Smart Home Energy Management Using Deep Q-Learning Networks for Cost Reduction Under ToD Tariff

Smart home energy management plays a vital role in reducing electricity costs and enhancing grid efficiency, especially under Time-of-Day (ToD) tariff schemes. This paper presents a comparative analysis between the classical Reinforcement Q-Learning (RLQ) and a Deep Q-Learning Network (DQLN) approach for intelligent appliance scheduling in residential settings. The objective is to minimize electricity costs while respecting user-defined operational constraints. Simulation results over a 24-hour horizon reveal that the proposed DQLN approach achieves a 17.76% reduction in energy cost. Moreover, DQLN demonstrates faster convergence, better load shaping, and higher adaptability to RTP fluctuations. It effectively shifts appliance loads to low-tariff periods, reduces peak demand, and maintains user comfort, thereby supporting demand-side management. The findings establish DQLN as a scalable and cost-effective solution for smart home automation. This study also outlines potential extensions involving renewable energy integration, multi-objective optimization, and edge deployment for real-time operation.