing trustworthy AI frameworks that emphasize interpretability and reliability, which are essential when predictions impact engineering decisions and emergency operations.”
According to Jahan, this intelligence not only reduced computation time but also reduced prediction error compared to the NWP model.
“Overall, AI expands what is scientifically and operationally possible. AI does not replace physical modeling; it enhances it, fills gaps, and increases reliability,” Astisa said.
center. Photo by Peter Morenus/UConn Photo, UConn Today
This approach from Astitha’s research group integrates physically-based models and AI to improve predictions of extreme weather events.
“I lead the Atmosphere and Air Quality Modeling Group, where we combine physically-based atmospheric models with cutting-edge AI and machine learning to improve predictions of high-impact events such as storms, snowstorms, and tropical cyclones,” Astisa said. “These storms have huge impacts on society, from power outages to infrastructure damage, so there is an urgent need for improved forecasting.”
Astitha said the group is currently developing a next-generation AI framework that combines physical knowledge and machine learning to capture storm dynamics.
In addition to these models, they are also determining the impact of climate change on wind resources and the offshore wind industry by using “high-resolution modeling and downscale climate projections to understand how hubheight winds change over time, which directly informs regional energy planning,” Astisa said.
He also said the group has also built a physics-based ML system for freshwater ecosystems that integrates “meteorology, hydrology, air quality, and fertilizer application data to predict the dynamics of chlorophyll a, phosphorus, and dissolved oxygen in lakes.”
Astitha began working with Eversource Energy at UConn in 2015 prior to the opening of the Eversource Energy Center.
Astisa said the program was created to “bring together academic research and utility expertise to reduce storm-related power outages and improve emergency preparedness.”
Predictive methodologies, data sharing, modeling infrastructure, and translating research innovations into operational tools for use are created in tandem.
The research group led by Astisa is “focused on improving the prediction of extreme weather events using advanced modeling and machine learning approaches,” Jahan said. “My specific role in this collaboration is to develop and validate machine learning (ML) and deep learning (DL) models for wind gust prediction and ensure that they outperform benchmark numerical weather prediction (NWP) models.”
Astisa and Jahan aren’t the only ones researching weather forecasting. They say a second research group is running an outage prediction model that inputs various weather variables to determine the number of power outages that could occur within Eversource’s service area due to future storms.
“The overall goal is to provide Eversource with more accurate outage predictions by improving extreme weather forecasts to support operational preparedness and outage recovery efforts,” Jahan said.
According to Asisa, this long-standing partnership has grown not only through UConn’s research programs, but also through support from outside organizations.
“This research also catalyzed broader support from NSF’s IUCRC program and DOE, allowing us to expand our research into climate resilience, renewable energy, and next-generation AI systems,” Asisa said.
