13. Developing a model for power load demand forecasting using ensemble learning techniques

Tóm tắt

Accurate forecasting of enterprise energy consumption plays a vital role in optimizing electricity production and distribution. This study proposes a time series forecasting model that applies powerful ensemble learning algorithms such as XGBoost and Random Forest to predict daily electricity demand based on historical consumption data. The dataset used for training, testing, and model development was collected from an industrial zone. In addition, we evaluate and compare the performance of models using several commonly adopted data normalization techniques. The proposed model demonstrates high adaptability to volatile data caused by noise and missing observations. Our model outperforms existing approaches in terms of MAE and R² metrics, achieving 0.027 and 98%, respectively, with the XGBoost algorithm. This modeling framework shows strong potential for energy demand management and investment cost optimization, as well as for the operation of smart grids. It is particularly effective in supporting decision-making under complex operating conditions, contributing to improved reliability and efficiency of modern power systems.