Machine Learning for Predicting Vehicle and Power Plant Emissions in the United States
Disciplines
Environmental Engineering | Transportation Engineering
Abstract (300 words maximum)
Machine Learning for Predicting Greenhouse Gas Emissions from Electric Vehicle Charging in the United States
Electric vehicles (EVs) produce significantly fewer emissions compared to traditional vehicles running on fossil fuels. However, the increase in electrical energy consumption of EVs from the power grid, driven by their rising adoption and higher vehicle miles traveled due to lower maintenance and travel costs, results in higher greenhouse gas emissions from power plants. This research develops a machine learning model to predict greenhouse gas emissions from power plants at the state level as EV energy consumption from the power grid increases. The proposed model is built on the Meta Prophet platform, which is specifically designed to capture temporal patterns and seasonality. Trained using the National Renewable Energy Laboratory (NREL) Cambium database, the proposed model provides accurate predictions of CO2, CH4, and N2O emission rates from power plants over the coming months and years under eight distinct pricing, taxation, and development scenarios. We apply the proposed model to predict the increase in emissions of CO2, CH4, and N2O from power plants in response to the growing energy consumption of EVs from the power grid across the United States.
Academic department under which the project should be listed
SPCEET - Civil and Environmental Engineering
Primary Investigator (PI) Name
Dr. Mahyar Amirgholy
Machine Learning for Predicting Vehicle and Power Plant Emissions in the United States
Machine Learning for Predicting Greenhouse Gas Emissions from Electric Vehicle Charging in the United States
Electric vehicles (EVs) produce significantly fewer emissions compared to traditional vehicles running on fossil fuels. However, the increase in electrical energy consumption of EVs from the power grid, driven by their rising adoption and higher vehicle miles traveled due to lower maintenance and travel costs, results in higher greenhouse gas emissions from power plants. This research develops a machine learning model to predict greenhouse gas emissions from power plants at the state level as EV energy consumption from the power grid increases. The proposed model is built on the Meta Prophet platform, which is specifically designed to capture temporal patterns and seasonality. Trained using the National Renewable Energy Laboratory (NREL) Cambium database, the proposed model provides accurate predictions of CO2, CH4, and N2O emission rates from power plants over the coming months and years under eight distinct pricing, taxation, and development scenarios. We apply the proposed model to predict the increase in emissions of CO2, CH4, and N2O from power plants in response to the growing energy consumption of EVs from the power grid across the United States.