Prediction Model for Electricity Energy Consumption and Pricing Using Xgboosts With L1 Regularization

Abstract

The increasing global demand for electricity, driven by rapid industrialization, urbanization, and digitalization, necessitates more accurate and efficient predictive models for electricity generation and consumption. Traditional statistical techniques often fall short in incorporating the complex environmental factors influencing energy demand. This study presents an advanced machine learning model, leveraging XGBoost with L1 regularization, to improve the accuracy of electricity consumption and generation predictions. By integrating comprehensive datasets, including economic indicators and weather variables, the model addresses the critical gaps in existing forecasting methods. The research involved rigorous data pre-processing, feature engineering, and model validation using metrics such as RMSE, MAPE, and MAE. The model achieves a RMSE of 2.212 and MAPE of 2.393%, indicating good predictive performance. Effective data cleaning, feature engineering, and the application of PCA contributed to this result. The results demonstrate significant improvements in prediction accuracy, highlighting the potential of the proposed model to enhance energy management practices, optimize grid operations, and support the integration of renewable energy sources. This study provides a robust framework for future research and practical applications, contributing to the advancement of sustainable energy systems and informed policy-making. The findings underscore the importance of integrating diverse data sources and sophisticated machine-learning techniques to meet the growing challenges in the energy sector.