ML-Enabled Solar PV Electricity Generation Projection for a Large Academic Campus to Reduce Onsite CO₂ Emissions

Abstract

Mitigating CO2 emissions is essential to reduce climate change and its adverse effects on ecosystems. Photovoltaic electricity is 30 times less carbon-intensive than coal-based electricity, making solar PV an attractive option in reducing electricity demand from fossil-fuel-based sources. This study looks into utilizing solar PV electricity production on a large university campus in an effort to reduce CO2 emissions. The study involved investigating 153 buildings on the campus, spanning nine years of data, from 2015 to 2023. The study comprised four key phases. In the first phase, PVWatts gathered data to predict PV-generated energy. This was the foundation for Phase II, where a novel tree-based ensemble learning model was developed to predict monthly PV-generated electricity. The SHAP (SHapley Additive exPlanations) technique was incorporated into the proposed framework to enhance model explainability. Phase III involved calculating historical CO2 emissions based on past energy consumption data, providing a baseline for comparison. A meta-learning algorithm was implemented in Phase IV to project future CO2 emissions post-solar PV installation. This comparison estimated a potential emissions reduction and assessed the university’s progress toward its net-zero emissions goals. The study’s findings suggest that solar PV implementation could reduce the campus’s CO2 footprint by approximately 18% for the studied cluster of buildings, supporting sustainability and cleaner energy use on the campus.

Keywords

• solar PV
• ensemble learning
• carbon emissions forecasting
• net-zero emissions
• university campus
• meta-learning