From Static to Dynamic: Improving Scope 2 Emission Factor Predictions with ARIMA Data-Driven Models

This study examines the efficacy of Autoregressive Integrated Moving Average (ARIMA) models in forecasting Scope 2 emission factors, focusing on the temporal dynamics of Consumption-based Hourly Emissions Factors (CHEFs). The research method involves selecting ARIMA models based on the lowest Akaike Information Criterion (AIC) values. Once identified, the best-performing model for each timestep underwent further evaluation for predictive accuracy using a suite of metrics. The analysis demonstrated that ARIMA models, optimized with consistent differencing and maximal autoregressive and moving average terms, excel in emissions trend forecasting. Notably, models forecasting 1-hour and 3-hour horizons achieved high predictive accuracy, suggesting their potential in operational planning despite limitations for demand response applications due to the short forecasting horizon. The strategic value of the 24-hour forecasts emerges as a pivotal tool for building management, offering a critical balance between accuracy and practical applicability, thereby facilitating more effective long-term operational adjustments. The study concludes with the recommendation to enhance ARIMA models' precision by incorporating external factors like temperature and to consider advanced machine learning techniques for future research. These findings offer insights into improving emission factors forecasting, thereby aiding dynamic demand response measures which seek to lower a building’s operational GHG emissions.