This study presents an innovative approach for predicting energy consumption by combining autoregressive (AR) models and differential evolution (DE) algorithms. The research addresses the growing need for accurate energy consumption prediction due to increasing demand and environmental concerns. AR models are known for their ability to capture temporal dependencies in time series, while DE algorithms optimize these models by finding the optimal parameters. Residential energy consumption data from an electrical feeder were used, comprising 100 observations for estimation and 30 for validation, with predictions made for a horizon of 24 samples. The process focused on data analysis and visualization to identify trends and patterns, without applying differencing to ensure stationarity. Five variants of the DE algorithm were evaluated, with the DE/best/1/bin strategy standing out for its rapid convergence and accuracy in minimizing the root mean square error (RMSE). The results show that combining AR models with DE significantly improves prediction accuracy, providing a robust framework for optimizing predictive models in the energy sector. This approach enables better planning and management of energy resources, reducing costs and enhancing sustainability. Additionally, the potential integration of hybrid techniques for future research is highlighted, which could offer further improvements in prediction accuracy and stability.