In smart grids, the Demand Response (DR) strategy constitutes a vital necessity for the electricity management system. DR benefits from the two-way communication between utilities and consumers within the smart grid. However, implementing an efficient DR model basically relies on the real-time adjustments of the load consumption pattern, which requires access to accurate load consumption data. In this paper, a Deep Learning (DL) predictive model is proposed to accurately predict the hourly load consumption. In addition, a DR scheme is illustrated to reduce the peak load demand and avoid the energy deficit. Compared with the state-of-the-art techniques in residential load prediction, the proposed DL predictive model and DR scheme outperforms Linear Regression by 59.82%, Tree Regression by 52%, Support Vector Regression by 57.76%, and Ensembled Boosted Trees by 59.43% in terms of the Root Mean Square Errors.