Pollutant emissions of conventional energy generators with increased renewable energy sources: A 2030 New York case study

This repository contains the custom code used in “Pollutant emissions of conventional energy generators with increased renewable energy sources: A 2030 New York case study” (Scientific Reports, under review). The code implements the emissions-forecasting workflow that combines historical generator operations and emissions reporting (hourly 2019 NYISO generator data) with forward-looking 2030 operational conditions generated through Security-Constrained Unit Commitment (SCUC) and Security-Constrained Economic Dispatch (SCED) simulations. The workflow includes: (i) data preparation and consistency checks; (ii) feature engineering capturing operational variables, generator characteristics, spatial attributes, and temporal seasonality (including lagged features and Fourier components); (iii) model training and evaluation for CO2, NOX, and SO2 emission rates using regularized regression baselines (Lasso/Ridge/Elastic Net) and an ensemble gradient-boosting approach (XGBoost); and (iv) generation of 2030 emissions forecasts using simulated SCUC/SCED operating outputs under renewable-generation and load scenario variability. Model interpretation utilities (e.g., SHAP and LIME) are included where applicable. Access and data restrictions: The Zenodo record is public but the files are restricted. Certain input datasets used in the study (notably emissions data from the U.S. EPA Clean Air Markets Division and PERFORM-derived scenario inputs) are subject to licensing and/or third-party access restrictions and cannot be redistributed publicly. The code is archived here to preserve the exact version used for the study and is available from the corresponding author upon reasonable request.