Comparative Analysis of ARIMA, LSTM, Prophet, and XGBoost for Energy Consumption Forecasting

Accurate energy consumption forecasting is vital for resource planning, cost optimization, and sustainable energy management. This study presents a comprehensive comparative analysis of four prominent forecasting models: ARIMA, LSTM, Prophet, and XGBoost. Using real-world energy consumption data from AEP, COMED, PJME, and DAYTON regions (2004-2019), we evaluate these models based on forecasting accuracy, robustness, and computational efficiency. Additionally, we explore the impact of preprocessing techniques on performance. Results highlight the superiority of LSTM, achieving an RMSE of 76.5 and MAPE of 8.1%, in capturing nonlinear dependencies and seasonality. Our findings offer practical guidance for stakeholders seeking to enhance energy forecasting systems.