This study presents an artificial intelligence-based system for predicting energy consumption in homes with 3-phase power installations. Using time-series data stored in a NoSQL database and leveraging deep learning techniques on GPUs, the system achieves high-precision forecasts of future electricity consumption. Several contextual factors influencing energy usage, such as time of day, day of the week, holidays, weather conditions, and occupancy status, are incorporated into the predictive model. To facilitate data collection, a comprehensive IoT-based architecture has been developed, integrating high-performance, low-bandwidth technologies. The system includes a communication protocol based on subscription messaging, information retrieval via RESTful services, and data storage in JSON format. Additionally, an Android application has been designed to calibrate the energy consumption of household appliances, enabling the identification of specific contributors to energy fluctuations. This feature enhances prediction accuracy and opens possibilities for feedback mechanisms that optimize electricity usage. The proposed system is designed for scalability, making it adaptable to residential communities or larger energy networks. Technologies such as MongoDB and MQTT were selected for their big data capabilities. Although predictive experiments were conducted using a limited dataset over a short time frame, the model demonstrated remarkable accuracy. These findings highlight the potential of AI and IoT integration in improving residential energy efficiency and optimizing electricity consumption.