The continuous increase in global energy demand has intensified the challenges associated with balancing energy production and consumption. Accurate forecasting of energy consumption is essential for optimizing production schedules, resource allocation, and maintaining grid stability. This paper presents a comparative study of short-term and long-term energy consumption forecasting in the town of Belyounech, Morocco. We employ a combination of statistical and machine learning models, including XGBoost, LightGBM, and SARIMA, as well as ensemble methods to enhance prediction accuracy. The results demonstrate that ensemble methods, particularly those excluding suboptimal models, outperform individual models in terms of Mean Absolute Error (MAE) and Mean Absolute Percentage Error (MAPE). Recommendations for model updating and future research directions are discussed.